AI and silkworm biomanufacturing: a new paradigm for malaria therapeutics access

1 Introduction

Malaria, caused by protozoan parasites of the genus Plasmodium, continues to impose an enormous burden on global public health. According to the World Health Organization’s World Malaria Report 2025, there were an estimated 282 million cases and 610,000 deaths in 2024 (World Health Organization, 2025). The WHO African Region bears a disproportionate share of this burden, accounting for 94% of global cases and 95% of fatalities, with children under 5 years of age constituting the most vulnerable population (Medicines for Malaria Venture, 2025). The development and deployment of highly efficacious malaria vaccines represents the most promising pathway toward achieving global elimination targets (World Health Organization, 2015). The complexity of the Plasmodium falciparum life cycle, encompassing the pre-erythrocytic (liver) stage, the blood stage responsible for clinical disease, and the sexual (mosquito) transmission stage, requires multi-stage immune responses to block infection, prevent severe disease, and halt onward transmission (Longley et al., 2015).

To provide a clear understanding of these intervention points, Figure 1 illustrates the P. falciparum life cycle and highlights the specific stages targeted by current vaccines, monoclonal antibodies, and antimalarial drugs.

The recent WHO recommendation of the RTS,S/AS01 and R21/Matrix-M vaccines, which target the pre-erythrocytic circumsporozoite protein (CSP), marks a critical breakthrough (Gavi, the Vaccine Alliance, 2024). However, these vaccines exhibit partial, waning efficacy and lack sterilizing immunity, underscoring the urgent need for next-generation vaccines capable of overcoming the extensive antigenic diversity of Plasmodium species (Takala et al., 2009).

Equally problematic are the structural constraints of current vaccine manufacturing architectures. Current biological production systems rely on highly specialized mammalian or microbial cell culture facilities, requiring massive capital expenditure, from $50 million to $500 million per antigen, and complex scale-up dynamics that can span 9 to 12 months from gene sequence to Good Manufacturing Practice (GMP) batch release (Plotkin et al., 2017; Kumraj et al., 2022). For LMICs, this results in a geography of dependence, where countries rely entirely on imports from a small number of centralized manufacturers. This centralization leaves endemic regions vulnerable to supply disruptions and a fundamental lack of responsiveness to local epidemiological shifts as depicted during the COVID-19 period (Peacocke et al., 2021; Kumraj et al., 2022).

To overcome these scientific and manufacturing limitations simultaneously, a dual approach is required. First, artificial intelligence (AI) and machine learning (ML) are transforming malaria research, enabling the rapid identification and design of novel, broadly protective antigens and potent monoclonal antibodies. Recent advancements demonstrate that AI models can rapidly screen vast chemical libraries to identify novel antimalarial compounds and predict the stability of complex immunogens (Khanday et al., 2025; Nishan, 2025). Second, novel biomanufacturing platforms are needed to address the production bottleneck. Whole-insect production systems, specifically the Bombyx mori (silkworm) baculovirus expression vector system (BEVS), offer the prospect of rapid scale-out, enabling decentralized production of both vaccines and therapeutic antibodies with significantly lower upstream infrastructure requirements (Kato et al., 2010). By utilizing the living silkworm as a natural bioreactor, this platform bypasses the need for large-scale stainless-steel fermenters. This review synthesizes the current state of AI-driven malaria therapeutic design and demonstrates how the silkworm platform uniquely supports the Partnerships for African Vaccine Manufacturing (PAVM) in delivering equitable access to next-generation malaria therapeutics.

2 AI-driven innovations in vaccine research: from global pathogens to malaria

The transformative impact of AI and ML in vaccinology has been unequivocally demonstrated across a spectrum of infectious diseases and oncology, driving breakthroughs in structural stabilization, immunogen design, and personalized medicine. These successes in other disease areas provide a critical proof-of-concept for the computational methodologies now being applied to malaria.

In the development of vaccines for respiratory syncytial virus (RSV), computational modeling and ML were critical in solving a decades-old structural biology challenge. The RSV fusion (F) protein is highly unstable in its prefusion conformation, which is the primary target for potent neutralizing antibodies. Through iterative computational design, researchers engineered specific amino acid substitutions and disulfide bonds to lock the F protein in its prefusion state, leading to the highly efficacious RSV vaccines recently approved for older adults and maternal immunization (McLellan et al., 2013; Falloon et al., 2017). More recently, AI screening models have been established to predict the stability and immunogenicity of mRNA-encoded RSV prefusion F proteins, further accelerating the optimization pipeline (Swanson et al., 2020).

Similarly, in HIV research, machine learning frameworks are being deployed to decipher the complex signatures of broadly neutralizing antibodies (bNAbs). Tools such as the RAIN (machine learning-based identification for HIV-1 bNAbs) framework analyze vast libraries of antibody sequences to predict neutralization breadth, informing the rational design of sequential immunogens intended to guide the immune system step-by-step toward bNAb development (Sok and Burton, 2018; Foglierini et al., 2024). In the realm of oncology, AI is revolutionizing personalized cancer vaccines. Deep learning algorithms are now routinely used to sequence a patient’s tumor genome, identify neoantigens unique to the cancer cells, and predict which of these neoantigens are most likely to be presented by the patient’s HLA molecules and recognized by T cells, enabling the rapid synthesis of personalized mRNA cancer vaccines for clinical trials (Yale School of Medicine, 2026). Furthermore, platforms utilizing AI are expanding into universal vaccine design for pathogens like dengue, demonstrating the broad applicability of these computational approaches to global health challenges (ImmunoPrecise Antibodies Ltd., 2025).

2.1 Applying AI methodologies to the malaria challenge

Building upon these cross-disciplinary successes, AI is now being leveraged to overcome the historical roadblocks specific to malaria vaccine development. Traditional malaria vaccine design focused on immunodominant antigens, which inadvertently favored highly polymorphic proteins that evolved under intense immune pressure, thereby limiting cross-strain efficacy. Reverse vaccinology inverts this paradigm by computationally screening entire pathogen genomes to identify conserved, surface-exposed antigens that may have been overlooked by classical approaches (Rappuoli, 2000). The application of ML has substantially enhanced this strategy. A landmark 2024 study implemented a positive-unlabeled (PU) learning framework analyzing 272 variables across the P. falciparum proteome. By learning the properties of known protective antigens, the model predicted novel candidates, prioritizing those with low genetic diversity and high expression across multiple life stages, enabling the identification of conserved antigens that may elicit strain-transcending immunity (Chou et al., 2024).

AI is also fundamentally transforming the structural design of malaria immunogens, mirroring the successes seen with RSV. Deep learning tools like AlphaFold2 have elucidated the complex structures of critical malaria targets, such as the transmission-blocking gamete surface protein Pfs48/45, revealing its three domains in a dynamic, planar, triangular arrangement and guiding the rational design of stabilized recombinant fragments (Ko et al., 2022). Building on these structural predictions, generative AI now enables the de novo design of multi-epitope vaccine scaffolds. A 2025 study published in PNAS employed a machine learning-driven, structure-guided pipeline to engineer novel antigens presenting multiple copies of key PfCSP epitopes in an optimally oriented configuration, designed to elicit potent neutralizing antibodies against the parasite (Nelson et al., 2025). The same study demonstrated that these computationally designed scaffolds could be expressed in recombinant systems and elicited antibody responses superior to those induced by the native antigen in preclinical models, representing a major advance in rational malaria vaccine design.

A critical and historically intractable challenge in malaria vaccinology is the definition of immune correlates of protection. ML algorithms are uniquely suited to analyze high-dimensional immunological datasets from controlled human malaria infection (CHMI) trials. For instance, Wistuba-Hamprecht and colleagues applied a multitask Support Vector Machine (SVM) approach to analyze P. falciparum-specific antibody profiles from CHMI studies, significantly outperforming standard approaches in predicting sterile protection and identifying specific antigens whose antibody responses were most predictive of efficacy (Wistuba-Hamprecht et al., 2024). This provides a data-driven roadmap for the design of future multi-component vaccine formulations. AI is further revolutionizing adjuvant discovery, with ML models predicting the efficacy of novel adjuvants and optimizing lipid nanoparticle (LNP) formulations to enhance vaccine stability and immunogenicity (Zhang et al., 2024; Ji et al., 2025).

2.2 AI in antimalarial drug discovery and adjuvant design

Beyond vaccine antigens, AI is profoundly reshaping pharmacological development and adjuvant design. In antimalarial drug discovery, the emergence of artemisinin-resistant P. falciparum strains necessitates the rapid identification of novel therapeutic compounds. Machine learning and deep learning algorithms, such as convolutional neural networks (CNNs), are now employed to rapidly screen massive chemical libraries, identifying compounds with antimalarial properties far faster than traditional high-throughput screening (Nishan, 2025). For example, AI-driven platforms like DeepMalaria predict the inhibitory properties of thousands of compounds against the parasite, prioritizing those most likely to bypass existing resistance mechanisms (Keshavarzi Arshadi et al., 2020).

Furthermore, the application of AI extends to the design of vaccine adjuvants. A recent perspective highlights the potential for AI-driven design of potent antigens to progressively replace or augment classical chemical adjuvants (Marchan-Alvarez, 2026). By embedding adjuvant-like functions, such as specific pattern recognition receptor (PRR) binding motifs, directly into computationally designed antigens, AI could enable the creation of self-adjuvanting vaccines. This approach promises to reduce the reactogenicity and manufacturing complexities associated with traditional adjuvants like alum or complex emulsions, streamlining the formulation process for next-generation malaria vaccines (Marchan-Alvarez, 2026).

2.3 AI-driven innovations in malaria diagnostics

Beyond vaccine and therapeutic design, AI is transforming a third, equally critical pillar of the malaria response: diagnostics. Accurate and timely diagnosis is the cornerstone of effective case management, rational drug deployment, and epidemiological surveillance, yet it remains a profound bottleneck in malaria-endemic settings. The gold standard for malaria diagnosis, expert microscopic examination of Giemsa-stained thick and thin blood smears, requires a high level of technical skill and sustained concentration, making it vulnerable to error under the high-throughput conditions typical of overburdened health facilities in sub-Saharan Africa (Faratisha et al., 2026). Rapid Diagnostic Tests (RDTs) based on the detection of P. falciparum histidine-rich protein 2 (HRP2) have partially addressed this gap, but their reliability is increasingly compromised by the emergence and spread of pfhrp2/pfhrp3 gene deletions, which cause false-negative results and have now been documented in 42 countries, posing a direct threat to malaria control strategies (Feleke et al., 2021).

Deep learning and CNNs are now being deployed to automate and standardize blood smear analysis, removing the dependency on scarce expert microscopists. A 2026 systematic review and meta-analysis demonstrated that AI-based diagnostic systems achieved a pooled sensitivity of 87.7% and a specificity of 91.4% relative to expert microscopy, performance metrics that are comparable to trained human microscopists (Faratisha et al., 2026). Purpose-built AI-integrated platforms, such as the miLab MAL automated instrument, have demonstrated 100% sensitivity and 100% specificity for P. falciparum detection in prospective clinical evaluations, while reducing turnaround time to approximately 18 minutes per sample (Nagendra et al., 2025). Similarly, the iMAGING system, a low-cost robotized microscope integrated with a smartphone application and YOLOv5-based CNN, achieves over 92% precision and 93% recall in detecting Plasmodium trophozoites, and was specifically designed for deployment in resource-limited, malaria-endemic settings (Maturana et al., 2023).

Beyond microscopy automation, AI is being applied to the interpretation of RDT results through Connected Diagnostics (ConnDx) platforms. A prospective cross-sectional study in Kenya demonstrated that an AI-driven smartphone application could interpret mRDT images with 96.1% sensitivity and 98.0% specificity, performance statistically indistinguishable from that of trained healthcare workers (Bahati et al., 2025). Critically, this system simultaneously uploads results to a cloud-based dashboard, enabling real-time malaria surveillance and outbreak detection at the population level.

3 Overcoming the manufacturing bottleneck: classical platforms versus silkworm-based biomanufacturing

3.1 The limitations of classical platforms

While AI accelerates the in silico design of malaria antigens and antibodies, translating these digital sequences into physical therapeutics remains critically constrained by the structural limitations of current biomanufacturing paradigms. The COVID-19 pandemic starkly exposed the fragility of centralized, capital-intensive biomanufacturing systems, demonstrating that discovery speed alone is insufficient without a parallel revolution in manufacturing. Addressing this vulnerability, particularly for malaria-endemic regions, demands a radical shift from classical platforms to highly scalable, decentralizable production systems.

Current biologics manufacturing is predominantly reliant on two complementary but imperfect paradigms. The first is mammalian cell culture systems (e.g., CHO or HEK293 cells), which form the backbone of traditional recombinant protein vaccine and monoclonal antibody production. While highly reliable for routine manufacturing and capable of complex post-translational modifications, these systems require massive capital investment. Establishing a conventional mammalian cell facility typically demands over €50 million in capital expenditure (CapEx) for specialized stainless-steel bioreactors, sterile fermentation environments, and complex cell-culture media (Plotkin et al., 2017). Furthermore, scaling up these volumetric processes is slow, typically requiring 9 to 12 months to move from gene sequence to GMP batch release (Kumraj et al., 2022).

The second paradigm is mRNA platforms. While mRNA technology demonstrated extraordinary development speed during the COVID-19 pandemic, it faces significant deployment challenges in LMICs. mRNA vaccines require stringent ultra-cold storage or costly lyophilization to ensure thermostability (Uddin and Roni, 2021). Furthermore, the manufacturing of complex lipid nanoparticles (LNPs) is cost-intensive and technically demanding, requiring sterile GMP-compliant infrastructure that severely limits the number of global manufacturing sites (BusinessWire, 2024). Consequently, neither mammalian cell culture nor mRNA platforms are optimally suited for the rapid, decentralized, and low-cost production required for global malaria equity.

3.2 The silkworm platform: rapid, scalable, and decentralizable

The Bombyx mori (silkworm) baculovirus expression vector system (BEVS) represents a structural disruption of the biologics manufacturing value chain. By utilizing the living insect as a programmable natural bioreactor, the silkworm platform overcomes the fundamental limitations of classical systems across four key dimensions: speed, cost, scalability, and sustainability. Table 1 summarizes the comparative advantages of the silkworm platform relative to mammalian cell culture, mRNA systems, and conventional insect cell culture.

Table 1. Comparative analysis of current biomanufacturing platforms versus the whole-insect silkworm BEVS platform.

The silkworm platform enables rapid translation from digital sequence to biological product. Because it bypasses the months-long process of establishing and expanding stable mammalian cell lines, the silkworm system can theoretically deliver an AI-designed candidate ready for GMP batch release within 100 days of sequence availability, directly aligning with the CEPI 100 Days Mission for pandemic preparedness (Hernandez-Ruiz et al., 2025; Brühlmann Consulting, 2026).

Silkworm rearing does not require capital-intensive stainless-steel bioreactors or sterile fermentation environments. The upstream production process is radically simplified. Even when rearing silkworms under Specific Pathogen Free (SPF) conditions, which adds a modest biosafety overhead, the infrastructure remains orders of magnitude less capital-intensive than cell-based facilities. A silkworm-based production unit is estimated to cost between €2 and €5 million, representing a ten-fold CapEx reduction compared to classical platforms (Felberbaum, 2015; Brühlmann Consulting, 2026). Consequently, the estimated production cost per dose is expected to drop from approximately €3–5 for mammalian systems to less than €1 for silkworm-derived antigens (Bunnak et al., 2016; Amasawa et al., 2021).

Because each insect can be considered an individual production unit, the silkworm-based platform achieves scale-out rather than scale-up. A single silkworm pupa yields, on average, the equivalent recombinant protein amount produced by approximately 100 to 1,000 mL of conventional insect cell culture, with some proteins reaching much higher equivalencies (Park, 2025; Brühlmann Consulting, 2026). This scale-out model eliminates the engineering risk associated with building larger bioreactor tanks and reduces infrastructure burden, supporting long-term cost efficiency. This low barrier to entry democratizes access to biologics manufacturing, opening the possibility of truly distributed, decentralized biomanufacturing across resource-constrained malaria-endemic regions.

3.3 Technical and regulatory considerations for silkworm-based biomanufacturing

While the advantages of whole-insect biomanufacturing are profound, several technical and regulatory hurdles must be addressed to enable widespread adoption for human therapeutics.

The regulatory pathway for baculovirus-expressed proteins is already well-established. Major regulatory agencies (EMA, FDA) have approved several human vaccines produced in insect cell BEVS, including GSK’s Cervarix (HPV), Sanofi’s FluBlok (influenza), and Novavax’s Nuvaxovid (COVID-19) (Hong et al., 2023). This provides a strong regulatory foundation regarding the safety of baculovirus vectors and insect-derived glycosylation patterns. For whole-insect systems, the regulatory framework is emerging. Companies like Algenex (using Trichoplusia ni pupae) have successfully navigated the EMA to secure approval for veterinary vaccines, proving that whole-insect platforms can meet stringent GMP standards (Stanton, 2019). Similarly, KAICO in Japan is advancing silkworm-derived human norovirus vaccines into Phase I clinical trials, working closely with the Pharmaceuticals and Medical Devices Agency (PMDA) to define raw material controls and CMC expectations for human use (Brühlmann Consulting, 2026). For African regulatory authorities, aligning with these precedents through WHO prequalification pathways will be critical.

A common question regarding BEVS is whether the production of the baculovirus seed stock itself requires large-scale bioreactors. Fortunately, the baculovirus inoculum required to infect silkworms is produced in small-volume insect cell cultures (e.g., Sf9 cells in disposable shake flasks or wave bags) (Kwang et al., 2016). Because the amplification of the protein occurs in vivo within the silkworm, the upstream virus production footprint remains exceptionally small and manageable, avoiding the capital-intensive bioreactors required for mammalian cell culture.

Utilizing a whole organism as a bioreactor introduces unique downstream processing challenges compared to monoculture cell lines. Silkworm lysates or hemolymph are a complex biological matrix containing host proteins, lipids, and potential native insect pathogens. To mitigate this, larvae must be reared under Specific Pathogen Free (SPF) conditions to prevent adventitious viral or bacterial infections (Park, 2025). Downstream purification requires robust clarification steps (e.g., centrifugation and depth filtration) to remove the bulk biomass, followed by standard chromatographic techniques (affinity, ion-exchange, and size-exclusion chromatography) to isolate the recombinant protein and clear residual baculovirus and host cell proteins (Kwang et al., 2016). While the initial extraction is more complex than harvesting secreted proteins from CHO cells, modern filtration and chromatography resins are highly capable of achieving the >99% purity required for human injectables.

4 Expanding the arsenal: AI-designed monoclonal antibodies and silkworm production

The integration of AI and silkworm biomanufacturing extends beyond prophylactic vaccines to encompass passive immunization through monoclonal antibodies (mAbs). This represents a critical new frontier in malaria prevention and treatment, particularly for highly vulnerable populations such as infants and pregnant women, or for rapid outbreak control where the delayed onset of vaccine-induced immunity is insufficient.

4.1 AI-driven discovery of anti-malarial monoclonal antibodies

Historically, the discovery of highly potent mAbs relied on laborious B-cell sorting and screening from convalescent patients or vaccinees. Today, machine learning and structural biology are revolutionizing this process. AI-driven approaches are beginning to accelerate antibody discovery and optimization, enabling the rapid design of high-affinity, broadly protective antibodies against conserved malaria targets. A paradigm shift occurred with the identification of CIS43, a highly potent neutralizing antibody targeting the junctional region of the P. falciparum circumsporozoite protein (PfCSP) (Kisalu et al., 2018).

Recent phase 2 clinical trials in Mali demonstrated that a single intravenous infusion of the engineered variant CIS43LS, which incorporates an Fc-domain mutation (LS: M428L/N434S) for extended half-life, provided up to 88.2% sterile protection against P. falciparum infection over a 6-month malaria season (Wu et al., 2022; Lyke et al., 2023). Similarly, the L9LS monoclonal antibody, which targets a distinct, highly conserved epitope on PfCSP, has demonstrated profound protective efficacy in early clinical trials, with a phase 2 study in Malian children reporting 77% efficacy against clinical malaria (Steinhardt et al., 2026).

AI is now accelerating the discovery of next-generation anti-malarial mAbs. Deep learning models can analyze the structural interactions between antibodies like CIS43 or L9 and their target epitopes, predicting amino acid substitutions that enhance binding affinity, broaden cross-strain reactivity, or improve thermodynamic stability (Nelson et al., 2025). Furthermore, AI-guided platforms are being used to identify entirely novel, non-dominant epitopes across various Plasmodium life stages, enabling the rational design of mAb cocktails that target the parasite simultaneously at the sporozoite, blood, and transmission stages. The combination of AI-accelerated discovery and silkworm-based production creates a uniquely powerful pipeline for rapidly translating computational insights into deployable biological therapeutics.

4.2 Silkworm-based production of monoclonal antibodies

The primary barrier to the widespread deployment of mAbs in LMICs is their exorbitant cost, driven almost entirely by the reliance on mammalian cell culture (CHO cells) for production. The Bombyx mori platform offers a disruptive solution. The silkworm is uniquely capable of expressing not only simple subunit antigens but also complex, multimeric glycoproteins like full-length IgG monoclonal antibodies.

Unlike E. coli, which cannot perform the necessary post-translational modifications, such as disulfide bond formation and glycosylation, required for functional antibodies, the silkworm BEVS possesses a eukaryotic secretory pathway capable of proper antibody assembly. Studies have demonstrated that silkworm larvae can successfully express and secrete fully assembled, functional human IgG1 antibodies into their hemolymph (Park et al., 2009). In a landmark study producing the CR3022 antibody against SARS-CoV-2, researchers demonstrated that the silkworm-BEVS could produce high yields of scFv, Fab, and full-length IgG formats, with the silkworm-produced IgG exhibiting binding affinities almost identical to those produced in mammalian expression systems (Ebihara et al., 2021). This functional equivalence is critical for regulatory acceptance and clinical deployment.

Crucially, the silkworm platform offers immense yield advantages. Because the actively growing silkworm larva is a biological factory dedicated to protein synthesis, the yield of recombinant IgG can reach up to 36 µg per larva (equivalent to 8 mg/kg of larvae) (Park et al., 2009). By utilizing actively growing larvae rather than pupae, the system maximizes protein yield before the physiological disruptions of metamorphosis occur. This high-yield, low-CapEx production model has the potential to reduce the cost of goods for anti-malarial mAbs by an order of magnitude, transforming them from an exclusive luxury of the Global North into a viable, cost-effective public health tool for malaria-endemic regions.

5 Aligning African sericulture with a farm-to-GMP manufacturing model

The silkworm platform’s reliance on healthy, actively growing larvae aligns perfectly with existing agricultural infrastructure in Africa. This alignment creates a unique and historically unprecedented opportunity: to use Africa’s sericulture sector not merely as a textile industry, but as the upstream biological supply chain for a new generation of locally produced malaria vaccines and monoclonal antibodies. In this model, silkworm farms serve as the source of larvae, which are then collected and transferred to centralized GMP facilities where baculovirus infection and protein purification are performed under controlled pharmaceutical conditions (Figure 2).

Sericulture has a long history in Africa, encompassing the cultivation of the domesticated mulberry silkworm (Bombyx mori). Supported by institutions like the International Centre of Insect Physiology and Ecology (ICIPE) and the Kenya Agricultural & Livestock Research Organization (KALRO) through its National Sericulture Research Centre in Kenya, and major agricultural development projects in Uganda, Rwanda, and Ethiopia, sericulture is an established, low-technology agricultural practice across the continent (Food and Agriculture Organization, 2025; Pilorget and Loizel, 2025). While Africa is the primary focus of this initiative, the global distribution of Morus (mulberry) species extends extensively across Latin America and Asia, where robust sericulture industries already exist in countries like Brazil, India, and Vietnam (Sánchez, 2000). This widespread distribution ensures that the silkworm biomanufacturing model is highly scalable and directly applicable to other major malaria-endemic regions globally. The physical requirements for rearing silkworm larvae as a biological supply are identical to those for textile silk production: mulberry plantations for feed, basic rearing houses with temperature and humidity control, and a trained agricultural workforce.

Critically, the farmers themselves need not be involved in any pharmaceutical process; their role is to rear and supply healthy larvae, keeping the technical complexity of GMP manufacturing within specialized facilities. By leveraging this established agricultural base, African nations can repurpose existing sericulture infrastructure to supply the biological raw materials for advanced biomanufacturing. Instead of constructing $100 million stainless-steel bioreactor facilities for upstream cell expansion, existing sericulture farms can serve as a decentralized network producing healthy silkworm larvae. These larvae are then transported from rural farms to centralized GMP laboratories, where they undergo baculovirus infection and subsequent protein purification. This model creates a high-value cash crop that boosts rural incomes while providing a steady, low-cost supply of the living biomass required for malaria vaccine and mAb production. The farm-to-GMP model simultaneously addresses rural economic development and pharmaceutical manufacturing sovereignty, two goals that are rarely achievable through a single intervention.

6 Integrating AI and decentralized production to support the PAVM

The COVID-19 pandemic delivered an unambiguous verdict on Africa’s structural vulnerability in global health: despite bearing a disproportionate burden of infectious disease, the continent produced less than 1% of the vaccines administered on its territory and received less than 2% of global vaccine doses in the first year of rollout (Africa CDC, 2022; Fleck-Vidal et al., 2025). This catastrophic inequity galvanized continental political will. In 2021, the African Union launched the Partnerships for African Vaccine Manufacturing (PAVM) under the Africa Centres for Disease Control and Prevention (Africa CDC), establishing the most ambitious health manufacturing target in the continent’s history: to produce at least 60% of Africa’s vaccine needs locally by 2040 (Africa CDC, 2022). In February 2024, the PAVM mandate was further broadened to encompass all health products, vaccines, diagnostics, and therapeutics, reflecting a recognition that health sovereignty requires a comprehensive, integrated manufacturing ecosystem rather than a single-product focus (Africa CDC, 2024). Complementing this political framework, Gavi launched the African Vaccine Manufacturing Accelerator (AVMA) in 2024, committing over $1 billion over ten years in milestone-based financing to incentivize WHO-prequalified African manufacturers and create the predictable demand signals that have historically been absent (Fleck-Vidal et al., 2025).

Despite this momentum, Africa’s vaccine manufacturing ecosystem remains nascent and heavily concentrated. A 2025 Lancet Global Health landscape analysis found that only eight African countries have achieved WHO regulatory maturity level 3, the minimum standard required for credible national oversight of vaccine production, and that the continent’s existing manufacturers are predominantly engaged in fill-and-finish operations rather than upstream drug substance production (Fleck-Vidal et al., 2025). The transition from fill-and-finish to full end-to-end manufacturing of novel biologics represents the critical unmet challenge. It is precisely here that the silkworm-based platform, with its radically lower capital requirements and modular scalability, offers a structurally differentiated and strategically complementary pathway.

Kenya is rapidly emerging as the East African anchor for health product manufacturing, driven by strong presidential commitment and a structured national strategy. In October 2023, President William Ruto committed Kenya to producing 50% of its essential health products domestically by 2026 and was subsequently appointed as the African Union Champion for Local Manufacturing, a recognition of Kenya’s leadership in translating political ambition into institutional action (PATH, 2026). The Kenya BioVax Institute, established in 2021, is developing the country’s first dedicated vaccine manufacturing facility, with a mandate to manufacture and commercialize human vaccines and other health products (Kenya BioVax Institute Limited, ).

Kenya’s experience also illustrates the critical limitations of classical manufacturing models: in April 2024, Moderna paused its $200 million plan to build an mRNA manufacturing facility in Kenya, citing insufficient long-term demand signals from African procurement systems, a stark demonstration of the fragility of high-CapEx manufacturing investments in the absence of guaranteed market commitments (Moderna Inc., 2024). This episode underscores precisely why the silkworm platform’s low capital threshold could be of interest for Kenya and the broader East African Community (EAC). At €2–5 million in CapEx, a silkworm-based GMP facility is viable even in the absence of guaranteed multi-year procurement contracts, eventually enabling Kenya to establish drug substance manufacturing capacity for malaria vaccines and mAbs at a fraction of the investment risk that deterred Moderna. Furthermore, Kenya’s established sericulture sector provides an immediately available agricultural supply chain for silkworm larvae, eliminating the need to build upstream biological supply from scratch (Pilorget and Loizel, 2025).

Sub-Saharan Africa accounts for 94% of global malaria cases and 95% of malaria deaths, yet not a single malaria vaccine or anti-malarial monoclonal antibody is currently manufactured on the continent (Medicines for Malaria Venture, 2025; World Health Organization, 2025). The RTS,S/AS01 and R21/Matrix-M vaccines, both produced in facilities located in Europe and India respectively, must be imported at every stage of the supply chain, from bulk antigen to finished dose, leaving African health systems entirely dependent on the pricing decisions, production schedules, and logistical priorities of manufacturers operating thousands of kilometers away (Kumraj et al., 2022; Gavi, the Vaccine Alliance, 2024).

AI and the silkworm platform together provide the two essential drivers for breaking this dependency. AI addresses the scientific bottleneck: by enabling the rational, in silico design of next-generation malaria antigens and broadly protective monoclonal antibodies, it compresses the discovery timeline from years to months and eliminates the costly trial-and-error cycles that have historically made malaria vaccine development so protracted (Nelson et al., 2025; Nishan, 2025). Once an AI-optimized antigen sequence or mAb construct has been validated computationally, it is encoded into a recombinant baculovirus vector and transferred directly to an African GMP facility, a process that requires no proprietary mammalian cell line, no specialized bioreactor infrastructure, and no technology transfer agreement with a foreign manufacturer. The silkworm platform addresses the manufacturing bottleneck: by replacing capital-intensive mammalian cell culture with a biological production system rooted in an agricultural practice already present across East and West Africa, it enables African nations to establish end-to-end drug substance manufacturing capacity for malaria vaccines and mAbs at a CapEx of €2–5 million, a threshold that is achievable within the budget envelopes of national health product manufacturing strategies and accessible to AVMA milestone financing (Felberbaum, 2015; Brühlmann Consulting, 2026).

7 The health economics of malaria and next-generation therapeutics

The economic rationale for investing in AI-driven, silkworm-produced malaria therapeutics is compelling and operates at multiple scales. Malaria exerts a devastating macroeconomic toll on endemic nations. Recent modeling by the African Leaders Malaria Alliance (ALMA) warns that a resurgence in malaria could cost sub-Saharan Africa up to $83 billion in lost GDP by 2030, while sustained investment and elimination efforts could yield a $230 billion GDP boost over the same period (African Leaders Malaria Alliance (ALMA), 2024). At the microeconomic level, the direct medical costs of treating uncomplicated malaria, ranging from $8 to $31 per episode, consume a disproportionate and catastrophic share of household income for families living near the poverty line (Lamesgen et al., 2025).

Current first-generation vaccines, while historic achievements, offer only partial economic relief. The R21 vaccine, priced at approximately $2.99 to $3.90 per dose under the recent Gavi-UNICEF equitable pricing agreement, achieves a highly favorable cost-effectiveness ratio of roughly $39 per Disability-Adjusted Life Year (DALY) averted, comparable to insecticide-treated bed nets (1Day Sooner / Imperial College London, 2024; Gavi and UNICEF, 2025). However, because these vaccines achieve only approximately 36% (RTS,S) to 71% (R21) efficacy, substantial economic value remains unrealized. The cost-effectiveness of a vaccine achieving 85% or greater efficacy would be dramatically superior, averting a far greater number of clinical episodes, hospitalizations, and deaths per dollar invested.

The convergence of AI design and silkworm biomanufacturing acts as an economic multiplier. By utilizing AI to design highly efficacious, strain-transcending antigens and mAbs targeting >85% efficacy, the number of averted clinical cases and hospitalizations increases dramatically, compounding direct medical savings. Simultaneously, by shifting production to the silkworm platform, the capital expenditure required to build manufacturing capacity drops from >€50 million to under €5 million, and the cost of goods per dose falls below €1 (Amasawa et al., 2021; Brühlmann Consulting, 2026). This dual effect, higher efficacy driving greater health system savings, combined with radically lower production costs, fundamentally transforms the return on investment for malaria elimination, making the widespread deployment of next-generation vaccines and mAbs economically feasible for LMICs. The silkworm platform’s ability to produce both vaccine antigens and therapeutic mAbs within the same infrastructure further enhances the economic case, enabling a single investment in manufacturing capacity to serve multiple product pipelines simultaneously.

8 Conclusion

The persistent global burden of malaria and the structural limitations of centralized biologic manufacturing demand innovative, convergent solutions. Artificial Intelligence is transforming malaria research by enabling the rational design of optimized, strain-transcending immunogens and highly potent monoclonal antibodies, compressing discovery timelines from years to months. Furthermore, AI is accelerating the discovery of novel antimalarial drugs and optimizing adjuvant design, creating a comprehensive technological toolkit against the parasite. However, realizing the full potential of these AI-designed therapeutics requires a parallel revolution in manufacturing. The Bombyx mori platform offers a highly scalable, low-cost, and environmentally sustainable alternative to centralized mammalian cell culture. By maximizing protein yield in actively growing larvae and eliminating the need for capital-intensive bioreactor infrastructure, the silkworm platform enables a genuinely novel manufacturing model. African sericulture farms supply healthy larvae at scale; these larvae are then transferred to centralized GMP facilities where baculovirus infection and downstream protein purification take place under controlled pharmaceutical conditions. This farm-to-GMP supply chain aligns biological production with existing agricultural ecosystems, providing a sustainable and culturally embedded framework that directly supports the Partnerships for African Vaccine Manufacturing (PAVM). By uniting the precision of AI with the scalable potential of the silkworm platform, the global health community can accelerate the delivery of equitable, next-generation malaria vaccines and mAbs, advancing the African Union’s vision of health sovereignty and ensuring that the communities bearing the greatest burden of malaria are the first to benefit from the next generation of life-saving therapeutics.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. Generative AI was used to create the illustration presented in Figures 1 and 2. In addition, AI tools were used to improve the English language and readability of the 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: Nicolas F. Delahaye, Sanofi US, United States

Reviewed by: Adam Wesley Whisnant, Hannover Medical School, Germany

Jose Gregorio Marchan Alvarez, Karolinska Institutet (KI), Sweden