Draft — Africa Health-Tech Accelerator

MOTIVATION LETTER

The Africa Health-Tech Accelerator selects early-stage startups building AI and digital health solutions for African markets. My venture predicts drug resistance mutations from protein sequence alone, using a protein language model that requires no crystal structure. This matters for Africa because resistance to antimicrobials and oncology drugs is a silent crisis across the continent, and the tools to predict it have historically required structural data that does not exist for most clinically relevant proteins.

I am Eniola Olutogun, a pharmacist turned machine learning engineer. I built and validated a classifier that combines ESM-2 protein language model delta-embeddings with ECFP4 drug fingerprints. On the Platinum benchmark, 553 mutations with protein-grouped cross-validation, the model achieves an AUROC of 0.804 plus or minus 0.025. This beats the published state of the art, mCSM-lig, which scores around 0.70. The model covers 100 percent of mutations in the benchmark, while structure-limited tools cover roughly 18 percent. On SKEMPI 2.0, the AUROC is 0.634, which defines the current gap I am closing with a fine-tuned ESM-2 on the SKEMPI 3K mutation set.

The venture is pre-seed, proof-of-concept validated, and not yet incorporated. I am the sole author and operator. The roadmap is concrete: fine-tune ESM-2 on SKEMPI 3K to reach an AUROC of at least 0.70, then run a pilot with Servier in Suresnes, then convert that pilot into recurring revenue. Named partners include Servier, Paris-Saclay I2BC, Institut Pasteur, and Sanofi in Gentilly.

The Africa Health-Tech Accelerator fits this venture because it targets technology-driven healthcare solutions with clear impact and scalable business models. Drug resistance prediction is a digital health infrastructure problem. It sits upstream of diagnostics, treatment selection, and drug development. A platform that predicts resistance from sequence data can inform antibiotic stewardship programs, guide oncology treatment choices, and de-risk drug development for pathogens and cancers prevalent in African populations. The regulatory-aware model I have built, shaped by my pharmacy background, aligns with the accelerator's emphasis on compliant digital health infrastructure.

I am applying for the six-month structured programme because I need mentorship on regulatory pathways across African markets, investor readiness support, and access to a biopharma network. The compute credits offered by accelerators in this space would directly fund the ESM-2 fine-tuning phase. The current support pipeline includes SEMIA and Quest for Health meetings in progress, WILCO One BioTech in October 2026, and applications submitted to IncubAlliance and AI House. The EIC Accelerator and BPI i-Lab are future targets. The Africa Health-Tech Accelerator is the right next step because it matches the venture's stage, sector, and geographic focus.

RESEARCH STATEMENT

The problem is drug resistance. Antimicrobial resistance kills an estimated 1.27 million people directly each year, and oncology drug failure due to resistance accounts for a substantial fraction of cancer mortality. In African markets, the burden is compounded by limited diagnostic infrastructure and limited access to structural biology tools. Most clinically relevant mutations are in proteins with no experimentally determined crystal structure. Structure-based prediction tools simply cannot score them.

My approach removes the structure requirement. I use ESM-2, a protein language model pretrained on 65 million protein sequences, to generate embeddings for wild-type and mutant sequences. The delta between those embeddings captures the functional effect of a mutation. I concatenate that delta with an ECFP4 fingerprint of the drug in question, then train a Random Forest classifier to predict resistance. The method is fully sequence-based, which means it scales to any protein with a known sequence, regardless of structural data availability.

The validation status is honest and specific. On the Platinum benchmark, 553 mutations across protein groups, the model achieves an AUROC of 0.804 with a standard deviation of 0.025 under protein-grouped cross-validation. This is a rigorous evaluation because it prevents leakage between training and test mutations from the same protein. The published state of the art, mCSM-lig, achieves approximately 0.70 on the same benchmark. My model also achieves 100 percent mutation coverage, versus roughly 18 percent for structure-limited tools. On SKEMPI 2.0, a binding affinity benchmark, the AUROC drops to 0.634. That gap is the target of the next development phase.

The next phase is fine-tuning ESM-2 on the SKEMPI 3K mutation set. The goal is an AUROC of at least 0.70 on SKEMPI 2.0, which would close the gap between the Platinum performance and the binding affinity performance. Once that threshold is met, the model is ready for a pilot with Servier in Suresnes. The pilot would test the platform on a specific oncology or antimicrobial target selected by Servier's discovery team. Success in that pilot converts to an annual recurring revenue license.

The business model is a software-as-a-service platform for biopharma discovery teams. The customer is a pharma company that needs to prioritize which mutations to test experimentally, or which drug candidates are likely to fail due to resistance. The platform reduces the experimental burden and shortens the discovery cycle. The market is global, but the African angle is specific: the platform can be deployed for pathogens and cancers that are under-served by structural biology resources, and it can be accessed remotely, which suits markets with limited lab infrastructure.

The named partners are real and engaged. Servier in Suresnes is the pilot target. Paris-Saclay I2BC and Institut Pasteur provide domain expertise in structural biology and microbiology. Sanofi in Gentilly is a potential second pilot or commercial partner. The support pipeline includes SEMIA and Quest for Health meetings in progress, WILCO One BioTech in October 2026, and applications submitted to IncubAlliance and AI House. The EIC Accelerator and BPI i-Lab are future targets for non-dilutive funding.

The Africa Health-Tech Accelerator is the right programme because it explicitly targets early-stage African health-tech startups building AI solutions. The venture is pre-seed, proof-of-concept validated, and not yet incorporated, which matches the early-stage criterion. The technology is AI-driven and addresses critical healthcare delivery challenges, specifically drug resistance, which is a growing threat in African markets. The business model is scalable because it is software-based and does not require local hardware deployment. The market potential is strong because the platform serves both global pharma and regional health systems.

ESSAY RESPONSE: HEALTHCARE IMPACT

Drug resistance is a critical healthcare delivery challenge in African markets because it directly undermines the effectiveness of existing treatments. When a patient with tuberculosis or malaria or a resistant bacterial infection receives a drug that no longer works, the consequences are prolonged illness, higher mortality, and increased cost to an already strained health system. My platform predicts resistance from protein sequence alone, which means it can be applied to any pathogen or cancer with a sequenced genome. This is a diagnostic and treatment-guidance tool that does not require expensive structural biology infrastructure.

The proof-of-concept demonstrates the impact. An AUROC of 0.804 on the Platinum benchmark means the model correctly distinguishes resistant from non-resistant mutations in 80 percent of cases, across 553 mutations and multiple protein groups. The 100 percent mutation coverage means no mutation is excluded for lack of structural data. This is a direct improvement over structure-limited tools that cover only 18 percent of mutations. For African health systems, this means a wider net of actionable predictions.

The venture also addresses drug development cost. Predicting resistance early in the discovery pipeline reduces the number of experimental assays required and de-risks candidate selection. For a continent that needs affordable drugs developed for its specific pathogens, this is a meaningful contribution. The platform is software-based, so deployment is remote and low-cost. The regulatory-aware model, shaped by my pharmacy background, ensures the platform is designed for compliance from the start, which aligns with the accelerator's focus on compliant digital health infrastructure.

ESSAY RESPONSE: SCALABILITY AND MARKET POTENTIAL

The business model is software-as-a-service for biopharma discovery teams. The customer is a pharma company that needs to prioritize mutations for experimental testing or identify drug candidates likely to fail due to resistance. The platform is sequence-based, so it scales to any protein with a known sequence. There is no dependency on crystal structures, which are expensive and slow to produce. This removes a major bottleneck in the drug discovery workflow.

The market is global, but the African angle is specific and defensible. Pathogens prevalent in African markets, such as drug-resistant tuberculosis and malaria, have genomic data that is increasingly available. The platform can be deployed remotely, which suits markets with limited lab infrastructure. The pilot with Servier in Suresnes is the first revenue milestone. The roadmap to annual recurring revenue is clear: fine-tune ESM-2 on SKEMPI 3K to reach an AUROC of at least 0.70, run the Servier pilot, then convert the pilot to a license.

The accelerator's emphasis on scalable business models and strong market potential matches this venture. The platform is not a point solution; it is a platform that can be applied across therapeutic areas, from oncology to antivirals to antimicrobial resistance. The compute requirements are modest and can be supported by accelerator compute credits. The venture is pre-seed and not yet incorporated, which means the accelerator's six-month structured programme, mentorship, and investor readiness support are directly applicable to the current stage.

CHECKLIST

- [ ] Confirm the application deadline on the Africa Health-Tech Accelerator website
- [ ] Verify the programme's eligibility criteria for pre-incorporation ventures
- [ ] Confirm whether the programme requires a pitch deck or video submission
- [ ] Prepare a one-page executive summary of the venture for the application form
- [ ] Gather validation metrics: AUROC 0.804 on Platinum, 0.634 on SKEMPI 2.0, 100 percent mutation coverage
- [ ] Prepare a list of named partners: Servier, Paris-Saclay I2BC, Institut Pasteur, Sanofi
- [ ] Confirm the status of SEMIA and Quest for Health meetings for the application timeline
- [ ] Confirm the status of IncubAlliance and AI House applications for the application timeline
- [ ] Prepare a compute budget for the ESM-2 fine-tuning phase on SKEMPI 3K
- [ ] Draft a regulatory compliance summary for African markets, based on pharmacy background
- [ ] Prepare a one-slide visual of the model architecture: ESM-2 delta-embeddings plus ECFP4 fingerprints plus Random Forest
- [ ] Confirm the SKEMPI 3K dataset access and licensing for the fine-tuning phase
- [ ] Prepare a timeline for the Servier pilot, contingent on the AUROC threshold of 0.70

EDITOR NOTES

- Eligibility risk: the venture is not yet incorporated. Confirm whether the accelerator accepts pre-incorporation ventures or requires a legal entity. If required, plan for incorporation in a relevant African jurisdiction or France.
- Fact verification: confirm the Platinum benchmark AUROC of 0.804 plus or minus 0.025 and the SKEMPI 2.0 AUROC of 0.634 are from the latest validation run. Confirm the mCSM-lig comparison figure of approximately 0.70 is from the published benchmark paper.
- Gap to address: the profile does not specify the founder's current location or nationality. The accelerator is pan-African, so confirm the founder's eligibility and whether the venture is registered or planned to be registered in an African market.
- Personal detail needed: the application should include the founder's specific pharmacy background and any direct clinical experience with drug resistance in African settings. Insert a short paragraph on this in the motivation letter if the application form allows.
- Timeline risk: the WILCO One BioTech programme is listed for October 2026. Confirm whether the Africa Health-Tech Accelerator programme runs concurrently and whether participation in both is feasible or conflicts with the six-month structured programme.

MOTIVATION LETTER

The Africa Health-Tech Accelerator selects early-stage startups building AI and digital health solutions for African markets. My venture predicts drug resistance mutations from protein sequence alone, using a protein language model that requires no crystal structure. This matters for Africa because resistance to antimicrobials and oncology drugs is a silent crisis across the continent, and the tools to predict it have historically required structural data that does not exist for most clinically relevant proteins.

I am Eniola Olutogun, a pharmacist turned machine learning engineer. I built and validated a classifier that combines ESM-2 protein language model delta-embeddings with ECFP4 drug fingerprints. On the Platinum benchmark, 553 mutations with protein-grouped cross-validation, the model achieves an AUROC of 0.804 plus or minus 0.025. This beats the published state of the art, mCSM-lig, which scores around 0.70. The model covers 100 percent of mutations in the benchmark, while structure-limited tools cover roughly 18 percent. On SKEMPI 2.0, the AUROC is 0.634, which defines the current gap I am closing with a fine-tuned ESM-2 on the SKEMPI 3K mutation set.

The venture is pre-seed, proof-of-concept validated, and not yet incorporated. I am the sole author and operator. The roadmap is concrete: fine-tune ESM-2 on SKEMPI 3K to reach an AUROC of at least 0.70, then run a pilot with Servier in Suresnes, then convert that pilot into recurring revenue. Named partners include Servier, Paris-Saclay I2BC, Institut Pasteur, and Sanofi in Gentilly.

The Africa Health-Tech Accelerator fits this venture because it targets technology-driven healthcare solutions with clear impact and scalable business models. Drug resistance prediction is a digital health infrastructure problem. It sits upstream of diagnostics, treatment selection, and drug development. A platform that predicts resistance from sequence data can inform antibiotic stewardship programs, guide oncology treatment choices, and de-risk drug development for pathogens and cancers prevalent in African populations. The regulatory-aware model I have built, shaped by my pharmacy background, aligns with the accelerator's emphasis on compliant digital health infrastructure.

I am applying for the six-month structured programme because I need mentorship on regulatory pathways across African markets, investor readiness support, and access to a biopharma network. The compute credits offered by accelerators in this space would directly fund the ESM-2 fine-tuning phase. The current support pipeline includes SEMIA and Quest for Health meetings in progress, WILCO One BioTech in October 2026, and applications submitted to IncubAlliance and AI House. The EIC Accelerator and BPI i-Lab are future targets. The Africa Health-Tech Accelerator is the right next step because it matches the venture's stage, sector, and geographic focus.

RESEARCH STATEMENT

The problem is drug resistance. Antimicrobial resistance kills an estimated 1.27 million people directly each year, and oncology drug failure due to resistance accounts for a substantial fraction of cancer mortality. In African markets, the burden is compounded by limited diagnostic infrastructure and limited access to structural biology tools. Most clinically relevant mutations are in proteins with no experimentally determined crystal structure. Structure-based prediction tools simply cannot score them.

My approach removes the structure requirement. I use ESM-2, a protein language model pretrained on 65 million protein sequences, to generate embeddings for wild-type and mutant sequences. The delta between those embeddings captures the functional effect of a mutation. I concatenate that delta with an ECFP4 fingerprint of the drug in question, then train a Random Forest classifier to predict resistance. The method is fully sequence-based, which means it scales to any protein with a known sequence, regardless of structural data availability.

The validation status is honest and specific. On the Platinum benchmark, 553 mutations across protein groups, the model achieves an AUROC of 0.804 with a standard deviation of 0.025 under protein-grouped cross-validation. This is a rigorous evaluation because it prevents leakage between training and test mutations from the same protein. The published state of the art, mCSM-lig, achieves approximately 0.70 on the same benchmark. My model also achieves 100 percent mutation coverage, versus roughly 18 percent for structure-limited tools. On SKEMPI 2.0, a binding affinity benchmark, the AUROC drops to 0.634. That gap is the target of the next development phase.

The next phase is fine-tuning ESM-2 on the SKEMPI 3K mutation set. The goal is an AUROC of at least 0.70 on SKEMPI 2.0, which would close the gap between the Platinum performance and the binding affinity performance. Once that threshold is met, the model is ready for a pilot with Servier in Suresnes. The pilot would test the platform on a specific oncology or antimicrobial target selected by Servier's discovery team. Success in that pilot converts to an annual recurring revenue license.

The business model is a software-as-a-service platform for biopharma discovery teams. The customer is a pharma company that needs to prioritize which mutations to test experimentally, or which drug candidates are likely to fail due to resistance. The platform reduces the experimental burden and shortens the discovery cycle. The market is global, but the African angle is specific: the platform can be deployed for pathogens and cancers that are under-served by structural biology resources, and it can be accessed remotely, which suits markets with limited lab infrastructure.

The named partners are real and engaged. Servier in Suresnes is the pilot target. Paris-Saclay I2BC and Institut Pasteur provide domain expertise in structural biology and microbiology. Sanofi in Gentilly is a potential second pilot or commercial partner. The support pipeline includes SEMIA and Quest for Health meetings in progress, WILCO One BioTech in October 2026, and applications submitted to IncubAlliance and AI House. The EIC Accelerator and BPI i-Lab are future targets for non-dilutive funding.

The Africa Health-Tech Accelerator is the right programme because it explicitly targets early-stage African health-tech startups building AI solutions. The venture is pre-seed, proof-of-concept validated, and not yet incorporated, which matches the early-stage criterion. The technology is AI-driven and addresses critical healthcare delivery challenges, specifically drug resistance, which is a growing threat in African markets. The business model is scalable because it is software-based and does not require local hardware deployment. The market potential is strong because the platform serves both global pharma and regional health systems.

ESSAY RESPONSE: HEALTHCARE IMPACT

Drug resistance is a critical healthcare delivery challenge in African markets because it directly undermines the effectiveness of existing treatments. When a patient with tuberculosis or malaria or a resistant bacterial infection receives a drug that no longer works, the consequences are prolonged illness, higher mortality, and increased cost to an already strained health system. My platform predicts resistance from protein sequence alone, which means it can be applied to any pathogen or cancer with a sequenced genome. This is a diagnostic and treatment-guidance tool that does not require expensive structural biology infrastructure.

The proof-of-concept demonstrates the impact. An AUROC of 0.804 on the Platinum benchmark means the model correctly distinguishes resistant from non-resistant mutations in 80 percent of cases, across 553 mutations and multiple protein groups. The 100 percent mutation coverage means no mutation is excluded for lack of structural data. This is a direct improvement over structure-limited tools that cover only 18 percent of mutations. For African health systems, this means a wider net of actionable predictions.

The venture also addresses drug development cost. Predicting resistance early in the discovery pipeline reduces the number of experimental assays required and de-risks candidate selection. For a continent that needs affordable drugs developed for its specific pathogens, this is a meaningful contribution. The platform is software-based, so deployment is remote and low-cost. The regulatory-aware model, shaped by my pharmacy background, ensures the platform is designed for compliance from the start, which aligns with the accelerator's focus on compliant digital health infrastructure.

ESSAY RESPONSE: SCALABILITY AND MARKET POTENTIAL

The business model is software-as-a-service for biopharma discovery teams. The customer is a pharma company that needs to prioritize mutations for experimental testing or identify drug candidates likely to fail due to resistance. The platform is sequence-based, so it scales to any protein with a known sequence. There is no dependency on crystal structures, which are expensive and slow to produce. This removes a major bottleneck in the drug discovery workflow.

The market is global, but the African angle is specific and defensible. Pathogens prevalent in African markets, such as drug-resistant tuberculosis and malaria, have genomic data that is increasingly available. The platform can be deployed remotely, which suits markets with limited lab infrastructure. The pilot with Servier in Suresnes is the first revenue milestone. The roadmap to annual recurring revenue is clear: fine-tune ESM-2 on SKEMPI 3K to reach an AUROC of at least 0.70, run the Servier pilot, then convert the pilot to a license.

The accelerator's emphasis on scalable business models and strong market potential matches this venture. The platform is not a point solution; it is a platform that can be applied across therapeutic areas, from oncology to antivirals to antimicrobial resistance. The compute requirements are modest and can be supported by accelerator compute credits. The venture is pre-seed and not yet incorporated, which means the accelerator's six-month structured programme, mentorship, and investor readiness support are directly applicable to the current stage.

CHECKLIST

- [ ] Confirm the application deadline on the Africa Health-Tech Accelerator website
- [ ] Verify the programme's eligibility criteria for pre-incorporation ventures
- [ ] Confirm whether the programme requires a pitch deck or video submission
- [ ] Prepare a one-page executive summary of the venture for the application form
- [ ] Gather validation metrics: AUROC 0.804 on Platinum, 0.634 on SKEMPI 2.0, 100 percent mutation coverage
- [ ] Prepare a list of named partners: Servier, Paris-Saclay I2BC, Institut Pasteur, Sanofi
- [ ] Confirm the status of SEMIA and Quest for Health meetings for the application timeline
- [ ] Confirm the status of IncubAlliance and AI House applications for the application timeline
- [ ] Prepare a compute budget for the ESM-2 fine-tuning phase on SKEMPI 3K
- [ ] Draft a regulatory compliance summary for African markets, based on pharmacy background
- [ ] Prepare a one-slide visual of the model architecture: ESM-2 delta-embeddings plus ECFP4 fingerprints plus Random Forest
- [ ] Confirm the SKEMPI 3K dataset access and licensing for the fine-tuning phase
- [ ] Prepare a timeline for the Servier pilot, contingent on the AUROC threshold of 0.70

EDITOR NOTES

- Eligibility risk: the venture is not yet incorporated. Confirm whether the accelerator accepts pre-incorporation ventures or requires a legal entity. If required, plan for incorporation in a relevant African jurisdiction or France.
- Fact verification: confirm the Platinum benchmark AUROC of 0.804 plus or minus 0.025 and the SKEMPI 2.0 AUROC of 0.634 are from the latest validation run. Confirm the mCSM-lig comparison figure of approximately 0.70 is from the published benchmark paper.
- Gap to address: the profile does not specify the founder's current location or nationality. The accelerator is pan-African, so confirm the founder's eligibility and whether the venture is registered or planned to be registered in an African market.
- Personal detail needed: the application should include the founder's specific pharmacy background and any direct clinical experience with drug resistance in African settings. Insert a short paragraph on this in the motivation letter if the application form allows.
- Timeline risk: the WILCO One BioTech programme is listed for October 2026. Confirm whether the Africa Health-Tech Accelerator programme runs concurrently and whether participation in both is feasible or conflicts with the six-month structured programme.

AI Draft — Africa Health-Tech Accelerator