Hydrogels and artificial intelligence (AI) in tissue engineering and regenerative medicine (TERM): an update on ‘skin and bones’

1 Introduction

Health conditions that require tissue regeneration, including dermal (skin) and musculoskeletal (MSK) diseases, represent significant global healthcare burdens. As we age, the prevalence of the need for regenerative interventions continues to rise in both skin-related challenges and MSK diseases, intensifying pressure on healthcare systems worldwide due to our globally ageing population (World Economic Forum, 2025). This underscores the urgent need for innovative therapeutic strategies to effectively restore tissue function, rather than merely managing symptoms (World Economic Forum, 2025; Lysaght et al., 2008). Despite major advances, current clinical approaches to repairing or replacing damaged tissues remain limited by donor shortages, immune rejection, and poor long-term integration (Felix et al., 2025). Traditional approaches often fail to replicate the complex physiological microenvironments required for functional regeneration in skin and bones (Meretsky et al., 2024).

Tissue engineering and regenerative medicine (TERM) have thus emerged as critical fields for addressing the escalating socioeconomic impact of these conditions (Hoang et al., 2025; Olson et al., 2011). TERM combines biomaterials, cells, and molecular signals to create constructs or scaffolds that restore, maintain, and enhance tissue function in several organs, including skin and bone (Esdaille et al., 2021; Ferraz, 2024; Nourian Dehkordi et al., 2019). A wide variety of methods and biomaterials may be used to fabricate these constructs, including electrospinning, freeze-drying, solvent casting and self-assembly (Leong et al., 2008; Socci et al., 2023; Bakhshandeh et al., 2017). Among these fabricated structures, hydrogels have emerged as a promising tool for TERM and clinical applications for skin and bone in the last few decades (Lee and Mooney, 2001; Zhu and Marchant, 2011; Xu et al., 2022). While recent years have seen significant advances in the field, several challenges persist, including scalability, integration of vascularisation and the difficulty of reproducing native extracellular matrix (ECM) complexity in vitro (Ronfard et al., 2017; Williams, 2019; Somasekhar and Griffiths, 2023). These limitations reinforce the need for advanced fabrication technologies capable of producing physiologically relevant and biologically functional constructs for patients (Grounds, 2018; Naderi et al., 2011).

Three-dimensional (3D) printing and bioprinting have transformed the landscape of TERM by enabling precise, customizable fabrication of scaffolds and hydrogels (Uysal et al., 2025; You et al., 2018; Agrawal and Hussain, 2023). Unlike traditional scaffold fabrication methods, 3D bioprinting allows optimised cellular arrangements, biomaterials, and growth factors, facilitating the creation of biomimetics that can better mimic the physiology of the target tissue, both structurally and functionally (Koumentakou et al., 2024; Fang et al., 2023a; Advincula et al., 2021). The method has demonstrated reproducibility, adaptability towards patient-specific design, and the possibility to integrate multiple cell types within heterogeneous microenvironments (Jammalamadaka and Tappa, 2018; Tan et al., 2021; Jiang et al., 2019). This positions 3D bio-printed hydrogels as a powerful platform for engineering tissues with enhanced biological relevance, offering new opportunities for personalised regenerative therapies and precision medicine approaches (Cadamuro et al., 2023; Douglas et al., 2025; Fu et al., 2023).

Artificial intelligence (AI) is now accelerating progress in 3D bioprinting by addressing design optimisation, process control, and quality assurance based on the target tissue. AI-driven algorithms can enable predictive modelling of mechanical performance and cell viability, reducing trial-and-error experimentation and optimising bioink formulations (Gharibshahian et al., 2024; Bagherpour et al., 2025). Machine learning (ML) approaches have been applied to automate print parameter selection, improve structural accuracy, and detect defects in real time, enhancing reproducibility and reducing variability across scaffold fabrication (Zhang et al., 2024; Tănase, et al., 2025). As AI continues to advance, its integration with 3D bio-printed hydrogels promises to accelerate clinical translation by enabling smarter, reproducible, and predictive modelling of hydrogel behaviour (Zhang et al., 2025a; Bediaga-Bañeres et al., 2025; Logeshwaran et al., 2024).

As authors of this article, we want to focus on the novelty in our integrative approach, that uniquely combines AI with the techniques of hydrogel-based biomaterials and 3D printing for tissue engineering. While conventional studies have treated biomaterials and computational tools separately, we want to highlight on studies that have investigated at the intersection of the cutting-edge technologies for enhanced patient care. We bring together AI-driven design optimisations and the potential of personalisation of hydrogels across the entire regenerative pipeline for better patient outcomes. With this review, we thus aim to converge the current knowledge in the unique field of hydrogels, AI, skin, and bone TERM, and highlight emerging innovations to encourage future interdisciplinary collaborations. We also identify the current translational gaps and propose future frameworks for data-driven, patient-specific regenerative therapies. Finally, we outline how AI-driven hydrogel systems may improve wound healing, bone repair, and take a step towards a precision medicine approach that ultimately enhances clinical outcomes for these patients.

2 Methodology

Our search began with the key words, ‘tissue engineering’, ‘artificial intelligence’, ‘hydrogels’ and ‘3D printing’ for literature search. These were further narrowed down by adding either ‘bone’ or ‘skin’ to the search for the respective sections below. Both PubMed and Google scholar were used as platforms for extraction of information on these specific areas of research. Further, we included full texts, focussed on literature within the last decade, with research advancements since 2020. Review articles, research articles, and editorials primarily focusing on MSK diseases and skin were included. Publications that did not have full text were excluded. Articles discussing other organ/tissue systems or in combination with bone or skin were also excluded. Quality of publications used was checked by all the authors writing this manuscript (KH, YP, NR and PG) and additionally critically reviewed by co-author ML. We are aware that despite our best efforts of conducting a comprehensive literature search, we cannot completely ignore the risk of bias. It is possible that the strategy we employed for our manuscript and the selections we made using the keywords, inclusion/exclusion criteria and our interpretation that sculpted the article, may have overlooked some of the studies in the field. We do not claim that this review is exhaustive but rather a reflection of our judgement with the resources available to us, during manuscript preparation.

3 3D printing and AI in bone regeneration

Bone regeneration and repair following a fracture or surgery are long-winded processes with several stages involved and multiple considerations, including mechanical load, pain management, healing and restoring functional efficiency. These are closely regulated biological processes that restore the structural integrity and functionality of bones after injury (Zhang et al., 2025; Dimitriou et al., 2011; MA et al., 2023). It progresses through the different stages of inflammation, soft callus formation, hard callus formation, and finally remodelling. Inflammation stage recruits immune cells and growth factors, while progenitor cells, known as bone marrow mesenchymal stem/stromal cells (BM-MSCs) differentiate into chondrocytes and osteoblasts to form new matrix (Bahney et al., 2019; Xu et al., 2023b). Mineralisation follows this step, eventually restoring the strength, shape, and functionality of bones. Tissue engineering (TE) is increasingly being optimised and utilised as a framework to provide stability and a platform for new bone formation, especially after large bone defects (Iqbal et al., 2024; Premjit et al., 2025; Xue et al., 2022). Among the several TE methods, 3D printing has emerged as a method with exponential interest due to the ease of tunability it offers for scaffold optimisations.

3D printing is known to be adaptable and enables patient-specific solutions for large bone defects. The technology allows fabrication of scaffolds that match the biological target site and mimic bone’s porous structure, thereby supporting cell attachment, vascularisation, and tissue growth (Mayfield et al., 2022; Wang et al., 2024). Desired materials can be optimised to provide appropriate mechanical strength, biodegradability, and bioactivity. Additionally, 3D printing enables controlled delivery of growth factors, cells and drugs like antibiotics to prevent post-surgical infections and enhance healing outcomes (Bandyopadhyay et al., 2020; Samaila et al., 2020; Bai et al., 2025). Compared with traditional grafts, it reduces donor-site morbidity, improves reproducibility, and offers patient-centric interventions for large or irregular bone defects (Brachet et al., 2023; Yaneva et al., 2023).

The advent of AI opens an entire world of possibilities for the convergence of 3D printing and bone tissue engineering (BTE). AI can be applied to the different complexities and the various stages of BTE. AI can help optimise BTE by enabling evidence-based design, optimisation and prediction across multiple stages (Bagherpour et al., 2025; Ganguly, 2026). ML models can design scaffold architectures with desired porosity, mechanical strength and degradation profiles to provide platforms best suited to patient needs. AI will also be able to aid biomaterial selection and formulation by predicting cell-material interactions, behaviour, and toxicity (Kolomenskaya et al., 2023; Mackay et al., 2021). The next section explores the structure of the bone and its micro-environment to describe the complexity of the dynamic tissue.

3.1 Bone structure and its micro environment

The bone is one of the most dynamic organs in the human body, forming the base of our skeletal system. The bone is hard on the exterior and has the bone marrow (BM) inside, which is the site of blood formation (haematopoiesis) and serves as a reservoir for stem cells, immune cells, and growth factors (Yin and Li, 2006; Kandarakov et al., 2022). The BM’s unique microenvironment is often referred to as the BM niche, and it houses the two major types of progenitors or stem cells: hematopoietic stem cells (HSCs) and mesenchymal stem/stromal cells (MSCs). The HSCs further differentiate into common lymphoid cells and common myeloid cells. Lymphoid cells then differentiate into B cells, T cells and natural killer cells (NK cells); while myeloid cells further differentiate into granulocytes, monocyte/macrophages, megakaryocytes and erythrocytes as shown below in Figure 1 (Kandarakov et al., 2022; Lucas, 2021; Furuhashi et al., 2025). BM-MSCs are the progenitors of bone, fat and cartilage that differentiate into osteoblast, chondrocyte, and adipocyte, respectively. Osteoblasts further differentiate into mature osteoblasts and finally into osteocytes that form part of the bone matrix (Deb, 2019; Gao et al., 2022).

The BM niche is where all the inter-cellular interactions take place to maintain bone homeostasis. These components within the BM regulate stem-cell fate, quiescence, differentiation, and migration. The balance between bone formation and bone erosion, also known as bone remodelling, maintains bone homeostasis (Kenkre and Bassett, 2018; Bolamperti et al., 2022). New bone is formed from osteoblasts that eventually become mature osteocytes, cells known for their mechano-sensing properties. Osteoclasts, on the other hand, are the bone-eating or bone-eroding cells that are derived from the monocyte-macrophage lineage of cells. Normally, the amount of bone eroded by osteoclasts is balanced by new bone formed by osteoblasts (Wu et al., 2019; Kim et al., 2020). However, when the bone formation rate does not match the rate of bone erosion, bone loss becomes prevalent. This can be seen in age-related bone loss and in diseases such as osteoporosis (OP) (Ganguly et al., 2017; Smit et al., 2024). Large bone defects present a much more complex problem altogether. Degenerative diseases like osteoarthritis (OA) are also a challenge due to limited approaches in treatment, including medications for symptom management or surgery, which have a tremendous impact on a patient’s quality of life (QOL). This has increased the need for BTE in bone diseases and age-related bone loss (Liang et al., 2026; Sun et al., 2026). However, given the complexity and unique architecture of the bone as described above, creating a biomimetic platform that truly reflects the bone’s intricacies and the BM’s complexity will come at the intersection of BTE, patient-derived cells and strategies optimised for bone scaffolds (Doherty-Boyd et al., 2024; Rausch et al., 2023).

3.2 Need for BTE in bones

For bone conditions/diseases that require surgery, BTE is increasingly a popular alternative to traditional surgical approaches. While bones can regenerate, healing in case of degenerative diseases and trauma-based defects, is often impaired, resulting in large defects, non-unions and related complications. Traditional approaches, including autografts and allografts, face challenges such as donor-site morbidity, limited graft availability, immune rejection, infection risk, and inconsistent outcomes (Ganguly et al., 2022; Stahl and Yang, 2021). BTE has the potential to overcome these limitations by combining biomaterials, cells, and bioactive cues to promote controlled and predictable bone regeneration.

Engineered scaffolds optimised for the patient’s target site can provide structural support, guide tissue formation, and deliver cells, drugs and/or growth factors while gradually degrading as new bone forms (Ganguly et al., 2021b; Yildizbakan et al., 2023; Kumar et al., 2026). This framework is referred to as the diamond concept, which includes a mechanical substrate or environment, osteoconductive scaffolds (made from suitable biomaterials), osteogenic cells (like MSCs), and appropriate growth factors like bone morphogenic protein 2 (BMP2) and vascular endothelial growth factor (VEGF) to enable bone formation or osteogenesis. (Giannoudis et al., 2007; Andrzejowski and Giannoudis, 2019). Vascularity and consideration for host factors are increasingly being incorporated within this diamond concept, to acknowledge the challenges in replicating bone’s vasculature and the host’s microenvironment. Importantly, BTE supports the shift toward patient-specific regenerative solutions that are not only minimally invasive but also reduce dependence on implants and follow-up surgeries.

Over the past decade, 3D bioprinting has gained significant global attention across biological, medical, and pharmaceutical research. This technology focuses on the fabrication of complex biological structures using bioinks, which typically consist of living cells, biomaterials, and signalling molecules. Similar to conventional 3D printing, bioprinting relies on computer-aided design (CAD) and automated systems to deposit materials layer-by-layer (L-b-L) with high precision (Ansaf et al., 2023; Ng et al., 2019). The overall process involves four key stages: acquisition of biological data, bioink material selection, conversion into a digital model to guide the printing process, and subsequent functionalisation-fabrication of stable, tissue-like constructs (Ansaf et al., 2023; Olejnik et al., 2022; Gu et al., 2020).

Bioprinting, a specialised form of additive manufacturing (Weng et al., 2021), enables the L-b-L deposition of cells (Beheshtizadeh et al., 2020) along with scaffold materials, particularly hydrogels, in a controlled spatial and temporal manner (Weng et al., 2021). Among the various techniques, extrusion-based, inkjet, laser-assisted, and stereolithography-based bioprinting are most employed (Tarassoli et al., 2018). The choice of technique depends on material properties, resolution requirements, and desired tissue complexity (Singh et al., 2016; Weng et al., 2021). Overall, BTE addresses unmet clinical needs by enabling effective repair of complex bone defects, improving bone repair outcomes, and supporting long-term skeletal function and QOL of patients (Iqbal et al., 2023; Kostadinova et al., 2025).

3.3 3D printing hydrogels and materials for bone regeneration

In the past decade, 3D printing and bioprinting have been pivotal in BTE for fabricating scaffolds with precise architectural requirements and defect-specific geometries. This is relevant to hydrogel-based systems, as they provide a biocompatible environment that resembles the ECM. However, these systems often need improvements in their structure and composition to enhance printability, mechanical stability, and osteogenic functionality, all of which are crucial for effective bone repair. Recent studies have shifted towards hybrid bioinks and composite scaffolds that combine hydrogels with ceramics (Sajad Daneshi et al., 2024) like β-TCP/collagen composite that are known to enhance osteogenic differentiation and vascularisation by facilitating bone cell ingrowth and neovascularization. Hydrogel scaffolds with 3D bioprinted gelatin methacrylate (GelMA) modified with luteolin-loaded ZIF-8 (LUT@ZIF-8) nanoparticles demonstrated osteogenic, immuno-modulatory and antibacterial activity in rat model in vivo (Yu et al., 2025). 3D printed Hydrogels crosslinked with polymers like polyacrylamide (PAAM), polyurethane (PU) and Gelatin (Gel), displayed desirable osteogenic capabilities in animal models (Gao et al., 2023a; He et al., 2021). Similar hydrogel scaffolds with bioactive compounds (Lee et al., 2022; Wu et al., 2019; Jalilinejad et al., 2025) have been evidenced to improve osteoconductivity and vascularisation. While the exact mechanistic pathways of these studies are yet to be explored; the evidence demonstrated strong potential of the application of these hydrogels for enhanced bone repair and regeneration.

There has also been a transition to multifunctional printed systems designed specifically to address the key problems of bone regeneration, including poor mechanical strength (Jalilinejad et al., 2025; ElShebiney et al., 2025), insufficient vascularisation (Gao et al., 2023a; Zhou et al., 2025; Yang et al., 2024), and limited osteoinductive activity (Yu et al., 2025; Gao et al., 2023b). These approaches include mineral-reinforced hydrogels, cell-laden bioprinted structures, systems for the release of growth factors or pharmaceuticals, and bioinks tailored for specific defect scenarios, such as cranial, calvarial, maxillofacial, and long-bone repair. Specifically in the use of polysaccharides, the molecular weights can have a direct impact on the functions. Low molecular weight (LMW) polysaccharides are often used as an additive in BTE and can promote cell migration, proliferation and osteogenic differentiation. LMW hyaluronic acid fragments act as damage-associated molecular patterns (DAMPs) that activate wound healing (Ferreira et al., 2022). High molecular weight (HMW) polysaccharides on the other hand are known to provide enhanced structural support, better viscoelasticity and mechanical stability and degrades at a lower rate (Ho et al., 2021). This is usually used to ensure a longer lasting tissue construct support. While these will vary depending upon the actual biomaterial, it is imperative that molecular weights of polysaccharides are taken into consideration for TERM applications. Supplementary Table 1 provides a list of polysaccharides, their general use, mechanism of action, advantages and disadvantages.

Table 1 summarises relevant studies published since 2020 for 3D-printed hydrogel-based and related composite systems for bone regeneration. We explore the diversity of hydrogel formulations, integrated material phases, printing methodologies, cell lines and factors, target bone applications, and principal findings of studies.

TABLE 1. 3D-printed hydrogel-based systems for bone regeneration and BTE.

3.4 3D printing and AI in bone TERM

Recent advancements in BTE have progressively integrated 3D printing with computational techniques to enhance scaffold design and formulation selection. In this broader context, current applications of AI have predominantly focused on predicting printability and screening formulations (Chen et al., 2023; Nadernezhad and Groll, 2022), optimising printing parameters and quality (Madadian Bozorg et al., 2023 and Rafieyan et al., 2024), designing scaffolds by their mechanical properties (Liu et al., 2023b; Wu et al., 2024a; Drakoulas et al., 2024; Wang et al., 2025); and monitoring processes during 3D printing (Barjuei et al., 2024). Nevertheless, there have been comparatively few studies that have employed AI to predict bone regeneration outcomes, including osteogenesis, vascularisation, or healing of defects in vivo (Logeshwaran et al., 2024; Li et al., 2025).

ML has mostly been employed to minimise empirical screening by identifying correlations among material composition, rheology, and printability. Nadernezhad et al. employed an interpretable ML framework utilising rheological descriptors from 180 hydrogel formulations to determine key characteristics associated with printability. They demonstrated that printability can be understood through collaborative rheological assessments rather than a single universal threshold (Nadernezhad and Groll, 2022). In a similar vein, Chen et al. developed decision-tree, random-forest, and deep-learning models from 210 direct-ink-writing formulations containing 16 biomaterials and four solvents. All models achieved accuracies exceeding 80%, with the random forest model demonstrating the best classification performance, while deep learning produced a more refined printability map (Chen et al., 2023). By training a multilayer perceptron to predict storage and loss moduli (G′ and G″) of 3D-printed polyacrylamide hydrogels from seven formulation variables, Mohammad et al. went beyond printable versus non-printable classification. The research further used variational autoencoder and conditional variational autoencoder models to generate candidate material compositions for target rheological outputs (Mohammad et al., 2024). Collectively, these investigations support the application of AI as a formulation-screening tool in hydrogel and hybrid biomaterial systems. This approach is particularly efficient given the numerous interacting variables that render the conventional trial and error development ineffective (Chen et al., 2023; Nadernezhad and Groll, 2022).

A similar, but separate, application pertains to the optimisation of the printing process itself. Bozorg et al. used design of experiments, image-based non-destructive characterisation, and ML to investigate the pneumatic semisolid extrusion of model biomedical gels. The study examined the impact of printing speed, pressure, and infill percentage on quality characteristics like weight, surface area, and heterogeneity. Their findings indicated that ML may offer a more flexible and adaptable framework than conventional design of experiments alone for identifying robust operating windows in soft-material printing (Madadian Bozorg et al., 2023). Rafieyan et al. extended this process-oriented perspective by presenting a large open-source dataset of 1,171 3D-printed scaffolds. The research included various ML techniques to predict printability, scaffold integrity, and cellular response (Rafieyan et al., 2024). The relevance of this study is that it demonstrates how larger datasets can facilitate more systematic prediction of manufacturing outcomes in tissue-engineering scaffolds, even though it was broader in scope than bone-specific bioprinting. These investigations collectively suggest that AI is progressively beneficial for enhancing process robustness and alleviating experimental stress; nevertheless, they do not yet demonstrate a direct correlation between optimal printing conditions and enhanced bone-regeneration efficacy.

Additionally, AI is becoming increasingly important in the design of scaffold architecture, particularly for replicating the mechanical properties of bone. Liu et al. introduced a back-propagation neural network integrated with a regenerative genetic algorithm for the inverse design of anisotropic (Triply Periodic Minimal Surfaces) TPMS bone scaffolds based on specific targeted elastic parameters. They reported average errors of less than 3% for three design targets and approximately 5% for six targets (Liu et al., 2023a). Drakoulas et al. also established a probabilistic reduced-order framework to evaluate scaffold design under coupled mechanical loading and fluid flow conditions. The objective was to optimise the scaffold surface area linked to bone-cell differentiation while minimising other differentiation states (Drakoulas et al., 2024). In a related yet fabrication-focused study, Wu et al. employed Bayesian optimisation to conduct mechano-biological optimisation of functionally graded ceramic TPMS scaffolds. This was designed for a sheep tibia segmental-defect scenario, subsequently followed by lithography-based ceramic manufacturing and geometric validation of the printed structures (Wu et al., 2024b). Wang et al. demonstrated that a data-driven inverse-design framework could generate spinodoid scaffolds with elastic constants and anisotropies that closely aligned with those of the target bone. This approach combined a back propagation neural network with a genetic algorithm, and the results were subsequently validated through finite-element analysis and compression testing (Wang et al., 2025). Benaatou et al. have recently expanded AI-guided design from bulk architecture to implant functionalisation. They employed an AI-guided workflow to select a surface functionalization approach for 3D-printed Polyetheretherketone (PEEK) implants coated with a tryptophan-enriched alginate hydrogel. The resulting construct exhibited improved cytocompatibility, reduced roughness and friction, increased hydrophilicity, antibacterial activity, sustained-release behaviour, and favourable stress distribution, as determined by finite-element analysis. Nevertheless, in vivo validation of osseointegration was not reported (Benaatou et al., 2026). The studies illustrate that AI can facilitate the efficient investigation of scaffold structures that would be challenging to identify through conventional iterative design methods. The primary outcomes continue to be mechanical, structural, or mechanobiological rather than definitive evidence of improved bone healing in vivo (Liu et al., 2023a; Wu et al., 2024a; Drakoulas et al., 2024; Wang et al., 2025).

Barjuei et al. developed a computer vision-based approach for robotic bioprinting that detected discrepancies between deposited filaments and the planned trajectory, thereby facilitating the generation of a corrected tool path and enhancing geometric precision in printed vessel-like structures (Barjuei et al., 2024). Although this was not a bone-specific study, it is relevant to bone bioprinting because small deviations in strand placement, pore geometry, or layer fidelity can influence mechanical performance and local biological behaviour in printed scaffolds (Barjuei et al., 2024; Logeshwaran et al., 2024). Li et al. recently created a deep-learning model (OIODNet) that accurately predicts the osteogenic potential of orthopaedic implant surfaces based on early cell morphology with high accuracy, demonstrating the feasibility of incorporating AI into early biological responses rather than solely for printing or mechanical properties (Li et al., 2025). This approach was developed specifically for implant surface evaluation and screening, rather than for comprehensive 3D-printed bone scaffold systems. The current literature indicates that AI in bone-related 3D printing is most established in formulation screening, process optimisation, and scaffold design, while its application to direct prediction of regenerative outcomes remains at an earlier stage (Ganguly, 2026). Thus, the future of the intersection of AI, BTE and TERM holds great promise in the various parameters for BTE, ultimately aimed at enhanced patient outcomes and clinical impact (Figure 2).

4 3D printing and AI in skin regeneration

The skin, the largest organ of the human body, is complex and anisotropic, and its function is entirely dependent on the highly ordered arrangement of secondary protein structures within its ECM (Pissarenko and Meyers, 2020). For TERM to be effective, engineered constructions must move beyond simple biocompatibility and towards true biomimicry; towards the replication of the skin’s multi-layered hierarchy while still maintaining the precise biochemical signalling mediated by its protein networks. Ultimately, impactful regenerative outcomes are essentially a balancing act of successfully managing the interdisciplinary synergy between AI-driven design and the inherent thermodynamics of the skin’s cellular environment (Chaudhry and Czekanski, 2023; Unagolla and Jayasuriya, 2020; Talebi et al., 2025). TERM is pushing the boundary to not only replicate a cellular surface but also engineer an environment conducive to functionality.

4.1 The structure of skin

4.1.1 The skin microenvironment and peptide-mediated recognition

A biomimetic approach requires the precise spatial arrangement of specific cell types to restore metabolic and protective functions, largely governed by Protein-Protein Interactions (PPIs) at the cell-scaffold interface (Shaw et al., 2025). Hydrogel-based materials have been proposed and worked with to elaborate hydrogel printing technologies for 3D functional tissue by exchanging their properties, their printing conditions and ultimately scaffold design (Askari et al., 2021). Despite extensive research on hydrogels, the complexity of skin, due to its interconnected layering and PPI-mediated interactions, remains a challenge to replicating skin physiology and enhancing 3D skin bioprinting (Table 2).

TABLE 2. Summary of skin layers with biological complexity and current understanding of the bioprinting difficulty for each layer.

Primarily, the epidermis is an avascular layer and dominated by keratinocyte cells (approximately 90%) arranged in a stratified sheet formation (Figure 3). This cellular homogeneity and its perceived geometric simplicity should arguably provide the best starting point for bioprinting research, offering ease in bioink formulation and, given its avascular nature, removing the need for complex capillary network engineering. However, despite the presumed simplicity, its stability relies on the basement membrane, a specialised ECM rich in laminin and collagen IV. The ECM provides suitable environment for PPIs which in turn brings about an increase in complexity for biomimicry; since too firm or too soft a hydrogel will affect the PPIs that govern adhesion to the following layer, by prevention or disorganization of signalling respectively (Chaudhry and Czekanski, 2023; Priya et al., 2024; Luo et al., 2023; Di Stefano et al., 2025; Martin et al., 2026). Moreover, successful epidermal regeneration depends on the bioavailability of ligand motifs that facilitate rapid dermo-epidermal adhesion or the effective recognition between epidermis and dermis (Askari et al., 2021). It would not be enough for the motifs that are involved in recognition to be appended within the hydrogels as their availability for binding is crucial. AI-driven precision modelling can thus be deployed to support and predict the flexibility of the hydrogel and the arrangement of motifs (Priya et al., 2024; Zhang et al., 2026; Li et al., 2024).

Known as the ‘mechanical engine’ of the skin, the dermis is populated by fibroblasts embedded in a dense ECM, thereby raising the bar in terms of complexity for TERM. It is characterised by a gradient of structural proteins that govern tissue stiffness, thereby limiting flexibility in bioink formulation. Within this environment, fibroblasts are recruited and directed through specific PPIs. Maintaining these interactions is a significant challenge in TERM, as the local environment must remain compliant during recognition events, despite the metabolic stressors of injury (Zhang et al., 2026; Li et al., 2024).

Lastly, the hypodermis is the vital reservoir for adipose-derived stem cells (ASCs), not simply a fat storage layer but an important factory for cells uniquely in a blank-slate state (Figure 3). ASCs are positioned to become either adipocytes needed for skin volume, osteoblasts used in bone repair, or, lastly, used as endothelial cells for blood vessels¹. ASCs are important not only for their differentiation potential but also for their ability to promote growth factors such as VEGF via paracrine secretion. These signalling pathways for secretion that function through alpha-helical or beta-sheet mediated PPIs, are essential for the neo-vascularisation of thick hydrogel constructs (Zhang et al., 2023; Rahimnejad et al., 2017; Sawadkar et al., 2025). Unsurprisingly, to leverage the VEGF-mediation in TERM, two of the most vital building blocks used in modern bioink formulations are chitosan and peptides. The former for its physical and structural adjusting and bulk, while the latter to encode for biologically reactive components that can form alpha-helices or beta-sheets and thus encourage PPIs (Shaw et al., 2025; Lazaridou et al., 2025). The manipulation of peptides bridges the boundary mentioned above, not only to replicate a cellular surface but also to engineer an environment conducive to functionality.

4.1.2 Collagen: the structural and functional scaffold

Collagen serves as the fundamental structural protein of the skin, accounting for 70%–80% of its dry weight. Native skin is primarily composed of Type I and Type III collagen, which, through their organisation of triple-helical fibres, give skin its strength or mechanical resilience along defined natural tension lines referred to as Langer’s Lines (Pissarenko and Meyers, 2020). Advances in AI-driven path planning now allow for the alignment of hydrogel microstructures to match these specific tension lines, thereby enhancing overall graft integration (Sawadkar et al., 2025; Alberts et al., 2025). Synthetic hydrogels replicate this tension functionality by controlling the ratio of Type I to Type III collagen (Kanniyappan et al., 2025).

Beyond structural support, collagen contains specific amino acid sequences, such as the Arg-Gly-Asp (RGD) motif, that act as ligands for cell integrins (Figure 4). A critical challenge in utilising collagen-derived bio-inks, such as Gelatin Methacryloyl (GelMa), is ensuring that these bioactive motifs remain accessible and resistant to premature proteolysis during remodelling. If the peptide motifs lack conformational constraint, they quickly become tagged for protease degradation, losing their signalling efficacy (Martin et al., 2026). In 2024 Zhu et. al. designed polysaccharide-based inks, where hydrogels were decorated with RDG peptides, making these available for recognition events. This success showed the ability to stick, preventing the graft from sliding and being washed away. However, as is often the case, this example replicated crosslinking in the hydrogel and was too strong to allow mobility and rearrangement in response to environmental needs in the skin.

4.2 Need for TE in skin regeneration

While the skin possesses innate repair mechanisms, this natural regeneration window is not infinite, leading to or from a breakdown in native PPI signalling and the subsequent formation of non-functional scar tissue (Talebi et al., 2025; Zhang et al., 2023; Sawadkar et al., 2025; Lazaridou et al., 2025; Boyce and Lalley, 2018; Froelich et al., 2023). The depth of the defect can disrupt PPIs as well, once a defect exceeds the time needed for the PPIs in the skin to restore itself to its original tissue architecture, scarring occurs. This fragility in depth and window of repair requires the development of a scaffold preloaded with modular peptide-based units to serve as the missing links in PPI pathways. Organisation needs to be reinstated via PPIs, which creates a bottleneck in current biomimetic scaffolds.

Of course, one of the oldest clinical practices for skin repair has been skin grafting with roots stretching to ancient India, mentioned in Sushruta Samhita, an ancient Sanskrit text (Ameer et al., 2013; Alsaif et al., 2023). The western world experienced an expansion of this technique in the 1800 and grafts have been the gold standard in wound-closing efficacy with added biological safety. However traditional grafts act as patches rather than true replacements which leads to two main issues: a. Donor site morbidity: the act of harvesting skin creates secondary trauma at the donor site, introducing new vectors for infection and chronic pain (Ameer et al., 2013; Alsaif et al., 2023; Yeh and Lin, 2025); and b. Quantity and appendage loss: traditional grafts fail to regenerate complex appendages like sweat glands and hair follicles, which rely on intricate, developmentally timed PPI-driven pathways; a common feature they share with TERM. This leads to permanent loss of thermoregulation and sensation (Luo et al., 2023).

Chronic wounds, such as diabetic foot ulcers (DFUs), are characterised by a stalled normal healing pathway due to an imbalance in the protein interactions that govern inflammation. As of 2026, the US healthcare system spends over $25 billion annually on chronic wound care (Roots Analysis, 2026). AI-optimised 3D hydrogels can be designed to deliver localised peptide-based modulators that ‘re-set’ these disrupted PPI networks, effectively jumpstarting the healing process (Rahimnejad et al., 2017).

Although the complexities of PPI allowance in hydrogel formulation proves to be challenging, a three-dimensional, gelatin methacryloyl hydrogel was possible in 2025 (Di Stefano et al., 2025). The research group neatly designed a structural model with a degree of substitution and containing stratified epidermal cells, fibroblasts (dermis), and endothelial cells (specifically human umbilical vein endothelial cells, HUVECs). The interlayer bonding provided the closest native design of skin structure. Despite the bulk composition and layering fusion, there was a lack of signalling pathway or functionality for full PPI.

More recently, with the proliferation of AI models, a shift towards leveraging AI in analysis and design to synergistically integrate 3D skin models is underway (Li et al., 2026). Arranging cells in 3D space with precision has been demonstrated but remains static. Thus, leveraging AI could lead to a surge in new designs and formulations. Ideally, the intersection of AI and 3D bioprinting should allow for the transition from passive dressings to patient-specific regenerative therapies, the blueprint from static hydrogels to functional completeness. By utilising patient-derived cells and AI-modelled scaffolds, TERM provides a minimally invasive approach to reconstruction. This ensures that the mechanical base of the hydrogel degrades at a rate matched to the patient’s own tissue synthesis, moving the field towards the goal of functional, scarless regeneration (Martin et al., 2026).

4.3 3D printing and materials for skin regeneration

As discussed above, native skin has a complex, multilayered architecture with diverse cells and extracellular components, making it difficult to replicate using conventional TE approaches. Limitations such as poor spatial control, uneven cell distribution, and inadequate vascularisation persist. 3D bioprinting enables precise deposition of cells and biomaterials, improving the fabrication of biomimetic skin constructs (Haykal et al., 2026).

The materials used in these approaches for skin regeneration are broadly classified into synthetic and natural polymers. Common synthetic options include polycaprolactone (PCL) and poly lactic-co-glycolic acid (PLGA), while natural polymers such as chitosan, collagen, and fibrinogen are also widely used (Lee et al., 2014). Molecular weight also plays an important role in the selection of the materials. LMW polysaccharides have shorter chains which results in lower viscosity and is ideal for when the aim is to fabricate rapidly degrading hydrogels in skin. These materials are also known to stimulate the growth of new blood vessels. HMW polysaccharides have longer chains with stiffer characteristics, providing a stronger mechanical and structural support for hydrogels (Ho et al., 2021). Cell selection is another key factor, with most research focusing on keratinocytes due to their ease of expansion, along with fibroblasts, which possess multilineage differentiation potential. However, these cells alone have not yet produced completely functional skin substitutes, and recreating structures like sebaceous glands and hair follicles in vitro remains a challenge (Lee et al., 2014; Tarassoli et al., 2018).

Hydrogels play a central role in bioprinting as bioinks due to their ability to mimic the native ECM and provide a hydrated microenvironment, conducive to cell survival and organisation. Natural hydrogels, including collagen, gelatin, alginate, hyaluronic acid, fibrin, and chitosan, are widely used owing to their biocompatibility and bioactivity (Lima et al., 2022; Fang et al., 2023b). These materials enhance cell adhesion, proliferation, and differentiation, closely resembling the native cellular environment. Figure 5 represents the schematic overview of 3D printing for skin TERM applications. The schematic illustrates the integration of AI-based tools with advanced 3D bioprinting systems for skin tissue engineering and regenerative medicine. AI approaches, including machine learning, bioink optimization, and print parameter prediction, support design optimization and real-time process monitoring to improve fabrication precision and reproducibility. Various classes of biomaterials, including natural biomaterials, synthetic polymers, composite hydrogels, decellularized extracellular matrix (dECM), and bioactive components such as cells and growth factors, are incorporated into bioinks for bioprinting applications. These biomaterial-based constructs enable the generation of multilayered skin substitutes comprising epidermal, dermal, and vascular regions, thereby enhancing tissue regeneration and functional skin reconstruction.

Recent advances have focused on composite biomaterials that combine natural and synthetic components to achieve an optimal balance between bioactivity and mechanical performance. For instance, GelMa–PEG and collagen–PCL systems have demonstrated improved printability, stability, and cell-supportive properties, enabling the fabrication of full-thickness skin equivalents (Weng et al., 2021). Additionally, decellularised ECM (dECM)-based bioinks have gained attention for their ability to retain native biochemical cues, including growth factors and structural proteins, thereby promoting tissue-specific regeneration, and enhanced cellular responses (Ruberu et al., 2021).

Nanocomposite hydrogels further expand the functional capabilities of bioinks by incorporating nanoparticles such as silver, gold, or graphene oxide. These additives impart antimicrobial properties, improve mechanical strength, and enhance electrical conductivity, which can be beneficial for wound healing and tissue maturation (Lima et al., 2022). Moreover, the integration of bioactive components, including keratinocytes, fibroblasts, endothelial cells, and growth factors (e.g., VEGF, EGF, FGF), enables the reconstruction of complex, multilayered skin tissues with vascularisation potential (Ng et al., 2019; Fang et al., 2023b). The selection of biomaterials and additives directly influences the type and behaviour of incorporated cells, with keratinocytes, fibroblasts, endothelial cells, and stem cells being the most used to recapitulate the multilayered structure and function of native skin. Table 3 not only summarises existing studies but also distinguishes between true bioprinting (cell-laden) and scaffold-based approaches.

TABLE 3. Hydrogel-based 3D bioprinting strategies for skin tissue engineering: a comparative overview of bioink composition, incorporated cell types, printing techniques, and biological loading approaches across recent studies.

4.4 3D printing and AI in skin TERM

The integration of AI, specifically its subset of ML, into 3D bioprinting is reshaping skin TE by enabling data-driven optimisation, predictive modelling, and automation of complex fabrication processes (Zhang et al., 2026; and Greener et al., 2022). Conventional bioprinting relies heavily on empirical optimisation of bioink formulations and printing parameters, a process that is time-consuming and often lacks reproducibility. AI approaches, including machine learning (ML), deep learning, and computer vision, allow the analysis of multidimensional datasets to improve accuracy, scalability, and consistency in bioprinted constructs (Rezapour Sarabi et al., 2022; Song et al., 2023).

A key application of AI lies in the design and optimisation of hydrogel-based bioinks. These materials exhibit complex rheological and crosslinking behaviours that directly influence printability and cell viability (Naghieh and Chen, 2021). ML models have been successfully used to predict print fidelity and structural outcomes based on bioink composition and processing conditions, reducing experimental iterations (Rezapour Sarabi et al., 2022; Choi et al., 2023). Recent studies on AI-informed hydrogels demonstrate improved control over mechanical properties, degradation kinetics, and biological performance, enabling the development of next-generation smart biomaterials for skin regeneration (Zhang et al., 2025a; Zhang et al., 2025b).

AI also plays a pivotal role in optimising bioprinting process parameters. Variables such as extrusion pressure, nozzle speed, and temperature significantly affect construct integrity and cell survival, and thus the final application at the target site in the skin. ML-enabled predictive frameworks, including those applied to microneedle fabrication, have demonstrated strong correlations between process parameters and structural features, allowing efficient parameter tuning (Rezapour Sarabi et al., 2022; Biswas et al., 2024). Furthermore, closed-loop AI systems integrated with real-time imaging and sensor feedback enable adaptive control of the printing process, minimising defects and enhancing reproducibility (Ruberu et al., 2021).

Beyond process optimisation, AI facilitates the design of complex tissue architectures. AI is emerging as a transformative tool in hydrogel research, driven by its advanced capabilities in data processing, algorithmic modelling, and pattern recognition. It enables high-throughput screening of hydrogel systems, facilitates prediction of optimal compositions, and supports the optimisation of material properties alongside a robust model development (Wang et al., 2023; Zhang et al., 2025a). For example, digital light processing-based 3D printing combined with advanced modelling has enabled the fabrication of full-thickness skin equivalents with improved structural fidelity (Choi et al., 2023). Additionally, AI-integrated electronic skin platforms have demonstrated the ability to combine biosensing, data acquisition, and machine learning for real-time physiological monitoring, highlighting the convergence of tissue engineering and wearable bioelectronics (Song et al., 2023; Xu et al., 2023).

AI is also advancing the development of intelligent and responsive biomaterials. AI-enabled hydrogels can be engineered to respond to environmental stimuli, allowing dynamic modulation of mechanical properties, drug release, and cellular interactions. Such systems are particularly promising for wound healing applications, where adaptive responses to biological cues are critical (Zhang et al., 2025a). Moreover, AI-driven modelling approaches have been used to predict hydrogel behaviour under different printing conditions, further enhancing design efficiency and functional outcomes (Bediaga-Bañeres et al., 2025). Another critical application of AI-ML is in quality control and post-printing evaluation. Computer vision and deep learning techniques enable automated assessment of print fidelity, cell distribution, and structural integrity. These tools provide rapid, non-invasive evaluation of constructs and are essential for scaling bioprinting technologies toward clinical translation (Sujeeun et al., 2020; Tourlomousis et al., 2019).

Despite these advances, challenges remain, including limited availability of standardised datasets, variability in experimental protocols, and regulatory uncertainties (Biswas et al., 2024; Goh et al., 2021). Addressing these limitations will require interdisciplinary collaboration and the development of robust data-sharing frameworks. Nevertheless, the convergence of AI with 3D bioprinting represents a paradigm shift toward intelligent, adaptive, and clinically translatable skin regeneration systems, with significant implications for personalised medicine and advanced therapeutic applications (Cai et al., 2023; Thomazi et al., 2024).

5 Challenges, conclusions and future directions

While AI-integrated regenerative medicine has the potential to have a powerful impact on biomimetics, the initial stages will face several challenges in replicating the complex physiological structures of bones and skin. For bone TE, fabricating a scaffold that mimics bone’s external hardness, internal architectural intricacies and complexity of the BM niche is difficult. Various approaches involving organoids, spheroids, and organ-on-chips have been attempted, but they rarely tick all these boxes. For skin, variability due to limited standardised datasets, difficulty in replicating epidermal–dermal biology, and intricacies in keratinocyte–fibroblast interactions can restrict model reliability (Rezapour Sarabi et al., 2022; Gharibshahian et al., 2024). In 3D bioprinting, personalisation and scale-up are hindered by complex design spaces and poor reproducibility (Zhang et al., 2025a), challenges faced by both skin and bones. Technical limitations in sensing and automation reduce fidelity, while vascularisation, immune compatibility, and regulatory challenges delay clinical translation. Moreover, currently investigations are still focussed on optimisation of materials and/or techniques to obtain measurable data; with limited studies being able to suggest the exact mechanistic pathways involved in their results.

Other challenges around the fabrication and translation of these hydrogels include bioavailability, dose standardisation, fermentation variability and toxicity. Bioavailability remains a major challenge, as encapsulated cells and bioactive molecules (e.g., growth factors) often exhibit reduced stability and inconsistent release profiles within hydrogel matrices, limiting therapeutic efficacy (Kapoor et al., 2025). Dose standardization is difficult due to variability in hydrogel composition, crosslinking density and printing parameters, which directly influence drug/growth factor release kinetics and cell distribution; preventing reproducible clinical outcomes. Fermentation variability, particularly for polysaccharide-derived bioinks (e.g., alginate, hyaluronic acid), introduces batch-to-batch inconsistencies in molecular weight, purity and rheological behaviour, affecting printability, degradation and biological performance (Parvin et al., 2025). This variability complicates regulatory approval and large-scale manufacturing consistency. Toxicity and safety concerns also hinder translation. Residual crosslinking agents, degradation by-products and impurities can induce cytotoxicity or adverse immune responses, despite the general biocompatibility of hydrogels. Additionally, inadequate vascularization and nutrient diffusion in printed constructs may result in cell death post-implantation, further limiting clinical success (Omidian et al., 2024).

Despite these challenges, research and innovation in TERM for bone and skin are moving towards their peak, with a ‘from bench to bedside’ vision to enhance patient care. The convergence of TERM and AI is transforming regenerative medicine and bioprinting through predictive modelling, biomaterial optimisation, and process automation (Rezapour Sarabi et al., 2022; Gharibshahian et al., 2024). The involvement of AI enhances design efficiency, reduces experimental burden, and improves construct quality (Zhang et al., 2025a). Clinical translation for viable skin substitutes remains constrained by scalability, biological complexity, and regulatory barriers (Hoang et al., 2025), as it does for bone. The use of patient-derived cells in scaffolds has shown encouraging results in cancer (Jeong et al., 2021; Schroyer et al., 2025), and with the help of AI, several stages of tissue engineering can be streamlined for enhanced patient outcomes. Studies described in Sections 3.4, 4.4 that used ML have a better chance to predict biocompatibility, toxicity, mechanical strength and cellular behaviours (Fu et al., 2025). This will play a critical role in optimising TE for both skin and bones for precision-based approach for the patient. Use of AI and ML also have the potential to analyse the parameters outlined above faster and reduce the need for animal models (Jeznach et al., 2024).

For fabrication of large volume of hydrogels/scaffolds, AI can significantly enhance advanced path planning in multi-nozzle 3D bioprinting for large-volume tissue engineering scaffolds by enabling data-driven optimization and coordination. ML algorithms can process geometric models, material properties and printing constraints to generate optimized toolpaths that improve deposition accuracy and structural fidelity. In multi-nozzle systems, AI can facilitate dynamic task allocation, synchronizing nozzles to reduce collisions and enable parallel fabrication, thereby increasing scalability. AI-driven models may also be able to predict filament behaviour, flow dynamics alongside scaffold porosity, allowing precise control over microarchitecture; that is essential for bone and skin regeneration. Furthermore, integration with real-time sensors enables adaptive path correction and error minimization during printing. AI may transform traditional toolpath planning into an intelligent, adaptive process, improving efficiency, reproducibility and fabrication of complex hydrogels for tissue engineering.

Thus, future efforts should prioritise standardised datasets and multi-modal data integration of bone and skin microenvironments to improve AI robustness (Rezapour Sarabi et al., 2022; Gharibshahian et al., 2024). Training modalities to scan through the large variety of datasets for TE in both skin and bone will play a crucial role in identifying strengths and weaknesses for robust TE models. AI-driven Quality by Design (QbD) frameworks and digital twins will enhance reproducibility and scalability (Zhang et al., 2025a). Autonomous bioprinting with real-time monitoring is anticipated, while addressing vascularisation, immune integration, and regulatory pathways will be essential for clinical translation. Figure 6 below shows a schematic overview of futuristic trends in 3D bioprinting and AI in skin TERM. The schematic illustrates the integration of AI-driven tools, including machine learning, bioink optimization, and predictive modeling, with advanced 3D bioprinting platforms for skin regeneration. Future developments will focus on smart hydrogel bioinks, adaptive bioprinting systems, and digital-clinical integration to enable automated fabrication, real-time process monitoring, patient-specific constructs, vascularized tissues, and personalized regenerative therapies.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. Generative AI was used to support the formation of figures only in the manuscirpt.

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.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fbiom.2026.1886858/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Lorenzo Vannozzi, Sant'Anna School of Advanced Studies, Italy

Reviewed by: Sivakumar Singaravelu, Rajalakshmi Institute of Technology (RIT), India

Christopher S. O'Bryan, University of Missouri, United States