Comments
AI Safety in Biotech: Epitope Design for Phage and AAV Vaccines in the Global South
5 min readSep 16, 2025
5
TL;DR
AI-driven epitope prediction is revolutionizing vaccine development by rapidly identifying immune targets for phage-based and AAV (adeno-associated virus), based platforms. These tools hold promise for affordable cancer immunotherapies, antimicrobial resistance, and pandemic preparedness. Yet, without attention to AI safety—bias in datasets, opaque black-box pipelines, and dual-use risks—these innovations could reinforce inequities or pose new hazards, especially in the Global South. The path forward requires inclusive, open-source pipelines, collaborative pilot projects, and capacity building. My vision is to transform CenPBaT into a regional hub, develop globally inclusive pipelines, and train the next generation for safe, equitable AI-biotech innovation
When we discuss AI safety, the conversation often centers on general-purpose systems like large language models or speculative future AGI. However, AI is already transforming another domain with enormous stakes for human health: biotechnology.
Specifically, epitope prediction—the use of AI to identify antigenic regions for vaccine development—is advancing rapidly. These tools promise to accelerate the design of phage-based and AAV-based vaccines, opening up possibilities for affordable cancer immunotherapies and rapid-response platforms against infectious diseases.
Yet, without attention to AI safety in biotech, these same tools could amplify risks: unreliable epitopes, biased datasets, opaque design pipelines, or even dual-use misuse. The Global South, where health burdens are greatest, may be the most vulnerable to these hazards.
The need for new vaccine platforms is urgent:
• Rising Burdens: Breast cancer in sub-Saharan Africa has some of the highest mortality rates worldwide, largely due to late diagnosis and unaffordable therapies. At the same time, antimicrobial resistance (AMR) is eroding the effectiveness of antibiotics.
• Pandemic Uncertainties: The COVID-19 pandemic demonstrated how quickly pathogens can upend societies. The next pandemic—whether viral, bacterial, or fungal—may require vaccines that are faster, cheaper, and more globally adaptable than our current platforms.
• The Opportunity: Phage display systems and AAV vectors are promising vaccine platforms. AI-driven epitope design could make them adaptable for cancers, AMR pathogens, or emerging pandemics.
AI-powered epitope prediction is at the forefront of modern vaccine development, particularly in a pandemic context. Instead of relying on traditional, time-consuming lab methods, AI algorithms can analyze vast amounts of viral genetic data to rapidly identify the most promising vaccine targets. This process, often called reverse vaccinology, significantly compresses the timeline from pathogen identification to vaccine candidate selection.
• Accelerating Timelines: AI tools can sift through a pathogen’s entire proteome and identify potential epitope “hotspots” in a fraction of the time it would take human researchers. For COVID-19 vaccines, AI was crucial in quickly identifying the spike protein as the optimal target.
• Predicting Efficacy: AI models can simulate how different epitopes will interact with human immune systems, predicting their effectiveness across a wide range of HLA genotypes.
• Universal Vaccine Design: AI can identify conserved epitopes—shared across multiple viral strains or families—enabling universal vaccines that offer broad protection against variants and future pathogens.
AI is not a magic bullet. An unsafe AI pipeline could propose epitopes that are non-immunogenic or even harmful, wasting scarce resources or creating risks in clinical use.
Key concerns include:
• Bias: Most epitope prediction models are trained on Euro-American datasets. Without correction, vaccines designed this way may underperform in African, Asian, or Latin American populations due to diverse HLA alleles.
• Transparency: Closed, “black-box” AI systems could centralize vaccine discovery in wealthy countries, excluding Global South labs from the innovation process.
• Dual-Use Risk: The same tools used to predict protective epitopes could also be misapplied to design immune-evasive or harmful peptides.
This is not theoretical. Several vaccine candidates have already been designed using AI or immunoinformatics pipelines:
• mRNA-4157/V940 (Moderna & Merck) – A personalized cancer vaccine where AI algorithms select tumor neoantigens from patient sequencing (clinical trials ongoing).
• Evaxion Biotech’s AI-Derived Cancer Vaccines – Identifies antigenic “hotspots” from endogenous retroviruses (ERVs), demonstrating T-cell activation and tumor inhibition in preclinical models.
• UB-612 COVID-19 Vaccine – A peptide-based vaccine built using epitope prediction from spike and non-spike proteins, evaluated in clinical trials.
These examples illustrate both the promise and the risk: AI can compress timelines and broaden antigen discovery, but without robust safety checks, the outputs may fail or create unforeseen hazards.
To build a safer, more equitable future for AI in biotech, we must act now. The problems of bias, transparency, and dual-use risk are not theoretical—they are present today.
Recommended steps:
1. Partner with Researchers – Collaborate with Global South labs to stress-test safe AI pipelines for epitope prediction.
2. Support Pilot Projects – Fund small-scale projects showing responsible AI use in low-resource vaccine development.
3. Create Open-Source Prototypes – Develop an AI-enabled open-source vaccine for a local health challenge, such as breast cancer or endemic infections.
4. Build Biotech Infrastructure – Register a biotech company in Nigeria dedicated to AI-enabled vaccine design and innovation.
5. Establish a Working Group – Bring together AI researchers, immunologists, and biosecurity experts to shape AI safety in biotech.
My long-term career goal is to anchor this work in the Global South, bridging local health challenges with global AI-biotech innovation.
Ways to help:
• Career Capital: Visiting researcher opportunities; mentorship on responsible AI use in biotech.
• Pipeline Development: Collaborate on inclusive epitope prediction pipelines adapted to global populations with diverse HLA profiles.
• CenPBaT Transformation: Support the Centre for Phage Biology and Therapeutics (CenPBaT) as a hub for AI-enabled vaccines and phage therapeutics.
• Training & Capacity Building: Partner to create programs training African and Global South students in AI-for-biology.
• Funding: Provide seed funding for pilot AI-driven vaccine projects and help establish a biotech company in Nigeria.
Looking ahead, I envision the following steps shaping my trajectory in AI-enabled biotechnology and global health:
1. Global Research Exposure – I will plan to visit leading laboratories at the intersection of AI and biotechnology to gain hands-on expertise in responsible epitope design. These experiences will anchor my ability to integrate technical excellence with safety frameworks.
2. Development of a Next-Generation Pipeline – I aim to adapt and validate globally inclusive epitope prediction pipelines, ensuring they account for diverse HLA genotypes and disease burdens across populations. This future pipeline will serve as a reliable tool for designing vaccines against cancers, AMR pathogens, and emerging infectious threats worldwide.
3. Transforming CenPBaT into a Regional Hub – Over the coming years, I will work to position the Centre for Phage Biology and Therapeutics (CenPBaT) as a leading futuristic hub for AI-enabled vaccine and phage therapeutic development in Africa, while actively linking to global networks of innovation.
4. Training the Next Generation – I will establish forward-looking programs to prepare students and young scientists in the Global South and beyond to lead in AI-biotech innovation. These initiatives will seed a self-sustaining, globally connected ecosystem capable of shaping the future of safe biotechnology.
Epitope prediction may seem like a technical niche, but it is a critical test case for how AI will shape biotechnology more broadly.
• If we ignore AI safety, we risk inequitable, unreliable, or unsafe applications.
• If we act now, we can build open, safe, and globally inclusive AI pipelines that deliver the next generation of affordable vaccines against cancer, AMR, and future pandemics.
This is where AI safety meets global health—and the Global South cannot be left behind.
