Linking genomes to metabolites in antimicrobial discovery: integrative metagenomic and metabolomic approaches to combat antimicrobial resistance

1 Introduction

The past few decades have witnessed a substantial rise in AMR, significantly increasing the global burden of infectious diseases. The re-emergence of previously controlled pathogens has further complicated disease management and surveillance. This resurgence, coupled with the rapid evolution of AMR, has significantly reduced the effectiveness of existing therapeutic strategies (Laprise, 2023). Environmental changes, increased human mobility, and the continuous evolution of microbial pathogens further contribute to the persistence and global spread of infectious diseases (El-Sayed et al., 2026). WHO has identified AMR as one of the most serious global public health threats of the 21st century, underscoring the urgent need for innovative and sustainable approaches to drug discovery and disease management (Salam et al., 2023). AMR was associated with approximately 4.95 million deaths globally in 2019, including 1.27 million directly attributable deaths (Murray et al., 2022). By 2050, AMR is projected to cause nearly 10 million deaths annually worldwide (Naghavi et al., 2024). The stagnation of the antimicrobial drug discovery pipeline and the declining efficacy of frontline antibiotics necessitate innovative and integrative strategies for identifying novel antimicrobial agents (Reid et al., 2019). Underexplored ecological niches and medicinal plants represent important reservoirs of chemically diverse secondary metabolites with antimicrobial potential (Flores et al., 2025). Natural products have historically contributed to the majority of clinically used antimicrobial agents. However, conventional antimicrobial discovery pipelines have shown declining productivity over the past few decades (Caesar et al., 2021; Schneider, 2021; Sakyi et al., 2025). High rediscovery rates of known compounds, challenges associated with culturing diverse microbial communities, and the rapid emergence of resistance to existing drugs have collectively limited the success of traditional screening approaches (Li, Qiu and Peng, 2026). Consequently, there is a pressing need to explore untapped biological resources and employ advanced molecular tools to identify novel bioactive compounds with unique mechanisms of action (Nguyen et al., 2024; ElSaadony et al., 2025). To overcome these limitations (Roy et al., 2022; Zandavar and Afshari Babazad, 2023), recent advances in omics technologies, particularly metagenomics and metabolomics, have revolutionized natural product discovery. Metagenomics enables access to the biosynthetic potential of uncultured microbial communities, while metabolomics provides direct insights into the chemical compounds produced under specific conditions (Miethke et al., 2021; Aminov, 2022). Their integration enables linking genetic information to metabolite production, providing a powerful framework for identifying novel antimicrobial agents.

1.1 The global antimicrobial resistance crisis and unmet therapeutic needs

AMR is a major global health challenge that threatens the effective treatment of infectious diseases. The widespread and often indiscriminate use of antimicrobial agents in human medicine, agriculture, veterinary practice, and aquaculture has accelerated the development of resistance among pathogens. Over-prescription, self-medication, incomplete treatment courses, and the use of antibiotics as growth promoters in livestock have collectively created strong selective pressure, enabling resistant strains to proliferate. As a result, infections caused by multidrug-resistant bacteria, viruses, fungi, and parasites are becoming increasingly difficult to treat, leading to prolonged illness, higher healthcare costs, and increased mortality (Oliveira et al., 2024). Several infections, including tuberculosis, pneumonia, urinary tract infections, and wound infections, are re-emerging due to multidrug-resistant pathogens such as Staphylococcus aureus, Mycobacterium tuberculosis, Enterobacteriaceae, and Pseudomonas aeruginosa (Muteeb et al., 2025). Global travel, trade and hospital environments further facilitate the rapid dissemination of resistant pathogens. Community-acquired resistance is also rising, making infections harder to treat even outside clinical settings (Cella et al., 2023). The burden of AMR is not only medical but also socio-economic. Resistant infections increase healthcare costs and strain healthcare systems, particularly in low- and middle-income countries, where access to advanced diagnostic facilities and second-line antibiotics may be limited. The impact of AMR is even more severe. The loss of effective antimicrobial therapies threatens the sustainability of healthcare infrastructures and may reverse many gains achieved in public health over recent decades (Nazir et al., 2025). In addition to affecting the treatment of common infections, antimicrobial resistance also jeopardizes modern medical practices that rely heavily on effective prophylactic and therapeutic antibiotics. Surgical procedures, organ transplantation, cancer chemotherapy, and neonatal care depend on the ability to prevent and treat infections efficiently. The rising failure of antimicrobial agents increases the risk of complications, post-operative infections, and treatment-related morbidity, thereby threatening the success of life-saving medical interventions (Teillant et al., 2015). Despite the growing severity of the AMR crisis, the development of new antimicrobial drugs has slowed dramatically. Over the past few decades, pharmaceutical research has shifted its focus toward chronic and lifestyle-related diseases, which offer higher financial returns compared to antibiotics that are used for short durations. The scientific challenges associated with discovering new antibiotics, coupled with regulatory hurdles and limited economic incentives, have resulted in a declining pipeline of novel antimicrobial agents. Moreover, many of the antibiotics introduced in recent years are modifications of existing drug classes rather than entirely new chemical entities. While such modifications may temporarily improve efficacy, pathogens often develop resistance to them within a relatively short period. This cycle of resistance development followed by incremental drug modification is neither sustainable nor sufficient to address the magnitude of the problem (Muteeb et al., 2025). A major biological challenge is the diversity of resistance mechanisms, including enzymatic degradation, target modification, and efflux-mediated resistance. The free exchange of genetic material among bacterial species through horizontal gene transfer allows resistance traits to spread rapidly within microbial populations. The genetic adaptability of pathogens makes the development of long-lasting antimicrobial solutions particularly difficult. The emergence of pan-resistant strains represents a particularly alarming development. Infections caused by such strains leave clinicians with extremely limited or no therapeutic options, often resulting in increased mortality (Belay et al., 2024). Limited availability of rapid diagnostic tools further hinders accurate identification of resistant pathogens and guides targeted treatment decisions. Addressing this requires a shift from broad-spectrum screening toward mechanism-oriented discovery that targets conserved microbial vulnerabilities.

1.2 Importance of discovering new antimicrobial compounds

Natural products remain a vital source of novel antimicrobial agents, with many clinically important antibiotics historically derived from soil microorganisms and medicinal plants. However, a large proportion of microbial diversity remains uncultured, and the chemical diversity of plant-derived metabolites is still incompletely explored. These untapped resources hold significant potential for discovering new bioactive compounds capable of targeting resistant pathogens through novel mechanisms (Yarmolinsky et al., 2024). Medicinal plants, in particular, represent an underexploited reservoir of antimicrobial agents. Traditionally used in treating infectious diseases, these plants produce diverse secondary metabolites that function as natural defense compounds. Many exhibit antimicrobial, antioxidant, anti-inflammatory, and resistance-modulating activities, making them promising candidates for drug discovery (Vaou et al., 2021). However, translating traditional knowledge into therapeutics requires systematic approaches, including accurate species identification, phytochemical profiling, and rigorous bioactivity evaluation (Aćimović et al., 2026).

1.3 Secondary metabolites as chemical solutions to antimicrobial resistance

The accelerating spread of AMR has necessitated a paradigm shift in the search for new antimicrobial agents. As resistance kept increasing against the clinical utility of existing antimicrobials, attention has increasingly turned toward structurally diverse natural products, particularly secondary metabolites, as reservoirs of novel chemical scaffolds and resistance-modulating agents (Vaou et al., 2021). Unlike primary metabolites, which are directly involved in growth and reproduction, secondary metabolites are specialized molecules that mediate ecological interactions such as competition, defense, symbiosis, and signaling. Their structural complexity, stereochemical richness, and evolutionary optimization make them uniquely suited for pharmaceutical exploration (Jayawardena et al., 2025). Secondary metabolites encompass a broad spectrum of chemical classes, including alkaloids, terpenoids, polyketides, non-ribosomal peptides, phenolics, and glycosides. Plant-based secondary compounds such as Artemisinin from Artemisia annua and Paclitaxel from Taxus brevifolia demonstrate how plant secondary metabolites can lead to highly effective therapeutic drugs. Similarly, many clinically significant antibiotics, such as β-lactams, aminoglycosides, tetracyclines, and macrolides etc., have originated from microbial secondary metabolism. The continued relevance of these compounds underscores the importance of natural biosynthetic systems as blueprints for drug development (Muteeb et al., 2025).

One of the most compelling attributes of secondary metabolites lies in their mechanistic diversity. Rather than relying solely on bactericidal activity through inhibition of essential metabolic processes, many plant-and microbe-derived compounds exhibit multi-target or adjuvant effects. These include disruption of microbial membranes, inhibition of efflux pumps, attenuation of virulence factor production, interference with quorum sensing, and suppression of biofilm formation (Gorlenko et al., 2020). The structural complexity of secondary metabolites often exceeds that achievable through purely synthetic combinatorial chemistry. The presence of multiple chiral centres, macrocyclic rings, heterocyclic moieties, and unusual functional groups contributes to high target specificity and potent biological activity. Such structural diversity enables interactions with a wide range of molecular targets, including ribosomal subunits, DNA gyrase, topoisomerases, cell wall biosynthetic enzymes, and membrane components, e.g., certain phenolic compounds, flavonoids, and terpenoids have demonstrated the ability to destabilize biofilm architecture or inhibit its formation, thereby restoring susceptibility to conventional antibiotics (Mulat et al., 2025). Similarly, alkaloids and polyphenols have been reported to inhibit efflux pump systems, which are central to multidrug resistance in both Gram-positive and Gram-negative bacteria (Duda-Madej et al., 2025). Importantly, many natural products act through mechanisms distinct from those of existing antibiotics, thereby offering opportunities to circumvent established resistance pathways (Aware et al., 2022). Secondary metabolites also have the potential of acting as antibiotic potentiators. These compounds may not exhibit strong standalone antimicrobial activity but can enhance the efficacy of existing drugs through synergistic interactions. For instance, membrane-permeabilizing phytochemicals can facilitate antibiotic entry into resistant bacteria, while efflux inhibitors reduce intracellular drug expulsion (Zai et al., 2025). Such combinatorial approaches reduce the likelihood of resistance development and extend the therapeutic lifespan of current antibiotics. The ecological roles of secondary metabolites provide insight into their antimicrobial potential. In natural environments, microorganisms and plants are engaged in continuous chemical warfare, producing bioactive molecules to outcompete rivals and defend against pathogens. This evolutionary arms race has driven the selection of compounds with potent and highly specific biological activities.

2 Metagenomics and metabolomics: complementary omics for antimicrobial discovery

2.1 Principles and scope of metagenomics in natural product discovery

The Global escalation of AMR has transformed once manageable infections into existential threats, rendering traditional therapeutics increasingly obsolete. Within this crisis, Natural Products (NPs), small molecules synthesized by living organisms, remain the undisputed gold standard for drug discovery. Unlike synthetic libraries, NPs are evolutionarily pre-validated, as they have been refined over a million years to penetrate biological membranes and bind to essential microbial targets with high specificity (Atanasov et al., 2021; Butler and Capon, 2025). Historically, over 70% of the antibiotics currently in clinical use, including the β-lactams, tetracyclines, and macrolides, are derived from or inspired by microbial and plant secondary metabolites (Zhang et al., 2025). The Golden Age of Antibiotics was predicated on Waksman’s platform: culturing soil microbes in the lab, which has now reached a point of diminishing returns. The Great Plate Count Anomaly reveals that >99% of the microbial world, often termed as microbial dark matter, remains unculturable at standard conditions, leaving their vast chemical potential silent and unexplored (Rinke et al., 2013; Lloyd et al., 2018). Amid this backdrop, metagenomics has emerged as the definitive solution to this limitation, providing a culture-independent gateway to the collective genetic blueprint of complex microbial communities. The fundamental principle of metagenomics in NPs discovery lies in the direct extraction and analysis of total environmental DNA (eDNA). By detouring the need for isolation and cultivation, researchers can access the Biosynthetic Gene Clusters (BGCs), functionally linked groups of genes that encode the enzymatic machinery required to produce secondary metabolites directly from the source (Handelsman, 2004). This approach is particularly potent because many potent antimicrobials are secondary metabolites that microorganisms only produce under specific ecological stressors, not replicated in a Petri dish.

The scope of metagenomics in this field is defined by two primary technological pillars: highthroughput sequencing and advanced computational bioinformatics. The transition from short-read sequencing (e.g., Illumina) to long-read technologies (e.g., Pacific Biosciences and Oxford Nanopore) has revolutionized the assembly of complex, repetitive BGCs, such as those encoding Non-Ribosomal Peptide Synthetases (NRPS) and Polyketide Synthases (PKS) (Santana-Pereira et al., 2020). Furthermore, the development of predictive computational tools like antiSMASH and BiG-SCAPE allows for the in-silico identification of cryptic or silent clusters that possess no known homologues, thereby expanding the chemical search space into entirely novel scaffolds (Blin et al., 2019; Navarro-Muñoz et al., 2020; Bağcı et al., 2025).

Beyond mere identification, the scope of metagenomics extends to the ecological context. It allows for the exploration of niche-specific chemistry, where the selective pressures of extreme heat, salinity, or symbiotic competition drive the evolution of unique chemical molecules or secondary metabolites. These ecosystems are enriched in diverse and evolutionarily distinct BGC architectures, suggesting a high probability of structurally novel antimicrobial scaffolds. Importantly, metagenomic surveys consistently reveal that the majority of predicted BGCs do not correspond to known compounds, underscoring the vast untapped biosynthetic reservoir available for antibiotic discovery (Chevrette et al., 2022). By mapping these genetic potentials, metagenomics reveals the genetic blueprint for novel bioactive compounds, while metabolomics complements it by confirming the production and functional expression of these metabolites, thereby linking biosynthetic gene clusters to antimicrobial activity.

2.2 Functional and sequence-based metagenomic approaches

The methodological framework for exploring metagenomes in antimicrobial discovery has traditionally been divided into sequence (genotype-first) and functional-based (phenotype-first) strategies. Recently, the boundary between these two has been increasingly blurred by AI-driven predictive modelling, yet they remain the two primary lenses through which we view the microbial dark matter.

2.2.1 Sequence-based metagenomics

2.2.1.1 Mining the in-silico reservoir

Sequence-based approaches leverage the conserved nature of biosynthetic machinery. Historically, this relied on identifying signature domains such as the Ketosynthase (KS) domains in PKS clusters and acyl domains (A) in NRPS clusters using Hidden Markov Models (HMMs). However, in recent times, the reliance on simple homology has transitioned toward deep learning-based BGC identification. Modern tools like antiSMASH 7.0+ and the transformer-based BGC-Prophet have moved beyond identifying known clusters to recognizing distant or divergent BGCs that share minimal sequence identity with characterized genes (Blin et al., 2023; Lai et al., 2023). This is critical for overcoming the homology trap, where traditional methods only rediscover variations of known drugs. Furthermore, Phylogeny Guided Discovery now allows researchers to prioritize BGCs from rare taxa that occupy unique evolutionary branches, statistically increasing the likelihood of discovering novel chemical scaffolds (Milshteyn et al., 2018).

2.2.2 Functional metagenomics

2.2.2.1 Capturing the unpredictable

While sequence-based approaches are limited by what we already know (or can predict), functional metagenomics remains the ultimate tool for discovering truly unprecedented antimicrobial mechanisms. This involves the bottom-up approach of cloning large fragments of environmental DNA (eDNA) into high-capacity vectors (e.g., BACs or cosmids) and expressing them in a surrogate host, such as Escherichia coli or Streptomyces coelicolor. The power of functional metagenomics lies in its mechanism-agnostic nature: it detects activity, regardless of whether the gene looks like a classic antibiotic producer. This approach has led to the discovery of entirely new classes of inhibitors, such as the malacidins, which were identified through calcium-dependent antibiotic screening (Hover et al., 2018). However, the bottleneck of functional metagenomics, the failure of environmental genes to express in foreign hosts, remains a challenge. In 2026, this is being addressed through Predictive Heterologous Expression (PHE), where AI models analyze the GC content, codon usage, and promoter compatibility of a metagenomic fragment to select the optimal surrogate host, thereby increasing the hit rate of functional screens by over 40% (Ravi et al., 2015). The antimicrobials already discovered or in the pipeline through function or sequence-based screening are summarized in Table 1.

TABLE 1. Representative antimicrobial leads and specialized metabolites discovered via culture independent metagenomic workflows, along with their sources, biological activities and current development status.

2.2.3 The evolving 1% rule and culturomics-metagenomics synergy in recent times

It is important to contextualize microbial discovery within the evolving understanding of microbial cultivability. The traditional “Great Plate Anomaly”, which posited a rigid 1% cultivability rule, is being redefined in the era of recent high-throughput culturomics. By utilizing large-scale diversification of culture media, extended incubation periods, and metabolic cross-feeding strategies, researchers have successfully increased the cultivable fraction of the human and environmental microbiome to 30–50% in specific biomes (Staley and Konopka, 1985; Lagier et al., 2018). This nutritional language, deciphering the specific metabolic requirements and signaling molecules (such as siderophores or quorum-sensing peptides) needed for growth, acts as a vital complement to metagenomic mining. While metagenomics provides the blueprint of biosynthetic potential, culturomics enables the biological validation by providing live isolates for secondary metabolite induction. To overcome the limitations of conventional cultivation-dependent microbiology, contemporary antimicrobial discovery increasingly integrates culturomics and culture-independent “microbial dark matter” approaches. While culturomics expands the cultivable fraction of microbial diversity through high-throughput and condition-optimized cultivation strategies, metagenomic dark matter approaches provide direct access to uncultured microbial genomes and cryptic biosynthetic potential. As illustrated in Figure 1, these complementary methodologies collectively broaden access to hidden chemical diversity while also presenting distinct experimental and computational limitations. In the recent discovery paradigm, this is often termed Targeted Culturomics, where metagenomic data identifies a high-value BGC, and culturomics is then used to design specific media that mimic the organism’s natural niche, effectively fishing out the producer for laboratory study (Li et al., 2024; Armetta et al., 2025).

Nevertheless, the 99% figure still holds therapeutic weight for complex ecological niches like those living in the extreme niches, deep-sea sludges or complex plant endosphere discussed in this review. In these environments, over 85% of the metabolic potential remains locked within dark genes, sequences that lack homology to known compounds and cannot be assigned a function through conventional bioinformatics (Rodríguez del Río et al., 2022; Pavlopoulos et al., 2023). While culturomics expands the reachable fraction, metagenomics remains the primary lens for exploring the truly novel chemical space of the Dark Proteome. The integration of these complementary approaches thus offers a comprehensive framework for microbial exploration.

2.3 Metabolomics platforms for secondary metabolite profiling

While the metagenomics provide blueprint of microbial potential by identifying biosynthetic gene clusters, metabolomics serves as the functional validation. It captures the chemical reality of an environment, bridging the gap between genetic possibility and actual molecular production. Metabolomics is the comprehensive, high-throughput analysis of low-molecular-weight metabolites (typically <1,500 Da) produced by an organism under specific physiological or environmental conditions. These metabolites represent the final downstream products of gene expression and enzymatic activity, thereby providing a direct snapshot of cellular phenotype and biochemical functionality (Johnson et al., 2016). The metabolome refers to the complete qualitative and quantitative repertoire of low-molecular-weight molecules (metabolites) present within a cell, tissue, or organism at a given time. These metabolites include intermediates and end products of metabolic pathways that participate in fundamental biochemical processes essential for cellular survival. They encompass compounds involved in energy production, biosynthesis, signaling, redox balance, and stress responses, all of which collectively support cellular maintenance, growth, adaptation, and normal physiological function. Unlike the genome or proteome, which represent relatively static blueprints, the metabolome is highly dynamic and responsive to environmental fluctuations, nutrient availability, developmental stage, and stress conditions. As such, it provides a direct functional readout of cellular activity and phenotype. This dynamic nature makes the metabolome a powerful indicator of physiological state and a valuable resource for understanding metabolic regulation, disease mechanisms, and the production of bioactive secondary metabolites, including those with antimicrobial properties(Harrigan and Goodacre, 2003). In the context of antimicrobial discovery, metabolomics is particularly powerful because most clinically relevant antibiotics are themselves microbial secondary metabolites or their derivatives. The escalating AMR crisis has prompted a shift from traditional drug discovery methods, as bioassay-guided fractionation frequently leads to the rediscovery of known compounds, known as the dereplication bottleneck (Gaudêncio et al., 2015). Metabolomics actually overcome this limitation by enabling global metabolic profiling, allowing early discrimination between novel and known chemical entities based on mass spectral signatures and molecular features. Importantly, metabolomics captures chemical diversity independent of prior genomic knowledge, making it indispensable for studying uncultivable or poorly characterized microorganisms, such as those derived from environmental and host-associated metagenomes. When integrated with antimicrobial bioassays, metabolomics facilitates the correlation of specific metabolite features with observed bioactivity, accelerating lead identification and reducing time-to-discovery (Doroghazi et al., 2014). Targeted and Untargeted are the two major approaches of metabolomics (Lelli et al., 2021). Untargeted metabolomics mainly focuses on capturing a comprehensive overview of the metabolome, enabling the detection of a wide range of metabolites without prior bias. However, this broad coverage often comes at the expense of complete structural characterization. In contrast, targeted metabolomics is designed for the precise and quantitative measurement of predefined metabolites. This approach offers greater analytical sensitivity and improved structural confirmation for the selected compounds, making it particularly suitable for hypothesis-driven studies.

Recent advances in metabolomics have led to the emergence of pseudotargeted metabolomics, a hybrid analytical strategy that combines the extensive metabolite coverage of untargeted profiling with the sensitivity, reproducibility, and quantitative precision of targeted approaches (Cajka and Fiehn, 2016). In this workflow, metabolite features are initially identified through untargeted LC–MS/MS analysis and subsequently incorporated into multiple reaction monitoring (MRM)-based targeted assays using triple quadrupole mass spectrometry, enabling the simultaneous monitoring of numerous metabolites with improved analytical stability and quantitative reliability (Xu et al., 2019). Unlike conventional targeted metabolomics, pseudo-targeted approaches do not require authentic standards for all metabolites during the preliminary screening stage, thereby expanding metabolite coverage while maintaining high throughput and reproducibility.

Owing to these advantages, pseudo-targeted metabolomics has become increasingly valuable in food metabolomics, natural product research, microbial metabolomics, and antimicrobial discovery studies involving chemically complex and partially uncharacterized metabolite mixtures. Its application is particularly relevant in studies involving medicinal plants and endophytes, where metabolite pools often contain numerous low-abundance or previously unidentified compounds. By enabling reproducible large-scale metabolite profiling, pseudo-targeted workflows facilitate the identification of metabolites associated with antimicrobial activity, environmental adaptation, and stress responses. Furthermore, the integration of pseudo-targeted metabolomics with molecular networking and in silico annotation tools enhances the prioritization and identification of novel bioactive metabolites from complex biological matrices (Dührkop et al., 2020). As metabolomics continues to evolve toward high-throughput and systems-level applications, pseudo-targeted strategies are expected to play an increasingly important role in bridging exploratory metabolite discovery with robust quantitative validation (Nothias et al., 2020).

2.4 Analytical techniques involved in metabolomics

Despite the availability of diverse protocols and well-established methodologies for metabolomics investigations, the comprehensive detection and accurate identification of the entire metabolome remain highly challenging. This limitation largely stems from the intrinsic complexity of the metabolome, which consists of a vast array of chemically diverse metabolites with distinct physicochemical properties, including differences in polarity, volatility, stability, and molecular weight. Such heterogeneity makes it difficult to simultaneously extract, separate, and characterize all metabolites using a single analytical technique. Consequently, metabolomics studies typically rely on the integration of multiple complementary analytical platforms to achieve broader metabolite coverage and more reliable metabolite identification. Commonly employed techniques include mass spectrometry (MS) coupled with separation methods such as liquid chromatography (LC) or gas chromatography (GC), direct infusion mass spectrometry (DIMS), as well as spectroscopic approaches including nuclear magnetic resonance (NMR) and Raman spectroscopy. The combination of these analytical platforms enhances detection sensitivity, facilitates structural elucidation, and improves the overall reliability of metabolite profiling. Liquid chromatography coupled with mass spectrometry (LC-MS) has emerged as one of the principal analytical platforms employed in modern metabolomics studies (Gika et al., 2019). Gas chromatography coupled with mass spectrometry (GC-MS) is widely employed for the analysis of thermostable volatile compounds. However, this technique often requires relatively complex and time-consuming sample preparation procedures before analysis (Wang et al., 2019). In addition to the use of multiple analytical techniques, recent advances in metabolomics have emphasized the importance of automation in analytical workflows. Automated sample preparation, data acquisition, and data processing platforms help minimize human-induced variability and improve reproducibility across laboratories. Furthermore, automation significantly increases analytical throughput, allowing a larger number of samples to be processed within shorter timeframes (Lehotay et al., 2011). Such high-throughput capabilities are particularly valuable in clinical, pharmaceutical, and systems biology applications, where large-scale metabolite datasets are required for biomarker discovery and comprehensive metabolic profiling. Some of the major examples of bioactive molecules reported from endophytes using different analytical tools have been listed in Table 2.

TABLE 2. Bioactive metabolites reported from endophytic microorganisms using analytical tools.

2.4.1 Mass spectrometry

Mass spectrometry (MS) is the most extensively utilized analytical technique in metabolomics due to its high sensitivity, speed, and selectivity. It enables both qualitative and quantitative analysis of metabolites while offering strong capabilities for compound identification. Contemporary mass spectrometry platforms incorporate a variety of technologies that vary in terms of operational principles and analytical performance. These differences are largely attributed to the ionization techniques employed, the type of mass analyzer used, and the resulting resolving power and mass accuracy. Among the ionization methods widely applied in metabolomics are electron ionization (EI), electrospray ionization (ESI), and atmospheric pressure chemical ionization (APCI) (Lei et al., 2011). Additionally, techniques such as chemical ionization, matrix-assisted laser desorption/ionization (MALDI), and more recent developments including desorption electrospray ionization (DESI) (Takáts et al., 2005) and extractive electrospray ionization (EESI) have expanded the analytical capabilities of MS-based metabolomics. Mass analysers with different resolving powers have also been used in metabolomics. A wide range of mass spectrometry instruments are currently applied in metabolomics investigations. High-resolution systems such as Fourier transform ion cyclotron resonance MS (FT-ICR-MS), Orbitrap MS, and multipass time-of-flight (TOF) MS offer superior resolving power and mass accuracy for detailed metabolite characterization. In contrast, lower-resolution instruments, including linear and three-dimensional ion traps as well as single quadrupole analysers, remain commonly used because of their operational simplicity and lower cost. Since each mass analyser presents distinct advantages and limitations, the selection of a suitable MS platform depends on the objectives of the metabolomics study, the desired analytical throughput, and economic considerations. These systems may be employed either for direct MS-based metabolite profiling or in combination with chromatographic separation techniques to analyze chemically complex metabolomes.

2.4.2 Chromatography coupled with MS

The integration of chromatographic techniques with mass spectrometry provides an effective approach for analysing complex mixtures and is widely applied in metabolomics studies (Metabolomics and 2013, 2013) A mass spectrometry (MS) instrument typically comprises three primary components: an ion source, a mass analyser, and a detector. The ion source converts neutral molecules from the sample into charged ions. These ions are then separated by the mass analyser based on their mass-to-charge (m/z) ratios before being recorded by the detector. Several advantages have been associated with chromatographic separation of metabolites prior to MS analyses, allowing precise quantification of each metabolite, effective separation of isomers, reduction of matrix effects and generation of additional information required for metabolite annotation. In MS-based metabolomics, gas chromatography (GC), liquid chromatography (LC), and capillary electrophoresis (CE) are the three primary separation techniques widely utilized for metabolite analysis. More advanced multidimensional separation methods, including two-dimensional GC (GC × GC) and two-dimensional LC (LC × LC), provide enhanced resolution for the analysis of highly complex biological mixtures. Nevertheless, these techniques remain relatively underutilized compared to conventional chromatographic methods. It should be emphasized that metabolomics is not confined to GC-MS and LC-MS platforms. In principle, any analytical technique that enables broad and comprehensive analysis of the metabolome may be employed in metabolomics studies.

Electrospray ionization (ESI) has become one of the most widely employed ionization techniques in LC-MS-based metabolomics studies. Its widespread use is mainly due to its classification as a soft ionization method, which allows ion formation through charge transfer processes in solution and often results in the generation of intact molecular ions that facilitate reliable metabolite identification (Ren et al., 2018). Moreover, chromatographic separation before the ionization step reduces sample complexity and minimizes matrix effects, thereby improving the efficiency and reproducibility of ionization during mass spectrometric analysis.

2.4.3 Gas chromatography mass spectrometry

GC-MS is commonly recognized as a powerful and versatile analytical technique in metabolomics research. Its widespread application is attributed to its high sensitivity, selectivity, robustness, reproducibility and superior separation performance (Beale et al., 2018). Electron ionization (EI) and chemical ionization (CI) are the two principal ionization methods utilized in GC-MS analysis. Currently, EI is the predominant technique applied in metabolomics workflows. The extensive availability of EI mass spectral databases and libraries significantly aids metabolite identification, making it the preferred ionization approach for many researchers (Poojary, A & 2016). Despite its advantages, GC-MS has an inherent limitation in that it is primarily suitable for the analysis of low molecular weight (approximately 50–600 Da) and volatile compounds. Therefore, the detection of polar, thermolabile, or non-volatile metabolites generally requires chemical derivatization before analysis (Jun et al., 2016).

2.4.4 Nuclear magnetic resonance

Nuclear magnetic resonance (NMR) spectroscopy represents an important analytical tool in metabolomics research. The technique provides several benefits, such as high reproducibility, rapid data acquisition enabling high-throughput analysis, reliable metabolite identification, and the ability to analyze samples without causing their destruction (Takis et al., 2019). Furthermore, NMR spectroscopy is highly compatible with automated analytical workflows and provides outstanding reproducibility. These characteristics make it especially advantageous for high-throughput and largescale metabolomics studies, often offering greater practicality than LC-MS or GC-MS-based approaches (Emwas et al., 2019). Proton (H¹NMR) Spectroscopy and Two-dimensional (2D) NMR Spectroscopy are among the major techniques of NMR that have potential applications in metabolomics.

Metabolomics has significantly advanced the identification and characterization of secondary metabolites in complex biological systems. By integrating high-resolution analytical platforms such as LC-MS, GC-MS, and NMR spectroscopy, it enables comprehensive detection, structural annotation, and comparative analysis of metabolites in plant and microbial extracts. These approaches facilitate the rapid identification of both known and novel compounds through spectral databases, molecular networking, and multivariate analysis (Wolfender et al., 2019). Additionally, metabolomic profiling provides insights into metabolic changes driven by environmental factors, microbial interactions, or genetic variations, thereby enhancing our understanding of secondary metabolite biosynthesis and function. Consequently, metabolomics has become an indispensable tool for accelerating the discovery of bioactive natural products, particularly antimicrobial compounds from medicinal plants and their associated endophytes.

2.5 Why single omics approaches fall short: The need for integration

Although individual omics technologies have significantly advanced our understanding of biological systems, relying on a single omics layer often provides only a limited perspective on complex biochemical processes. For example, metagenomics enables the identification of microbial diversity and the detection of biosynthetic gene clusters that encode enzymes responsible for secondary metabolite production. However, the presence of such genetic information does not necessarily indicate that the corresponding metabolites are actively synthesized under specific environmental or physiological conditions. Similarly, metabolomics allows comprehensive detection and quantification of metabolites within biological samples, but it cannot directly identify the genes or microbial sources responsible for their biosynthesis. Consequently, single-omics approaches often fail to establish a direct connection between genetic potential, metabolic pathways, and the production of bioactive compounds (Zampieri et al., 2017). To overcome these limitations, the integration of multiple omics approaches, particularly metagenomics and metabolomics, has emerged as a powerful strategy for natural product discovery. Metagenomics provides insights into microbial community composition and biosynthetic gene clusters, while metabolomics enables direct detection and structural characterization of the metabolites produced by these microorganisms. The integration of genomic and metabolomic datasets allows researchers to link biosynthetic genes with their corresponding metabolic products, thereby facilitating the identification of novel bioactive compounds. Such multi-omics approaches have become increasingly important in antimicrobial drug discovery, as they help uncover cryptic or silent biosynthetic pathways and accelerate the discovery of new natural products from complex microbial communities (Wolfender et al., 2019).

3 Extreme niches as reservoirs of novel antimicrobial chemistry

The search for the next generation of antibiotics has moved from the backyard soil to the Earth’s forbidden zones. Extreme niches, habitats characterized by geochemical parameters at the edge of biological tolerance, represent the most promising, yet underexplored, frontier for drug discovery. In these high-stress environments, microbial life is governed by a different set of rules. Traditional competition is replaced by extreme chemical warfare, where microorganisms must produce highly potent and structurally unique metabolites just to secure a foothold in nutrient-scarce or hostile conditions. This evolutionary pressure promotes the emergence of BGCs encoding enzymatic architectures distinct from those found in traditional antibiotic-producing microbes (Bull et al., 2018). In recent landscape, the consensus is that the microbial dark matter is richest in these untapped reservoirs. Recent systematic reviews have identified many novel BGCs from extreme niches like Antarctic permafrost and deep-sea vents, with 60–99% of these clusters showing zero homology to any known antibiotic classes in the MIBiG database (Prayogo, 2026). These hostile biomes serve as evolutionary laboratories for the creation of unprecedented antimicrobial chemistry.

Advances in metagenomics have substantially expanded our ability to interrogate the biosynthetic capacity of these environments. Large-scale sequencing of extreme microbiomes frequently reveals a high abundance of previously uncharacterized BGCs, many of which display minimal similarity to known clusters in existing databases (Charlop-Powers et al., 2015). Recent developments in artificial intelligence-assisted genome mining and machine learning-based metabolite prediction have further improved the prioritization of these clusters, allowing researchers to identify high-value antimicrobial candidates from increasingly complex environmental datasets (Aghdam and Brown, 2021; Boudza et al., 2026; Paerhati et al., 2026) Integrating metagenomics with metabolomics, synthetic biology, and computational prediction frameworks provides a powerful route to translate the genetic diversity of extremophilic microbiomes into chemically validated antimicrobial leads (Hannigan et al., 2019; Kautsar et al., 2020).

3.1 Extreme environments and microbial adaptive metabolism

Extremophiles have evolved distinctive metabolic strategies that frequently involve the production of protective or competitive secondary metabolites. These metabolic responses are not merely passive survival mechanisms; they represent an active, intricate reorganization of cellular priorities to maintain functionality where life ordinarily falters (Rothschild and Mancinelli, 2001; Cavicchioli, 2002). Adaptive metabolism in extremophiles often serves multiple ecological functions that are critical for navigating harsh niches. Secondary metabolites produced in these contexts may act as antimicrobial agents, signalling molecules, metal chelators, Osmo protectants, or stress-response modulators that help maintain cellular homeostasis (Sayed et al., 2020). In densely populated microbial communities, these compounds can mediate intense interspecies competition, enabling producers to inhibit competing organisms and secure limited resources. Such chemically mediated interactions contribute to the remarkable diversity of bioactive natural products encoded within extremophile genomes, effectively turning these harsh habitats into evolutionary laboratories of chemical innovation.

Recent genomic and metagenomic surveys have revealed that extremophilic microorganisms frequently harbor complex biosynthetic gene clusters (BGCs) encoding polyketides, Nonribosomal peptides, terpenoids, and Ribosomally synthesized peptides. These pathways often display unique enzymatic features, reflecting evolutionary adaptations to environmental constraints such as low temperatures or high osmotic stress (Wang et al., 2024). Particularly, Himalayan psychrophilic ecosystems represent highly dynamic evolutionary microenvironments characterized by low temperatures, freeze-thaw cycles, high ultraviolet (UV) radiation, hypoxia, nutrient limitation, and oxidative stress. These combined ecological pressures impose strong selective constraints on resident microbial communities, driving the evolution of specialized metabolic and membrane-adaptive strategies necessary for survival under extreme conditions. Unlike temperate soil microbiomes, where metabolic competition is often shaped by relatively stable environmental parameters, psychrophilic microorganisms from high-altitude Himalayan niches exist within fluctuating and resource-limited habitats that promote intense chemical competition and stress-responsive biosynthesis. Such environments function as natural evolutionary laboratories, fostering the emergence of biosynthetic gene clusters encoding structurally distinct secondary metabolites with enhanced membrane activity, oxidative stability, cold-adaptive enzymatic properties, and unconventional stereochemistry. Several psychrophilic microorganisms isolated from Himalayan glaciers and alpine soils have been reported to produce unique lipopeptides, polyketides, pigments, and antimicrobial peptides that differ significantly from metabolites typically recovered from temperate terrestrial ecosystems (Borchert et al., 2017; Kuvarina et al., 2022). For example, enzymes from psychrophilic microorganisms often exhibit enhanced catalytic flexibility to remain functional in the cold, whereas those from thermophiles possess increased structural stability to resist denaturation. These inherent physical properties influence the chemical diversity and functional range of their associated metabolites, often resulting in structures that are distinct from those found in more temperate environments (Margesin and Feller, 2010; Margesin, 2017). From an antimicrobial discovery perspective, the metabolic adaptations of extremophiles offer two key advantages that are vital for overcoming modern resistance challenges. First, the evolutionary divergence of extremophile biosynthetic pathways significantly increases the likelihood of identifying structurally novel antimicrobial scaffolds that have not been over-mined in common soil microbes. Second, the ecological pressures shaping metabolite production may favour compounds with potent bioactivity against competing microorganisms, providing a natural selection mechanism for antimicrobial efficacy. By evolving to kill competitors in the most unforgiving places on Earth, these microbes have already done much of the hard work of screening for potent molecules (Borchert et al., 2017; Bull et al., 2018). These ecological and biochemical adaptations substantially expand the accessible chemical space for antimicrobial discovery and increase the probability of identifying compounds with novel mechanisms of action against multidrug-resistant pathogens. Recent computational advances are beginning to reveal the full extent of this biosynthetic potential, moving beyond what we can simply culture in a lab. Machine learning-based genome mining tools can now analyze large metagenomic datasets from extreme environments to identify cryptic BGCs with high novelty scores and predicted antimicrobial activity. Coupled with metabolomic validation and synthetic biology-based expression systems, these approaches are enabling a systematic exploration of extremophile metabolic diversity (Hannigan et al., 2019). This synergy of evolutionary biology and modern technology is rapidly accelerating the identification of previously inaccessible antimicrobial chemotypes, providing a conceptual foundation for targeted discovery in these underexplored ecosystems.

3.2 Metagenomic mining of biosynthetic gene clusters from extreme niches

Many extremophiles possess highly specific and difficult-to-replicate growth requirements; traditional laboratory cultivation often misses over 99% of the microbial diversity present in samples from deepsea hydrothermal vents, hypersaline lakes, and polar deserts (Aulitto and Gallo, 2025). Metagenomic mining allows researchers to directly access the total eDNA, enabling the systematic identification of BGCs from organisms that have never been grown in a lab. This culture-independent approach has transformed extreme environments into high-priority targets for bioprospecting, revealing that these ecosystems are significantly more biosynthetically diverse than previously estimated (Paoli et al., 2022). Recent advances in sequencing technologies have significantly improved the resolution of metagenomic mining in these harsh landscapes. A significant bottleneck in metagenomic mining remains the BGC fragmentation inherent in short-read datasets. Large modular biosynthetic systems, such as NRPS and PKS, often span 50–100 kb dimensions that frequently exceed the N50 contiguity of standard environmental assemblies. When these clusters are broken into unusable fragments, the genomic context required to understand their regulation and metabolic integration is lost. While longread technologies like Oxford Nanopore provide a hardware solution for contiguity, the computational bridging of these fragments in existing Illumina datasets remains a primary objective for contemporary bioinformatics pipelines (Waschulin et al., 2021). These high-quality metagenome-assembled genomes (MAGs) provide essential genomic context for understanding how secondary metabolism is integrated into the survival strategies of extremophiles, such as the co-localisation of BGCs with stress response genes and defence systems (Mikheenko, Saveliev and Gurevich, 2016; Bowers et al., 2017). Computational genome mining has emerged as the primary engine for navigating the massive datasets generated from these niches. Standard platforms like antiSMASH remain the gold standard for rulebased detection, but the high degree of evolutionary divergence in extreme environments often requires the use of machine learning-based tools like DeepBGC and BGC-Prophet (Hannigan et al., 2019; Lai et al., 2025). These tools use deep learning to identify cryptic BGCs that do not follow known biosynthetic rules, which is particularly relevant in extreme niches where 60%–99% of detected BGCs lack characterized homologs in global databases like MIBiG (Rahimian et al., 2026). By identifying these high-novelty clusters, researchers can prioritize previously inaccessible chemical scaffolds for further experimental study. Despite the vast biosynthetic potential revealed by these computational surveys, with some recent studies identifying over 14,000 unique BGCs across global extreme environments, the transition from in silico prediction to experimental validation remains a significant bottleneck. Currently, only a small fraction of metagenomically predicted antimicrobial BGCs have been successfully expressed in heterologous hosts to confirm their bioactivity. However, the integration of synthetic biology, where entire BGCs are refactored and inserted into model organisms like Streptomyces or E. coli, is beginning to bridge this gap (Smanski et al., 2016; Newman and Cragg, 2026). This systematic approach ensures that the metabolic innovations evolved over eons in the Earth’s most forbidding niches can finally be harnessed for modern therapeutic applications, specifically in the fight against multidrug-resistant pathogens.

3.3 Machine learning and computational pipelines for novel BGC identification

The transition from traditional rule-based genome mining to data-driven machine learning frameworks has fundamentally altered the trajectory of natural product discovery. Historically, antiSMASH utilized Profile Hidden Markov Models (pHMMs) to identify biosynthetic gene clusters (BGCs) based on well defined sequence motifs associated with known biosynthetic enzymes, such as ketoacyl synthase or adenylation domains (Medema et al., 2011; Blin et al., 2023. While effective for characterizing canonical clusters, these rule-based approaches are inherently limited by their reliance on homology, often failing to detect highly divergent or cryptic BGCs that lack characterized analogues in databases like MIBiG (Doroghazi et al., 2014; Kautsar et al., 2020). Machine learning models, particularly deep learning architectures, circumvent these limitations by learning the underlying genomic grammar of biosynthetic logic directly from high-dimensional sequence data, enabling the discovery of entirely novel chemical scaffolds (Caesar and Cech, 2019; Hannigan et al., 2019). Deep learning approaches, specifically Recurrent Neural Networks (RNNs) and Convolutional Neural Networks (CNNs), have demonstrated superior performance in identifying BGC boundaries within fragmented metagenomic assemblies. For instance, DeepBGC employs a Bidirectional Long Short-Term Memory (Bi-LSTM) network to capture long-range dependencies between protein domains, significantly outperforming traditional HMMs in detecting non-canonical clusters (Hannigan et al., 2019; Miller et al., 2019). More recently, the development of GECCO (Gene Cluster COmpressor) has introduced Conditional Random Fields (CRFs) to provide faster and more scalable BGC identification, which is critical for processing the multi-terabyte datasets generated by global metagenomic surveys (Carroll et al., 2021). A critical limitation in BGC prediction lies in the accurate delineation of cluster boundaries, particularly in fragmented metagenomic assemblies where biosynthetic genes are interspersed with housekeeping genes. Rulebased approaches often rely on the presence of signature domains, which can lead to either truncation of clusters or inclusion of adjacent non-biosynthetic regions. Unlike rule-based systems, CRF-based models treat the genome as a sequence of probabilistic labels, allowing the algorithm to distinguish BGC genes from the genomic background of standard primary metabolic and housekeeping genes with high precision. By modeling the transition states between biosynthetic and non-biosynthetic genomic contexts, CRFs effectively resolve the challenge of boundary detection, reducing the risk of chimeric BGC assemblies and ensuring that only functionally relevant genes are prioritized for downstream heterologous expression, especially for non-canonical or hybrid clusters. However, while GECCO excels at rapid boundary detection on contiguous scaffolds, it is less equipped to handle the BGC fragmentation issues. In contrast, Transformer-based models, such as BGC-Prophet, leverage self attention mechanisms to bridge these assembly gaps. By recognizing long-range genomic dependencies and conserved syntenic patterns across disparate contigs, Transformers can predict the likelihood that two fragmented sequences belong to the same 50–100 kb modular system. This architectural shift from local (CRF) to global (Transformer) context allows researchers to reconstruct virtual BGCs from short-read data, a critical advancement for the characterization of complex PKS/NRPS systems in biomes where high-cost long-read sequencing is not feasible. significantly expanding the discovery potential in complex biomes where high-depth long-read sequencing is costprohibitive. However, challenges persist in accurately distinguishing accessory genes that indirectly contribute to biosynthesis, highlighting the need for continued refinement of context-aware prediction models. Furthermore, the integration of Transformer-based models, which utilize self-attention mechanisms, is beginning to allow researchers to predict not only the presence of a BGC but also its potential bioactivity and chemical class directly from primary sequence data (Aghdam and Brown, 2021; Lai et al., 2025).

As the volume of predicted BGCs continues to grow exponentially, the challenge has shifted from identification to prioritization. Computational pipelines must now filter through thousands of clusters to identify those with the highest probability of chemical novelty and biological relevance. Tools such as BiG-SLICE and BiG-FAM facilitate this by organizing millions of BGCs into biosynthetic gene families (BGFs) through massive-scale clustering, allowing researchers to pinpoint singleton clusters that have no close relatives in current databases (Navarro-Muñoz et al., 2020; Kautsar et al., 2021). This prioritization is often augmented by metabolic networking and BGC-to-metabolite matching, which integrates metagenomic data with tandem mass spectrometry (LC-MS/MS) via platforms like the Global Natural Products Social Molecular Networking (GNPS) (Wang et al., 2016; Andreani-Gerard et al., 2024). Such integrated omics approaches ensure that experimental resources are directed toward the most promising biosynthetic dark matter. The recent revolution in protein structure prediction is providing a new dimension to computational mining. By integrating tools like AlphaFold2 into the BGC mining workflow, researchers can now model the 3D structures of uncharacterized enzymes within a predicted cluster to infer their catalytic functions and substrate specificities (Jumper et al., 2021; Varadi et al., 2022). This structural genomics approach is particularly valuable for extremophiles, where enzymes often possess unique folds to maintain stability under harsh conditions. When combined with phylogenetic analysis and machine learning based novelty scoring, these pipelines provide a comprehensive roadmap for the targeted discovery of next-generation antimicrobials from previously inaccessible ecosystems (Gavriilidou et al., 2022; Rahimian et al., 2026). Beyond cluster detection, ML-driven models are increasingly used for prioritization of bioactive candidates. By integrating features such as domain organization, evolutionary conservation, resistance gene signatures, and genomic context, predictive algorithms can rank BGCs based on their likelihood of encoding antimicrobial compounds (Navarro-Muñoz et al., 2020; Gavriilidou et al., 2022). More recently, AI-based tools have incorporated metabolomics data to enable genome-metabolome correlation, facilitating the linkage of predicted gene clusters to observed chemical features in complex extracts. Another emerging direction involves structure prediction and retrosynthetic inference from BGC sequences. Computational frameworks such as PRISM and NPLinker can generate predicted metabolite scaffolds and match them with mass spectrometry-derived molecular networks, accelerating dereplication and reducing rediscovery rates. These integrative pipelines are particularly important in AMR-focused discovery, where rapid identification of structurally novel compounds is essential (Skinnider et al., 2017; Eldjarn et al., 2021).

3.4 Functional metagenomics and heterologous expression strategies

Functional metagenomics complements sequence-based mining by directly identifying bioactive antimicrobial compounds through phenotype-driven screening. This approach involves constructing large-insert metagenomic libraries, typically using Cosmid or BAC vectors, which are subsequently expressed in surrogate hosts and screened for antimicrobial activity against clinical pathogens (Gillespie et al., 2002; Handelsman, 2004). Unlike sequence dependent approaches, functional metagenomics does not rely on prior knowledge of biosynthetic gene cluster (BGC) sequences, enabling the discovery of entirely novel antimicrobial scaffolds that may escape homology-based prediction frameworks (Iqbal, 2016; Scherlach and Hertweck, 2021). Landmark discoveries such as malacidins and teixobactin demonstrate the power of phenotype-first screening strategies in identifying antibiotics with unique mechanisms of action through inhibition of cell-wall biosynthesis (Piddock, 2015; Hover et al., 2018). Successful recovery of antimicrobial BGCs depends heavily on the choice of the heterologous expression system and the vector architecture. For relatively small BGC, high-expression systems such as the pET vector series are widely employed due to their robust T7 promoter-driven transcriptional efficiency (Rosano and Ceccarelli, 2014; Palazzotto et al., 2019). These systems are particularly useful when specialized chaperones or disulfide-bond isomerases are required for the correct folding of complex antimicrobial proteins or peptides (Sørensen and Mortensen, 2005; Niazi and Magoola, 2023). In contrast, large modular systems such as NRPS and PKS clusters often require metabolically specialized hosts including S. coelicolor or Myxococcus xanthus, which provide compatible precursor pools, regulatory networks, and sigma factors necessary for efficient biosynthesis (Baltz, 2010; Myronovskyi and Luzhetskyy, 2019).

Despite these advances, transcriptional silence of environmental BGCs remains a major bottleneck in heterologous expression. Many clusters fail to express in surrogate hosts due to incompatibilities in promoter architecture, codon usage, and metabolic context. To address this challenge, recent Predictive Heterologous Expression (PHE) frameworks integrate artificial intelligence and comparative genomics to rationally guide host selection. These models evaluate parameters such as GC content, codon adaptation indices, promoter compatibility, and regulatory motifs to predict optimal expression platforms (Xu et al., 2022; Akbari et al., 2025). By aligning BGC specific biosynthetic requirements with host transcriptional and metabolic landscapes, PHE strategies improve cluster activation efficiency while reducing empirical trial-and-error experimentation. Complementary synthetic biology approaches, including BGC refactoring and CRISPR/Cas-mediated genome engineering, further enhance activation of cryptic pathways by replacing native regulatory elements with standardized inducible promoters and integrating clusters into optimized chassis strains (Smanski et al., 2016b; Zhang et al., 2020). Nevertheless, an incomplete understanding of pathway regulation and precursor availability continues to limit expression efficiency in many systems. The integration of functional metagenomics with analytical chemistry and metabolomics has further accelerated antimicrobial discovery. High-throughput screening of metagenomic clones against MDR pathogens, followed by bioassay-guided fractionation and LC-MS/MS analysis, enables rapid correlation of genetic clusters with their corresponding chemical products (Milshteyn et al., 2018; Ayon, 2023). In extremophilic environments, where biosynthetic pathways may require highly specialized physicochemical conditions, the development of extremophilic expression platforms capable of tolerating high salinity or temperature is emerging as a critical frontier (Aulitto and Gallo, 2025; Rahimian, Aghazadeh-Soltan-Ahmadi and Panahi, 2026). Collectively, these advances are transforming functional metagenomics into a predictive and scalable platform for next-generation antimicrobial discovery.

4 Medicinal plants and associated microbiomes in antimicrobial discovery

Since the dawn of human civilization, plants have constituted the foundation of therapeutic intervention and disease management. Through centuries of empirical observation and experiential learning, early societies systematically recognized plant species with distinct pharmacological properties, laying the groundwork for traditional medicine systems. This enduring plant-human relationship has profoundly shaped the evolution of healthcare, with medicinal plants continuing to serve as indispensable reservoirs of bioactive compounds for modern drug discovery. In ancient Indian texts such as the Vedas and the Bible, the widespread use of herbal remedies and healthcare preparations has been described. WHO defines a medicinal plant as any plant that contains bioactive substances in one or more of its organs, which can be utilized directly for therapeutic purposes or serve as precursors for the semisynthesis of pharmaceutical drugs (Kaul et al., 2012). This definition encompasses both plants that have undergone rigorous scientific validation and those traditionally used in indigenous or folk medicine systems without extensive experimental evaluation. The importance of medicinal plants in global healthcare is underscored by WHO estimates indicating that nearly 80% of the world’s population relies on plant-based remedies in some capacity for primary health care needs. This widespread dependence reflects not only cultural traditions and accessibility but also the significant pharmacological potential embedded within plant-derived natural products.

4.1 Medicinal plants as sources of antimicrobial secondary metabolites

While medicinal plants have traditionally been regarded as the primary source of therapeutic bioactive compounds, emerging evidence indicates that their associated microbiomes, especially endophytes, play a crucial and often overlooked role in the biosynthesis of pharmacologically active secondary metabolites. Endophytes are bacteria and fungi that reside asymptomatically within plant tissues such as leaves, stems, roots and seeds, forming intricate and often mutualistic associations with their host plants (Hashem et al., 2023). These interactions are mediated by a complex network of chemical signals and metabolic exchanges between the plant and its microbial inhabitants. Such plant-microbe chemical interactions play a crucial role in shaping the physiology, metabolism, and ecological fitness of both partners. Increasing evidence suggests that these interactions significantly influence the biosynthesis of secondary metabolites with important pharmacological properties. These endophytic communities contribute significantly to plant fitness by enhancing stress tolerance, modulating host metabolism, and protecting plants against different disease-causing pathogens. Through complex plant-microbe interactions, endophytes can influence nutrient acquisition, induce systemic resistance, and regulate host metabolic pathways, thereby improving plant adaptability to environmental stresses (Compant et al., 2019). Plants produce a wide array of specialized metabolites, including phenolics, terpenoids, alkaloids, and flavonoids, which serve as chemical signals that regulate microbial colonization and communication within plant tissues. These metabolites can act as attractants or inhibitors, selectively shaping the composition of endophytic microbial communities. In response, endophytes may produce signalling molecules, enzymes, or bioactive compounds that modulate host metabolic pathways, thereby contributing to plant growth, defense responses, and environmental adaptation. Such bidirectional communication forms the basis of chemical cross-talk that governs plant-endophyte symbiosis (Brader et al., 2014). Notably, many endophytes have been reported to produce the same or structurally similar bioactive compounds as their host plants, suggesting possible mechanisms such as horizontal gene transfer, co-evolutionary adaptation, or convergent evolution of biosynthetic pathways (Venugopalan and Srivastava, 2015).

A landmark example demonstrating this phenomenon is the discovery of Taxol-producing endophytic fungi isolated from the Pacific yew, T. brevifolia. This finding provided compelling evidence that microorganisms residing within medicinal plants can independently synthesize clinically important metabolites (Stierle et al., 1993). Endophytic microorganisms inhabiting medicinal plants are increasingly recognized as prolific producers of structurally diverse secondary metabolites with significant pharmacological potential. These metabolites, including alkaloids, terpenoids, polyketides, phenolic compounds, flavonoids, and non-ribosomal peptides, are not essential for primary cellular metabolism but play crucial ecological roles in microbial competition, host defense, and environmental adaptation (Kusari et al., 2012). Within plant tissues, endophytes exist in a chemically competitive microenvironment where the production of bioactive metabolites helps to protect the host plant against invading pathogens, herbivores, and environmental stressors. As a result, many of these microbial metabolites exhibit potent antibacterial, antifungal, antiviral, and anticancer activities, making them promising candidates for drug discovery. Several studies have demonstrated that endophytes associated with medicinal plants produce bioactive compounds with antimicrobial activity comparable to, and in some cases exceeding, those produced by their host plants. For example, endophytic fungi isolated from Camptotheca acuminata have been shown to produce the alkaloid camptothecin, a clinically important antitumor compound (El-Hady et al., 2024). Another notable example is podophyllotoxin, an important precursor for anticancer drugs, which has been isolated from endophytic fungi associated with Podophyllum species (Eyberger et al., 2006). A wide range of antimicrobial compounds such as pestalone, 2-hydroxy-6-methyl benzoic acid, clavatol, pseudomycins, ecomycins, penitrem A, javanicin, and cryptocandin Cryptocin, xylarenone B, have been reported from endophytic fungi, demonstrating potent activity against diverse pathogenic microorganisms (Gouda et al., 2016). Such chemical diversity reflects the remarkable metabolic adaptability of endophytes within plant tissues. For example, fungal endophytes isolated from the roots of Paris polyphylla, a Himalayan medicinal plant extensively used in traditional Chinese medicine, have been shown to produce pharmacologically important secondary metabolites with antimicrobial potential, further highlighting the therapeutic relevance of plant-associated microbiomes (Chen et al., 2026). Additionally, bacterial endophytes such as Bacillus and Pseudomonas species isolated from Himalayan medicinal plants have been shown to produce antimicrobial metabolites including lipopeptides and phenazine derivatives, which play a crucial role in pathogen suppression and plant defense (Hardoim et al., 2008). Furthermore, actinobacterial endophytes from high-altitude Himalayan regions have demonstrated the ability to produce polyketides and nonribosomal peptides, indicating their potential as sources of novel antimicrobial scaffolds (Thakur et al., 2022).

This biosynthetic versatility becomes even more pronounced in extreme environments such as the Himalayan high-altitude ecosystems, where abiotic stressors including intense ultraviolet (UV) radiation, low temperatures, hypoxia, and nutrient limitation impose strong selective pressures on both plants and their associated microbiota. These conditions drive adaptive evolution, leading to the diversification and activation of biosynthetic gene clusters (BGCs) that produce specialised metabolites (Ziemert et al., 2016). UV-induced oxidative stress promotes the synthesis of protective compounds such as antioxidants and antimicrobial agents, while temperature fluctuations favor the evolution of structurally flexible and stable metabolites with enhanced bioactivity (Thakur et al., 2022; Jiang et al., 2026). Collectively, these observations underscore that Himalayan medicinal plants and their endophytes represent a unique ecological niche where environmental stress, microbial interactions, and metabolic innovation converge, offering immense potential for the discovery of novel antimicrobial compounds (Gupta et al., 2019). These findings highlight the remarkable biosynthetic capacity of endophytes and their potential as valuable sources of novel therapeutic compounds. Table 3 provides an overview of the major classes of secondary metabolites produced by endophytic microorganisms and highlights their biosynthetic pathways and biological significance.

TABLE 3. Major classes of secondary metabolites produced by endophytic microorganisms and their biological significance.

4.2 Metabolomics-driven discovery of plant- and endophyte-derived antimicrobials

The discovery of novel antimicrobial compounds from natural sources has increasingly relied on metabolomics-based approaches. Metabolomics enables comprehensive profiling of small molecules produced by organisms under specific physiological or environmental conditions, thereby providing a powerful platform for identifying bioactive metabolites from plants and their associated microbiomes. In the context of medicinal plants and their endophytes, metabolomics offers a systematic strategy to explore the chemical diversity generated through plant-microbe interactions and to uncover antimicrobial compounds that may otherwise remain undetected using traditional isolation techniques (Wolfender et al., 2014). One of the major advantages of metabolomics in antimicrobial discovery is its ability to reveal previously uncharacterized chemical entities. Untargeted metabolomics platforms using techniques such as liquid chromatography-mass spectrometry (LC-MS), gas chromatography-mass spectrometry (GC-MS), and tandem mass spectrometry (MS/MS) enable comprehensive metabolite detection without prior knowledge of compound identity. Advanced analytical pipelines involving spectral alignment, feature extraction, and multivariate statistical analyses can then be used to identify metabolites that correlate with antimicrobial activity or specific biological phenotypes. Table 4 summarizes representative antimicrobial compounds isolated from endophytes associated with medicinal plants, along with their host organisms and reported biological activities. Together, these examples illustrate the immense potential of plant-associated endophytes as a reservoir of structurally diverse metabolites with promising antimicrobial applications.

TABLE 4. Secondary metabolites derived from endophytic microorganisms associated with medicinal plants and their reported biological activities.

4.3 Dereplication strategies in metabolomics guided antimicrobial discovery

Bioactivity-guided metabolomics, in combination with dereplication strategies, has emerged as a powerful and efficient framework for the discovery of bioactive secondary metabolites from complex biological systems. Traditional natural product discovery approaches primarily relied on bioactivity guided fractionation, which involves sequential extraction, separation, and testing of compounds. Although effective, this approach is labour-intensive, time-consuming, and often leads to the repeated isolation of previously known compounds (Wolfender et al., 2019). In contrast, metabolomics-driven approaches integrate high-throughput chemical profiling with biological screening, enabling the direct correlation of metabolite features with observed bioactivities. Analytical platforms such as LC-MS/MS and NMR spectroscopy are widely employed to generate comprehensive metabolite profiles, which are subsequently analyzed using multivariate statistical tools such as principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA) to identify metabolites associated with antimicrobial or pharmacological activity (Lima et al., 2023). A central component of this workflow is dereplication, which enables the rapid identification of known compounds at early stages of analysis, thereby minimizing redundant isolation efforts. High-resolution mass spectrometry, particularly LC-MS/MS, provides accurate mass measurements and fragmentation patterns that serve as molecular fingerprints for metabolite identification (Dührkop et al., 2021). These spectral features are compared with reference datasets to annotate known compounds, allowing researchers to prioritize novel metabolites for further investigation. As a result, dereplication significantly accelerates the natural product discovery pipeline and improves the efficiency of identifying biologically relevant compounds(Nothias et al., 2020).

A crucial advancement in dereplication strategies is the utilization of comprehensive spectral and chemical databases, which facilitate rapid and reliable metabolite annotation. Experimental MS/MS data are matched against curated databases such as the Human Metabolome Database (HMDB), METLIN, MassBank, and the Global Natural Products Social Molecular Networking (GNPS) platform (Wishart et al., 2007; Wishart, 2007). These resources contain extensive information on metabolite structures, fragmentation spectra, physicochemical properties, and biological annotations, enabling high-confidence identification of known compounds. Among these, GNPS is particularly valuable due to its community-driven framework, which allows sharing and continuous expansion of spectral data, thereby enhancing dereplication accuracy. In addition to general metabolomics databases, specialized natural product databases such as the Dictionary of Natural Products (DNP), NPAtlas, and AntiBase provide curated information on microbial and plant-derived secondary metabolites, further supporting dereplication efforts in antimicrobial discovery (Van Der Hooft et al., 2016). Recent developments in computational metabolomics have further strengthened dereplication workflows through the integration of molecular networking and in silico annotation tools. Molecular networking platforms, particularly GNPS, organize MS/MS data based on spectral similarity, allowing structurally related metabolites to be visualised as clusters. This facilitates not only the rapid identification of known compounds but also the detection of structurally related unknown metabolites that may represent novel chemical entities.

Although advanced analytical techniques such as LC-MS, GC-MS, and NMR have significantly improved the secondary metabolite detection and profiling, they do not necessarily guarantee the identification of secondary metabolites. In fact, in untargeted metabolomics, the majority of detected spectral features remain unannotated, constituting the so-called dark metabolome. Across the literature the term dark metabolome has been used in several ways (Da Silva et al., 2015). These m/z features remain structurally uncharacterized via MS or MS/MS analysis and may represent either known compounds or novel, previously undescribed molecules. It is crucial to reiterate a fundamental consensus in the field: while every metabolite is inherently a molecule, not every molecule detected within a biological sample by LC-MS or MS/MS can be classified as a metabolite. This group of undetected compounds contains both familiar chemical structures and completely new molecules that have never been documented before (Giera et al., 2024). In recent years, numerous computational tools have been designed to query chemical structure databases that are significantly larger than traditional spectral libraries. Notable examples of these tools include MS-FINDER (Tsugawa et al., 2016), DEREPLICATOR+ (Mohimani et al., 2018), MAGMa (Ridder et al., 2013), MetFrag (Ruttkies et al., 2016) MIDAS (Wang et al., 2014), CFM-ID (Allen et al., 2015a), CSI: FingerID (Dührkop et al., 2015). Among the listed ones, the benchmark tool for is CSI:FingerID being a part of SIRIUS software. SIRIUS performs metabolite identification in a two-step approach: Firstly, the molecular formula of the query compound is determined via isotope pattern analysis and fragmentation trees. Second, SIRIUS uses CSI: FingerID to predict a molecular fingerprint from the given spectrum and fragmentation tree. This predicted fingerprint can be searched against a structure database to identify the most likely candidate. It constructs a more comprehensive profile of metabolic systems, SIRIUS facilitates untargeted identification workflows (Da Silva et al., 2015). Recognizing that many biomolecules are not yet documented in traditional structure databases, the platform still provides critical insights for unknown compounds. It accomplishes this by generating probabilistic molecular fingerprints to aid in de novo structure elucidation and by enabling searches against databases of theoretical compounds, such as the in silico generated MINE databases (Jeffryes et al., 2015). Allen (2014) introduced CFM-ID (Competitive Fragmentation Modeling for Metabolite Identification), a freely accessible web server that addresses several challenges in untargeted metabolomics (Allen et al., 2015b). CFM-ID is a web-based platform designed to assist in identifying metabolites by analysing tandem mass spectrometry (MS/MS) data. It is an advanced in silico tool that models MS/MS fragmentation as a stochastic and homogeneous Markov process, referred to as competitive fragmentation modeling (CFM) (Allen et al., 2013). CFM-ID incorporates two modeling approaches: a Single Energy CFM model, which serves as a baseline for spectra acquired at a fixed collision energy, and a Combined Energy CFM model, which integrates spectral data obtained at multiple energy levels for improved predictive accuracy (Wang et al., 2021).

To capture fragmentation behaviour, the method utilizes binary feature vectors that encode structural characteristics such as specific atom adjacencies and common neutral losses. These features enable the model to learn fragmentation patterns associated with different molecular structures. Importantly, CFM-ID can be applied in two complementary ways: it can predict MS/MS spectra from known chemical structures and rank candidate structures for an unknown spectrum based on similarity scoring (Nijssen et al., 2023). This dual capability makes CFM-ID a powerful tool for metabolite identification, particularly in addressing the challenges of annotating unknown compounds within the dark metabolome (Nothias et al., 2020). Importantly, the integration of bioactivity-guided metabolomics with dereplication strategies allows the prioritization of metabolites that are both chemically novel and biologically relevant. By correlating metabolite abundance with bioactivity data, researchers can identify key compounds responsible for antimicrobial effects. This approach is particularly valuable in studies involving medicinal plants and their associated endophytes, where complex metabolite mixtures and dynamic plant-microbe interactions often lead to the production of diverse and previously unexplored secondary metabolites. Overall, this integrated strategy provides a robust, high-throughput, and efficient platform for natural product discovery, significantly enhancing the identification of novel antimicrobial agents.

4.4 Integrative metagenomics-metabolomics approaches for AMR-oriented discovery

The emergence of AMR has necessitated the development of innovative strategies for discovering novel antimicrobial agents. Integrative multi-omics approaches, particularly the combination of metagenomics and metabolomics, have emerged as powerful tools for exploring the biosynthetic potential of complex microbial communities. Metagenomics enables the identification of BGCs from unculturable microorganisms, while metabolomics provides direct evidence of metabolite production and chemical diversity. The integration of these approaches allows for the functional linkage between genetic potential and metabolite expression, thereby significantly enhancing the discovery of novel antimicrobial compounds from environmental biomes (Kenshole et al., 2021). The identification of BGCs and their corresponding metabolites represents a central challenge in natural product discovery, particularly in the context of complex microbial communities. Metagenomics has enabled the largescale exploration of microbial genomes, leading to the discovery of numerous BGCs encoding enzymes involved in the biosynthesis of structurally diverse secondary metabolites, including polyketides, nonribosomal peptides, terpenoids, and alkaloids. However, the presence of a BGC does not necessarily guarantee the production of its associated metabolite, as many gene clusters remain transcriptionally silent or are expressed only under specific environmental conditions. This disconnect between genetic potential and metabolite expression highlights the limitations of genome-based predictions alone and underscores the need for integrative approaches that combine metagenomic data with metabolomic profiling (Navarro-Muñoz et al., 2020).

Metabolomics provides a complementary strategy by enabling the direct detection and characterization of metabolites produced by microbial communities under defined conditions. High-resolution analytical platforms such as LC-MS/MS and NMR spectroscopy allow comprehensive profiling of metabolite signatures, which can be correlated with predicted biosynthetic pathways derived from metagenomic analyses. Advances in bioinformatics tools, including genome mining platforms such as antiSMASH and PRISM, have facilitated the prediction of chemical structures from BGC sequences, thereby enabling more accurate linkage between genes and metabolites. By integrating metabolomic datasets with genomic predictions, researchers can assign putative functions to BGCs and validate the production of corresponding metabolites, thus bridging the gap between genotype and chemotype (Medema et al., 2014). Recent developments in multi-omics integration have shifted toward frameworks that bridge the gap between Gene Cluster Families (GCFs) and Molecular families (MFs). A standout platform in this domain is NPLinker, which utilizes a data-centric approach to alleviate the manual bottleneck of matching BGCs with metabolites (Eldjarn et al., 2021; Louwen et al., 2023). The core methodological logic of such integration relies on statistical cooccurrence across diverse environmental samples. NPLinker calculates correlation scores between BGC abundance profiles, such as the Standardised Correlation Score, and metabolite intensity patterns across multiple environmental or experimental samples. The underlying principle is that metabolites originating from a specific BGC should exhibit abundance patterns that co-vary with the occurrence of that cluster across datasets. For a potential GCF-MF link, the system assigns a score (σ_corr) based on the intersection of samples where both the specific gene cluster and the metabolic feature are detected. To move beyond qualitative observation, these frameworks calculate p-values to distinguish true biosynthetic links from random co-occurrence, typically utilizing permutation testing or Cauchy combination tests to account for the inherent sparsity of metagenomic data (Navarro-Muñoz et al., 2020).

The standard integrative workflow proceeds through a structured sequence of comparative genomics, molecular networking, and data integration. Initially, BGCs are organized into GCFs using clustering algorithms such as BiG-SCAPE, while tandem mass spectrometry (LC-MS/MS) data are processed via the Global Natural Products Social Molecular Networking (GNPS) platform to group related spectra into MFs (Nothias et al., 2020b). Subsequently, NPLinker is utilized to compute link scores between these two distinct datasets, allowing researchers to prioritize high-scoring links with significant pvalues for targeted isolation or heterologous expression. A primary empirical validation of this approach is the discovery of the malacidins, a class of calcium-dependent antibiotics. By correlating the presence of specific biosynthetic signatures across more than 2,000 diverse soil metagenomes with corresponding MS/MS signals, researchers successfully bypassed the requirements for traditional cultivation to isolate a novel chemical scaffold targeting cell-wall biosynthesis in multidrug-resistant pathogens (Hover et al., 2018). Such metabologenomic pipelines represent the current state-of-theart for exploring high-altitude niches, such as those found in the Himalayan regions, where high degrees of endemism and niche specialization provide the necessary sample diversity to achieve the statistical significance required for robust correlation scoring (Rahimian et al., 2026; Li et al., 2026). Similarly, genome-guided discovery of darobactin from Photorhabdus symbionts integrated biosynthetic operon analysis with metabolomic and structural characterization, leading to the identification of a novel Ribosomally synthesized peptide targeting the BamA outer membrane protein in Gram-negative bacteria (Imai et al., 2019; Susanto et al., 2026). In the discovery of the taromycins, researchers utilized a metabologenomic approach to link a specific orphan BGC identified in the marine actinomycete Saccharomonospora sp. to its chemical product. By employing the BiG-SCAPE clustering algorithm to identify a GCF with high homology to the daptomycin biosynthetic pathway, and subsequently using mass spectrometry-based molecular networking to identify the corresponding chlorinated lipopeptide, they successfully validated the cluster’s activity. This case study is particularly significant because it utilized CNPS (Center for Natural Product Synthesis) logic to predict the metabolite’s structure in silico before experimental isolation was attempted (Reynolds et al., 2017). In addition, molecular networking–based metabolomics integrated with genome mining approaches has significantly improved functional annotation of cryptic natural products. For example, GNPS-assisted metabolome analysis combined with antiSMASH-based BGC prediction has enabled correlation of orphan gene clusters with previously uncharacterized metabolites in marine actinomycetes and environmental microbiomes, thereby accelerating discovery of novel nonribosomal peptides and polyketides (Wang et al., 2025). These studies collectively demonstrate how iterative integration of metagenomic predictions with metabolomic validation can improve prioritization of bioactive compounds while minimizing rediscovery of known metabolites.

The integration of MS/MS-based molecular networking with genomic data enables the association of metabolite families with their corresponding biosynthetic origins, thereby facilitating the discovery of structurally novel antimicrobial compounds. Such integrative workflows are particularly valuable in plant-associated microbiomes and endophytic communities, where complex ecological interactions drive the production of unique and previously uncharacterized secondary metabolites (Dührkop et al., 2020; Mantri et al., 2021). Collectively, the integration of metagenomic and metabolomic datasets through correlation-based and network-driven frameworks enables functional linkage between biosynthetic gene clusters and their associated metabolites, thereby accelerating resistance-oriented antimicrobial discovery (Figure 2). This iterative genome-to-metabolome framework enhances prediction accuracy, facilitates functional annotation, and minimizes rediscovery of previously characterized metabolites.

Beyond direct pathogen inhibition, integrative omics workflows are increasingly revealing secondary metabolites with immunomodulatory activities that contribute to host-directed antimicrobial responses. This dual functionality, targeting both microbial pathogens and host immune pathways, represents an emerging paradigm in antimicrobial discovery. Several microbial and plant-derived metabolites have been shown to regulate innate immune functions, including macrophage activation, phagocytosis, cytokine signaling, and reactive oxygen species (ROS) production, thereby enhancing pathogen clearance and immune homeostasis. For instance, microbiota-derived indole metabolites modulate host immune signaling through aryl hydrocarbon receptor (AhR)-dependent pathways and influence oxidative stress responses, highlighting the complex interplay between microbial metabolism and host defense mechanisms (Tintelnot et al., 2023). Similarly, recent metabolomics studies have demonstrated that microbiome-associated metabolites can regulate inflammatory signaling and immune homeostasis during infection and disease progression (Mafe and Büsselberg, 2025; Guo et al., 2026). These findings suggest that secondary metabolites may function not only as direct antimicrobial agents but also as modulators of host immunity, thereby expanding their therapeutic relevance beyond conventional bactericidal activity. Integrating immunometabolomics with metagenomics and metabolomics workflows may therefore provide a more comprehensive framework for developing next-generation antimicrobial and resistance-modulating therapeutics.

5 Chemical biology perspectives: targets and resistance-aware strategies

Chemical biology provides the essential toolkit to bridge the gap between genomic potential and therapeutic reality by elucidating the precise molecular interactions between bioactive small molecules and their biological targets. By integrating high-resolution omics technologies with sophisticated chemical probes, researchers can now interrogate the microbial cell at a systems level, identifying noncanonical vulnerabilities that are less prone to existing resistance mechanisms. Table 5 summarizes the major stages of the contemporary antimicrobial discovery pipeline, highlighting the associated omics approaches, computational tools, key strengths, and existing translational bottlenecks. This section explores the convergence of chemical proteomics, structural biology, and systems-level profiling in defining the next generation of antimicrobial targets and the strategies required to bypass the metabolic flexibility of multidrug-resistant pathogens.

TABLE 5. Summary of multi-omics and chemical biology frameworks in antimicrobial discovery.

5.1 Novel microbial and plant-derived targets revealed by omics approaches

The integration of multi-omics datasets has facilitated a transition from a one-drug-one-target model to a systems-level understanding of microbial vulnerability. Traditional antibiotics primarily target a narrow set of essential processes, such as cell wall biosynthesis, translation, and DNA replication, which has led to intense selective pressure and the rapid emergence of resistance (Brown and Wright, 2016). However, transcriptomic and proteomic profiling under various stress conditions have revealed a vast landscape of conditionally essential genes, targets that are critical for survival within the host environment but are not captured during standard in vitro laboratory growth (Farha and Brown, 2019). These include metabolic bottlenecks in central carbon metabolism and the biosynthesis of specialized cofactors, which offer a broader array of targets for structurally novel plant- and microbial-derived metabolites. Genomic and transcriptomic surveys have been particularly effective in identifying targets within specialized microbial processes, such as the regulation of virulence and bacterial proteostasis. For instance, the use of chemical proteomics has identified the Caseinolytic protease P (ClpP) as a high-value target; small molecules like acyldepsipeptides (ADEPs) dysregulate this machinery, leading to the uncontrolled degradation of essential proteins and subsequent cell death (Boutte and Crosson, 2013). Furthermore, omics-driven analysis of extremophilic microorganisms has pointed toward unique metabolic bypasses and specialized stress-response proteins that are absent in temperate pathogens, providing a conceptual basis for highly selective inhibitors that minimize collateral damage to the host microbiome.

In the realm of plant-derived compounds, metabolomics and chemoproteomics are revealing that many natural products do not target a single enzyme but rather disrupt multi-protein complexes or integrated signaling pathways. For example, several plant-derived polyphenols and alkaloids have been shown to target the FtsZ protein, a highly conserved tubulin homolog essential for bacterial cell division. By employing activity-based protein profiling (ABPP) alongside RNA-seq, researchers have discovered that these compounds often exert pleiotropic effects, simultaneously inhibiting cytokinesis while upregulating oxidative stress pathways, a multi-hit mechanism that significantly lowers the frequency of resistance development (Farha and Brown, 2019; Sass, Microbiology and 2013). The study of the microbial resistome through metagenomic sequencing has enabled the identification of pre-existing resistance determinants, allowing for the proactive design of inhibitors against these specific targets. By analyzing the structural biology of resistance enzymes (e.g., modified penicillin-binding proteins or aminoglycoside-modifying enzymes), medicinal chemists can utilize structure-guided design to create resistance-proof molecules or decoys that restore the efficacy of existing antibiotics (Huang and Zhou, 2025). This synergy between omics-based target discovery and chemical biology not only expands the druggable space within the bacterial cell but also ensures that the discovery process is inherently aware of potential resistance trajectories from the outset (Tyers and Wrights, 2019).

5.2 Mechanism-oriented antimicrobial discovery

The transition from traditional black-box phenotypic screening to mechanism-oriented discovery is a cornerstone of modern chemical biology. While phenotypic assays confirm that a compound kills or inhibits a pathogen, they offer little insight into the molecular target or the physiological changes leading to cell death. Deconvoluting the Mode of Action is essential for lead optimization, predicting potential cross-resistance, and ensuring the therapeutic index is sufficient for clinical progression (Farha and Brown, 2019). Today, this deconvolution is increasingly achieved through the integration of high-resolution omics assays that provide a global view of cellular perturbation. One of the most powerful tools for direct target identification is Thermal Proteome Profiling (TPP). By exploiting the principle that proteins become more thermally stable when bound to a ligand, TPP allows for the proteome-wide assessment of drug-target engagement in situ (Mateus et al., 2017). Recent advancements, such as Membrane-Mimetic TPP (MM-TPP), have addressed the historical difficulty of profiling transmembrane proteins, a critical class of targets for many antimicrobials derived from plant and microbial sources (Jandu et al., 2025). The primary limitation of standard TPP lies in the inherent instability of membrane proteins once they are liberated from the lipid bilayer. During the heating phase, these proteins typically undergo rapid irreversible aggregation and precipitation, regardless of ligand binding. MM-TPP overcomes this by utilizing specialized non-denaturing detergents or synthetic membrane mimetics (such as styrene-maleic acid lipid particles, SMALPs or lipid nano discs or amphipols) that maintain the native fold and structural integrity of transmembrane domains during thermal challenge. By stabilizing the hydrophobic surfaces of these proteins, MM-TPP allows for the detection of ligand-induced thermal shifts in targets like efflux pumps, transporters, and cell-wall synthases proteins that were previously invisible to soluble proteomic assays (Mateus et al., 2018; Pazicky et al., 2026). Furthermore, top-down TPP (TD-TPP) now allows for the characterization of intact proteoforms and their stability shifts under antibiotic stress, providing a more granular view of target engagement and downstream metabolic consequences (Mateus et al., 2018; Pazicky et al., 2026).

Complementary to TPP, CRISPR interference (CRISPRi) provides a chemical-genetics framework for identifying the genetic basis of drug sensitivity. By titrating the expression of specific genes, researchers can identify vulnerability hubs where a minor genetic knockdown significantly sensitizes the bacteria to a specific chemical scaffold. This approach has been particularly successful in defining the MoA of novel inhibitors in difficult-to-treat pathogens like M. tuberculosis and Acinetobacter baumannii (Peters et al., 2016). When paired with comparative genomics, CRISPRi reveals not only the primary target but also the intrinsic resistance factors that the bacteria upregulate to survive the chemical challenge (Peters et al., 2016). Metabolomic fingerprinting offers a rapid, real-time readout of the physiological state of the cell post-treatment. Advanced analytical platforms like KINEXO (Kinetics of Exometabolites) utilize NMR or Mass Spectrometry to monitor the secretion of key metabolites, providing a unique metabolic signature that can be matched against a library of known antibiotics to determine MoA novelty (Zampieri et al., 2018). These time-resolved metabolomics studies are particularly adept at differentiating between compounds that disrupt membrane integrity and those that inhibit specific enzymatic steps in central metabolism or biosynthetic pathways (Wolfender et al., 2019). Together, these integrative pipelines move the discovery process beyond mere observation toward a deep, structural, and physiological understanding of antimicrobial action.

5.3 Chemical strategies to overcome resistance: multi-targeting and synergism

The rapid evolution of AMR has rendered the one-drug-one-target approach increasingly vulnerable. To counter the genetic plasticity of pathogens, chemical biology has pivoted toward strategies that increase the genetic barrier to resistance. This is achieved through the design of multi-targeting agents and the implementation of synergistic combinations, which force the bacteria to acquire multiple simultaneous mutations to survive (Tyers and Wright, 2019; Walsh and Wencewicz, 2020). By hitting several nodes in a biological network, these strategies not only enhance potency but also restore the efficacy of legacy antibiotics that have been sidelined by widespread resistance. Multi-targeting strategies involve the development of single small molecules capable of inhibiting two or more essential enzymes. A classic example is the dual inhibition of DNA gyrase and topoisomerase IV, which significantly reduces the frequency of spontaneous resistance compared to targeting either enzyme alone (Lupien, 2015; Terreni et al., 2021). From the perspective of natural products, many plant-derived compounds such as certain flavonoids and terpenoids exhibit inherent multi-target activity, simultaneously disrupting membrane potential and inhibiting efflux pumps (Farha and Brown, 2019; Hogan et al., 2024). This pleiotropic effect is a natural evolutionary adaptation by plants to combat a broad spectrum of microbial attackers and serves as a blueprint for the design of synthetic hybrid antibiotics (Bollenbach, 2015).

Synergism, the use of two or more agents that together produce an effect greater than the sum of their individual parts, offers a powerful tool for drug repurposing. One of the most successful synergistic strategies is the use of antibiotic adjuvants. These molecules may not possess intrinsic antimicrobial activity but instead inhibit resistance determinants, such as β-lactamases or efflux pumps, thereby potentiating the primary antibiotic (Dhanda et al., 2023; Markley et al., 2019). Recent metagenomic surveys of extremophilic niches have identified novel classes of metallo-β-lactamase inhibitors, such as aspergillomarasmine derivatives, which can restore carbapenem susceptibility in multidrug-resistant Gram-negative pathogens (Rahimian et al., 2026; Gonzalez-Bello et al., 2019). The emergence of collateral sensitivity, where the development of resistance to one drug increases susceptibility to another, is being exploited to design sequential or combination therapies. Machine learning models trained on largescale chemogenetic datasets can now predict these sensitivity profiles, allowing for the rational selection of synergistic pairs that trap the bacteria in an evolutionary dead-end (Barbosa et al., 2018; Lázár et al., 2018). Coupled with metabolomics-driven discovery, these strategies are enabling the identification of metabolic bypass inhibitors that prevent the bacteria from utilizing alternative nutrient sources when primary pathways are blocked (Mohiuddin et al., 2020; Aulitto and Gallo, 2025). By integrating these chemical and evolutionary strategies, the next generation of therapies will be inherently designed to manage, rather than just treat, the emergence of resistance.

5.4 Role of chemical probes, activity-based profiling, and chemoproteomics

The final frontier in mechanism-oriented discovery is the ability to visualize and quantify drug-target interactions within the native cellular environment. Traditional biochemical assays often fail to account for the complex physiological factors, such as local pH, metabolite competition, and protein-protein interactions, that govern how an antimicrobial actually performs inside a bacterium. To bridge this gap, chemical biology utilizes chemical probes, which are small molecules designed to selectively bind and report on the activity of specific protein targets (Wright and Sieber, 2016). Activity-Based Protein Profiling (ABPP) has emerged as a cornerstone of this effort. Unlike standard proteomics, which measures total protein abundance, ABPP uses activity-directed probes to react only with the functional (active) form of enzymes (Berger et al., 2004). This is particularly critical in antimicrobial research, where many enzymes involved in cell wall synthesis or virulence are regulated post-translationally. By employing click-chemistry compatible probes, researchers can label active targets in live bacterial cultures and subsequently use mass spectrometry to identify exactly which enzymes are being inhibited by a novel plant or microbial extract (Hogan et al., 2024; Weis et al., 2022). Beyond simple target identification, chemoproteomics allows for the mapping of the off-target profile of a compound, which is essential for predicting toxicity and side effects. For example, competitive ABPP can reveal if a promising antimicrobial also binds to human homologs, providing an early-stage assessment of the therapeutic window (Mateus et al., 2020; Parker and Pratt, 2020). In the study of extremophile-derived metabolites, which often possess unique and reactive chemical handles, chemoproteomics has identified non-canonical targets such as riboswitches and orphan transporters that were previously invisible to genomic mining alone (Rahimian et al., 2026; Gavriilidou et al., 2022).

A critical dimension in the modern discovery story is the proactive assessment of resistance, where resistance evolution is investigated before the clinical deployment pipeline rather than retrospectively after resistance emergence. While resistance is traditionally studied as a reactive consequence of drug use, the modern discovery paradigm utilizes Proactive Resistomics to anticipate bacterial adaptation (Tyers and Wright, 2019). This involves using chemoproteomic tools to monitor how the bacterial active enzyme shifts when exposed to sublethal concentrations of a candidate molecule. Mechanistically, this involves using broadly reactive or suitespecific chemical probes to capture functional changes that total proteomics might miss, such as the posttranslational activation of latent bypass enzymes or the subtle upregulation of specialized efflux transporters. By identifying these early warning compensatory pathways, such as the recruitment of alternative transpeptidases or metabolic bypasses in central carbon metabolism, researchers can implement a resistanceaware design strategy (Mishra et al., 2023; Nimmana and Nguyen, 2026). This allows for the development of evolution-proof therapies, where the primary antimicrobial lead is paired with a secondary inhibitor specifically designed to block the identified compensatory nodes. By targeting both the primary enzymatic target and the anticipated evolutionary escape route simultaneously, this integrated chemoproteomic framework significantly raises the genetic barrier to resistance, ensuring the long-term therapeutic viability of novel scaffolds (Vincent et al., 2009; Gavriilidou et al., 2022; Alatawi et al., 2025). Collectively, the integration of ABPP, chemoproteomics, and proactive resistomics establishes a predictive framework for next-generation antimicrobial discovery, supporting the development of durable therapeutics capable of remaining effective under sustained evolutionary pressure.

6 Translational challenges, bottlenecks and future perspectives

The integration of multi-omics and chemical biology into antimicrobial discovery has expanded access to previously unexplored biosynthetic diversity in medicinal plants and extreme environments. However, significant technical and translational challenges persist. A major limitation lies in the heterogeneity of multiomics data, where genomic datasets are standardized, but metabolomic data remain highly variable due to differences in extraction methods, instrumentation, and environmental factors. This results in batch effects and poor cross-study comparability. Additionally, the lack of comprehensive databases for extremophiles and non model systems leads to a large proportion of unannotated metabolites, often referred to as the dark metabolome (Alseekh et al., 2021). Addressing these issues requires improved normalization strategies and standardized metadata frameworks. Beyond data integration, linking BGCs to functional metabolites remains challenging, particularly because BGCs are often silenced under laboratory conditions. (Farha and Brown, 2019) From a clinical standpoint, future discovery must move towards Proactive Resistomics using activity based probes to monitor shifts in the bacterial active enzymes during sub-lethal drug exposure, allowing researchers to design evolution-proof therapies that pre-emptively target escape mechanisms (de la FuenteNunez et al., 2017). Antimicrobial resistance (AMR) further complicated the translational success of newly discovered compounds. Disease causing pathogens can rapidly develop resistance mechanisms, thereby limiting long-term therapeutic efficacy. Therefore, integrated multi-omics approaches involving metagenomics, metabolomics, transcriptomics, proteomics, and systems biology are increasingly being explored to better understand microbial interactions, resistance pathways, and mechanisms of action (Crüsemann et al., 2017) Emerging artificial intelligence (AI) approaches are transforming this landscape by enabling predictive modelling of bioactivity and structure. Tools such as graph neural networks and AlphaFold-based workflows allow in silico prioritization of candidate molecules, supporting a more efficient “fail-fast” discovery model (Jumper et al., 2021). CRISPR-mediated activation of silent biosynthetic gene clusters, and engineered heterologous expression systems are facilitating the recovery and characterization of cryptic natural products (Ziemert et al., 2016). Similarly, molecular networking platforms such as GNPS and genome mining tools such as antiSMASH and MIBiG are enabling improved integration between metagenomic and metabolomic datasets. Perhaps most transformative is the transition towards personalized antimicrobial discovery. By utilizing real-time metagenomic profiling of an individual’s unique microbiome, clinicians could deploy narrow-spectrum, precision metabolites that selectively eliminate pathogens while preserving commensal microflora (Yaqub et al., 2025). This precision biotics approach minimizes the collateral damage to the host’s immune landscape and reduces the broad-spectrum selective pressure that historically accelerates the spread of AMR.

Future antimicrobial discovery is expected to increasingly rely on systems-level and evolution-informed strategies that integrate real-time resistome surveillance, multi-omics approaches, synthetic biology, and AI-driven design to enable targeted, predictive, and proactive development of next-generation therapeutics (Zhong et al., 2026). In this context, future research should focus not only on the discovery of novel biosynthetic pathways and bioactive metabolites but also on overcoming translational barriers through scalable fermentation systems, synthetic biology-assisted pathway engineering, toxicity reduction, and pharmacokinetic optimization (Farha and Brown, 2019). The integration of artificial intelligence with metagenomics and metabolomics is anticipated to accelerate metabolite annotation, target prediction, molecular optimization, and resistance forecasting, thereby improving the efficiency of antimicrobial discovery pipelines (Wolfender et al., 2019). Simultaneously, real-time monitoring of resistome dynamics and microbial adaptive responses may facilitate the development of evolution-aware therapeutics capable of pre-emptively targeting resistance mechanisms. Achieving successful bench-to-bedside translation will ultimately require strong interdisciplinary collaboration among microbiologists, chemists, computational biologists, pharmacologists, clinicians, and pharmaceutical industries. Such integrated and translationally oriented efforts are essential to bridge the gap between laboratory discovery and clinical application in the global fight against antimicrobial resistance (Atanasov et al., 2021).

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Yu Pi, Chinese Academy of Agricultural Sciences, China

Reviewed by: Trideep Saikia, Girijananda Chowdhury University, India

Sambamoorthy Unnam, Sree Dattha Institute of Pharmacy, India