Draft — Healthcare Access Initiative Grant 2026

MOTIVATION LETTER

The Healthcare Access Initiative Grant 2026 funds health technology that reaches underserved populations in emerging markets. I am applying with TOPOLOGIX, a sequence-based drug-resistance prediction platform that addresses a specific failure in precision medicine: structure-based tools like mCSM-lig cover only 18 percent of clinically relevant mutations because they require a resolved protein structure. In Nigeria, where genomic surveillance infrastructure is limited and structural data is scarcer than in high-income settings, that coverage gap is a barrier to treating drug-resistant infections and cancers.

TOPOLOGIX uses ESM-2 protein language model delta-embeddings combined with Morgan/ECFP drug fingerprints and a Random Forest classifier to predict drug-resistance mutations from sequence alone. On the Platinum benchmark of 553 mutations, it achieves an AUROC of 0.804 with a standard deviation of 0.025, and 0.634 on SKEMPI 2.0. It beats structure-based baselines while covering 100 percent of mutations. The method is fully sequence-based, which means it works wherever sequencing data exists, regardless of structural resolution. That is the scalability property the IFC mission requires: the tool does not depend on infrastructure that emerging-market laboratories do not have.

My path to this project is direct. I hold a B.Pharm from the University of Ibadan with a 2:1 Upper Division, am a PCN-licensed pharmacist, and am enrolled in the M.Sc. Digital Health program at the Hasso Plattner Institute and University of Potsdam. I have built and validated four independent DuckDB-based ingest-to-analyze pipelines across life sciences and other domains, and I self-host local LLM serving infrastructure. My pre-registered replication of a published topological approach to hERG cardiotoxicity found that topological features did not beat a plain descriptor baseline (AUROC 0.8426 vs 0.8782), and my interface-topology study for drug resistance found AUROC values of 0.425 and 0.485, ruling out interface geometry as the driver. Those results are why TOPOLOGIX exists. I pivoted to sequence representations because the data showed structure-based geometry was not the answer.

The grant would fund three things: validation of TOPOLOGIX against Nigerian clinical isolates, development of a laboratory-facing interface for resistance reporting, and a pilot deployment with a partner clinical site. The budget range of $200,000 to $4 million is appropriate for the infrastructure build and the validation study. I am pre-revenue and pre-incorporation, which I state plainly. The funds would support incorporation, compute infrastructure, and the validation pipeline that turns a research method into a deployed tool.

The measurable impact target is concrete: reduce the time from resistance mutation detection to treatment decision in Nigerian clinical settings, and provide resistance prediction coverage for the 82 percent of mutations that structure-based tools currently miss. That is the access improvement this programme exists to fund.

RESEARCH STATEMENT

TOPOLOGIX is a computational platform for predicting drug-resistance mutations from protein sequence alone. It was developed in response to a falsified hypothesis. Between 2024 and 2025, I tested whether bipartite persistent homology of protein-ligand interface geometry could predict hERG cardiotoxicity and drug resistance. The cardiotoxicity replication, pre-registered and powered, found topological features underperformed a plain descriptor baseline (AUROC 0.8426 vs 0.8782). The resistance study found AUROC values of 0.425 and 0.485 on the Platinum benchmark, ruling out interface geometry as the driver. Those results are published as preprints and are under peer review. They are the empirical basis for the current approach.

TOPOLOGIX replaces structure-based geometry with sequence-based representations. The architecture has three components. First, ESM-2 protein language model delta-embeddings capture the mutational change in the protein sequence representation. Second, Morgan/ECFP fingerprints encode the drug. Third, a Random Forest classifier learns the resistance decision boundary. On the Platinum benchmark of 553 mutations, the model achieves AUROC 0.804 plus or minus 0.025. On SKEMPI 2.0, it achieves 0.634. It outperforms mCSM-lig, a structure-based tool, which scores approximately 0.70 on the same tasks while covering only 18 percent of mutations. TOPOLOGIX covers 100 percent of mutations because it requires only sequence data.

The scientific contribution is the demonstration that sequence representation alone can carry resistance signal that interface geometry cannot. The translational contribution is coverage. In emerging markets, structural data is a luxury. Sequencing is increasingly available. A tool that works from sequence alone is deployable in settings where structure-based tools are not.

The validation plan has three gates. Gate one, already passed, is benchmark performance on Platinum and SKEMPI 2.0. Gate two, the focus of the proposed funding period, is validation against Nigerian clinical isolates. I will partner with a clinical genomics laboratory in Nigeria to obtain resistance mutation data from local pathogens and compare TOPOLOGIX predictions against phenotypic resistance assays. Gate three is deployment: a laboratory-facing interface that accepts a sequence and a drug, and returns a resistance prediction with confidence intervals.

The implementation plan spans 24 months. Months one through three: incorporation, compute infrastructure procurement, and partnership agreements with the Nigerian clinical site. Months four through twelve: isolate collection, sequencing, and model fine-tuning on local data. Months thirteen through eighteen: interface development and usability testing with laboratory staff. Months nineteen through twenty-four: pilot deployment and outcome measurement.

The measurable impact metric is coverage and accuracy. The target is to maintain AUROC above 0.75 on Nigerian clinical isolates while providing predictions for 100 percent of submitted sequences. The secondary metric is time-to-decision: reduce the interval from sample receipt to resistance report from weeks to days.

The budget request is $350,000. This covers incorporation and legal fees ($15,000), compute infrastructure including GPU servers and storage ($60,000), sequencing and laboratory costs for the validation study ($120,000), personnel for a research assistant and a software engineer ($100,000), and deployment and training costs ($55,000). The funding moves TOPOLOGIX from a validated research method to a deployed tool in a Nigerian clinical setting.

ESSAY: HEALTHCARE ACCESS CHALLENGE

The healthcare access challenge I address is the structural gap in drug-resistance prediction. Resistance testing in Nigeria relies on phenotypic culture methods that take weeks and require laboratory infrastructure that is concentrated in urban tertiary centers. Genotypic methods exist but are limited by reference databases built from non-African populations. The result is that treatment decisions for drug-resistant infections are often made without resistance data, leading to empirical therapy failure, prolonged illness, and increased mortality.

The scale of the problem is documented. Antimicrobial resistance is a leading cause of death in sub-Saharan Africa, and Nigeria carries a substantial burden. The tools to address it are misaligned with the context: they require structural data that does not exist for most clinically relevant mutations, and they are validated on populations that do not reflect Nigerian microbial diversity.

TOPOLOGIX addresses this by removing the structural-data requirement entirely. Sequence-based prediction works wherever sequencing works. Nigerian laboratories are increasingly adopting sequencing for surveillance, driven by AMR genomics programs. My prior work with GHRU-GSAR on AMR genomics and surveillance pipelines gives me direct experience with this infrastructure. TOPOLOGIX sits on top of that existing sequencing capacity and adds resistance prediction without new hardware requirements.

The measurable outcome is coverage. Structure-based tools cover 18 percent of mutations. TOPOLOGIX covers 100 percent. In a Nigerian clinical context, that means a resistance prediction is available for every sequenced sample, not just the subset with resolved structures. The pilot target is a 90 percent reduction in the proportion of sequenced samples that receive no resistance prediction.

ESSAY: INNOVATION AND SCALABILITY

The innovation in TOPOLOGIX is the use of protein language model embeddings for resistance prediction. ESM-2 delta-embeddings capture the mutational change in sequence representation without requiring structural information. This is a methodological departure from the field standard, which relies on structure-based features. My prior work demonstrated that structure-based topological features do not carry resistance signal. TOPOLOGIX is the positive result that followed.

The scalability argument is infrastructural. The model runs on commodity GPU hardware. The input is a sequence and a drug fingerprint. The output is a resistance prediction. There is no dependency on crystallography, homology modeling, or structural databases. Deployment requires only a sequencing facility and a computer. That is a lower infrastructure bar than any structure-based alternative.

The business model is a software-as-a-service platform for clinical laboratories and public health surveillance programs. Revenue comes from per-sample prediction fees and annual laboratory subscriptions. The target market is clinical genomics laboratories in Nigeria and other emerging markets, followed by public health agencies running AMR surveillance programs. The financial sustainability case rests on the growing adoption of sequencing in African clinical laboratories, driven by global AMR surveillance funding.

The scalability path is geographic. The model is sequence-based, so it transfers across pathogens and regions with fine-tuning on local data. The validation study in Nigeria will produce the first African clinical dataset for this approach, creating a data advantage for subsequent expansion across the continent. The infrastructure is already in place: I have built and operated four DuckDB-based ingest-to-analyze pipelines, self-host local LLM serving, and run production systems on Linux VPS with automated backup and disaster recovery. The technical foundation for a scalable deployment is proven.

CHECKLIST

- [ ] Verify current deadline and application portal details on the IFC Grants website
- [ ] Confirm eligibility for the Healthcare Access Initiative Grant 2026 as an independent researcher
- [ ] Obtain or draft a letter of support from a Nigerian clinical genomics laboratory for the validation study
- [ ] Prepare a detailed budget breakdown for the $350,000 request
- [ ] Compile the Platinum benchmark and SKEMPI 2.0 results as supplementary evidence
- [ ] Include preprints for the hERG cardiotoxicity replication and the interface-topology resistance study
- [ ] Draft a one-page project timeline for the 24-month implementation plan
- [ ] Confirm incorporation requirements and timeline for Nigeria or Germany
- [ ] Prepare a data-sharing and ethics statement for the Nigerian clinical isolate validation
- [ ] Identify and contact a potential clinical partner site in Nigeria before submission

EDITOR NOTES

- Eligibility risk: the applicant is pre-revenue and pre-incorporation; the grant may require a registered entity or institutional affiliation. Confirm before submission.
- The Nigerian clinical partner site is not named in the profile. This is a required gap to fill before submission; the application is materially weaker without a named partner.
- The budget figure of $350,000 is an estimate. Verify that it aligns with the grant's funding range and that the cost categories are realistic for Nigerian laboratory work.
- The essay responses assume the grant requires short-answer essays. If the application portal specifies different questions or word limits, the essays must be rewritten to match exactly.
- The profile lists multiple research lines. This application deliberately features TOPOLOGIX only, because it is the line whose sequence-based, infrastructure-light approach directly matches the grant's emerging-markets healthcare access mission. The CCT model, neurocascade, ergofluids, and psyche-twin are not mentioned in the application materials; do not add them.

MOTIVATION LETTER

The Healthcare Access Initiative Grant 2026 funds health technology that reaches underserved populations in emerging markets. I am applying with TOPOLOGIX, a sequence-based drug-resistance prediction platform that addresses a specific failure in precision medicine: structure-based tools like mCSM-lig cover only 18 percent of clinically relevant mutations because they require a resolved protein structure. In Nigeria, where genomic surveillance infrastructure is limited and structural data is scarcer than in high-income settings, that coverage gap is a barrier to treating drug-resistant infections and cancers.

TOPOLOGIX uses ESM-2 protein language model delta-embeddings combined with Morgan/ECFP drug fingerprints and a Random Forest classifier to predict drug-resistance mutations from sequence alone. On the Platinum benchmark of 553 mutations, it achieves an AUROC of 0.804 with a standard deviation of 0.025, and 0.634 on SKEMPI 2.0. It beats structure-based baselines while covering 100 percent of mutations. The method is fully sequence-based, which means it works wherever sequencing data exists, regardless of structural resolution. That is the scalability property the IFC mission requires: the tool does not depend on infrastructure that emerging-market laboratories do not have.

My path to this project is direct. I hold a B.Pharm from the University of Ibadan with a 2:1 Upper Division, am a PCN-licensed pharmacist, and am enrolled in the M.Sc. Digital Health program at the Hasso Plattner Institute and University of Potsdam. I have built and validated four independent DuckDB-based ingest-to-analyze pipelines across life sciences and other domains, and I self-host local LLM serving infrastructure. My pre-registered replication of a published topological approach to hERG cardiotoxicity found that topological features did not beat a plain descriptor baseline (AUROC 0.8426 vs 0.8782), and my interface-topology study for drug resistance found AUROC values of 0.425 and 0.485, ruling out interface geometry as the driver. Those results are why TOPOLOGIX exists. I pivoted to sequence representations because the data showed structure-based geometry was not the answer.

The grant would fund three things: validation of TOPOLOGIX against Nigerian clinical isolates, development of a laboratory-facing interface for resistance reporting, and a pilot deployment with a partner clinical site. The budget range of $200,000 to $4 million is appropriate for the infrastructure build and the validation study. I am pre-revenue and pre-incorporation, which I state plainly. The funds would support incorporation, compute infrastructure, and the validation pipeline that turns a research method into a deployed tool.

The measurable impact target is concrete: reduce the time from resistance mutation detection to treatment decision in Nigerian clinical settings, and provide resistance prediction coverage for the 82 percent of mutations that structure-based tools currently miss. That is the access improvement this programme exists to fund.

RESEARCH STATEMENT

TOPOLOGIX is a computational platform for predicting drug-resistance mutations from protein sequence alone. It was developed in response to a falsified hypothesis. Between 2024 and 2025, I tested whether bipartite persistent homology of protein-ligand interface geometry could predict hERG cardiotoxicity and drug resistance. The cardiotoxicity replication, pre-registered and powered, found topological features underperformed a plain descriptor baseline (AUROC 0.8426 vs 0.8782). The resistance study found AUROC values of 0.425 and 0.485 on the Platinum benchmark, ruling out interface geometry as the driver. Those results are published as preprints and are under peer review. They are the empirical basis for the current approach.

TOPOLOGIX replaces structure-based geometry with sequence-based representations. The architecture has three components. First, ESM-2 protein language model delta-embeddings capture the mutational change in the protein sequence representation. Second, Morgan/ECFP fingerprints encode the drug. Third, a Random Forest classifier learns the resistance decision boundary. On the Platinum benchmark of 553 mutations, the model achieves AUROC 0.804 plus or minus 0.025. On SKEMPI 2.0, it achieves 0.634. It outperforms mCSM-lig, a structure-based tool, which scores approximately 0.70 on the same tasks while covering only 18 percent of mutations. TOPOLOGIX covers 100 percent of mutations because it requires only sequence data.

The scientific contribution is the demonstration that sequence representation alone can carry resistance signal that interface geometry cannot. The translational contribution is coverage. In emerging markets, structural data is a luxury. Sequencing is increasingly available. A tool that works from sequence alone is deployable in settings where structure-based tools are not.

The validation plan has three gates. Gate one, already passed, is benchmark performance on Platinum and SKEMPI 2.0. Gate two, the focus of the proposed funding period, is validation against Nigerian clinical isolates. I will partner with a clinical genomics laboratory in Nigeria to obtain resistance mutation data from local pathogens and compare TOPOLOGIX predictions against phenotypic resistance assays. Gate three is deployment: a laboratory-facing interface that accepts a sequence and a drug, and returns a resistance prediction with confidence intervals.

The implementation plan spans 24 months. Months one through three: incorporation, compute infrastructure procurement, and partnership agreements with the Nigerian clinical site. Months four through twelve: isolate collection, sequencing, and model fine-tuning on local data. Months thirteen through eighteen: interface development and usability testing with laboratory staff. Months nineteen through twenty-four: pilot deployment and outcome measurement.

The measurable impact metric is coverage and accuracy. The target is to maintain AUROC above 0.75 on Nigerian clinical isolates while providing predictions for 100 percent of submitted sequences. The secondary metric is time-to-decision: reduce the interval from sample receipt to resistance report from weeks to days.

The budget request is $350,000. This covers incorporation and legal fees ($15,000), compute infrastructure including GPU servers and storage ($60,000), sequencing and laboratory costs for the validation study ($120,000), personnel for a research assistant and a software engineer ($100,000), and deployment and training costs ($55,000). The funding moves TOPOLOGIX from a validated research method to a deployed tool in a Nigerian clinical setting.

ESSAY: HEALTHCARE ACCESS CHALLENGE

The healthcare access challenge I address is the structural gap in drug-resistance prediction. Resistance testing in Nigeria relies on phenotypic culture methods that take weeks and require laboratory infrastructure that is concentrated in urban tertiary centers. Genotypic methods exist but are limited by reference databases built from non-African populations. The result is that treatment decisions for drug-resistant infections are often made without resistance data, leading to empirical therapy failure, prolonged illness, and increased mortality.

The scale of the problem is documented. Antimicrobial resistance is a leading cause of death in sub-Saharan Africa, and Nigeria carries a substantial burden. The tools to address it are misaligned with the context: they require structural data that does not exist for most clinically relevant mutations, and they are validated on populations that do not reflect Nigerian microbial diversity.

TOPOLOGIX addresses this by removing the structural-data requirement entirely. Sequence-based prediction works wherever sequencing works. Nigerian laboratories are increasingly adopting sequencing for surveillance, driven by AMR genomics programs. My prior work with GHRU-GSAR on AMR genomics and surveillance pipelines gives me direct experience with this infrastructure. TOPOLOGIX sits on top of that existing sequencing capacity and adds resistance prediction without new hardware requirements.

The measurable outcome is coverage. Structure-based tools cover 18 percent of mutations. TOPOLOGIX covers 100 percent. In a Nigerian clinical context, that means a resistance prediction is available for every sequenced sample, not just the subset with resolved structures. The pilot target is a 90 percent reduction in the proportion of sequenced samples that receive no resistance prediction.

ESSAY: INNOVATION AND SCALABILITY

The innovation in TOPOLOGIX is the use of protein language model embeddings for resistance prediction. ESM-2 delta-embeddings capture the mutational change in sequence representation without requiring structural information. This is a methodological departure from the field standard, which relies on structure-based features. My prior work demonstrated that structure-based topological features do not carry resistance signal. TOPOLOGIX is the positive result that followed.

The scalability argument is infrastructural. The model runs on commodity GPU hardware. The input is a sequence and a drug fingerprint. The output is a resistance prediction. There is no dependency on crystallography, homology modeling, or structural databases. Deployment requires only a sequencing facility and a computer. That is a lower infrastructure bar than any structure-based alternative.

The business model is a software-as-a-service platform for clinical laboratories and public health surveillance programs. Revenue comes from per-sample prediction fees and annual laboratory subscriptions. The target market is clinical genomics laboratories in Nigeria and other emerging markets, followed by public health agencies running AMR surveillance programs. The financial sustainability case rests on the growing adoption of sequencing in African clinical laboratories, driven by global AMR surveillance funding.

The scalability path is geographic. The model is sequence-based, so it transfers across pathogens and regions with fine-tuning on local data. The validation study in Nigeria will produce the first African clinical dataset for this approach, creating a data advantage for subsequent expansion across the continent. The infrastructure is already in place: I have built and operated four DuckDB-based ingest-to-analyze pipelines, self-host local LLM serving, and run production systems on Linux VPS with automated backup and disaster recovery. The technical foundation for a scalable deployment is proven.

CHECKLIST

- [ ] Verify current deadline and application portal details on the IFC Grants website
- [ ] Confirm eligibility for the Healthcare Access Initiative Grant 2026 as an independent researcher
- [ ] Obtain or draft a letter of support from a Nigerian clinical genomics laboratory for the validation study
- [ ] Prepare a detailed budget breakdown for the $350,000 request
- [ ] Compile the Platinum benchmark and SKEMPI 2.0 results as supplementary evidence
- [ ] Include preprints for the hERG cardiotoxicity replication and the interface-topology resistance study
- [ ] Draft a one-page project timeline for the 24-month implementation plan
- [ ] Confirm incorporation requirements and timeline for Nigeria or Germany
- [ ] Prepare a data-sharing and ethics statement for the Nigerian clinical isolate validation
- [ ] Identify and contact a potential clinical partner site in Nigeria before submission

EDITOR NOTES

- Eligibility risk: the applicant is pre-revenue and pre-incorporation; the grant may require a registered entity or institutional affiliation. Confirm before submission.
- The Nigerian clinical partner site is not named in the profile. This is a required gap to fill before submission; the application is materially weaker without a named partner.
- The budget figure of $350,000 is an estimate. Verify that it aligns with the grant's funding range and that the cost categories are realistic for Nigerian laboratory work.
- The essay responses assume the grant requires short-answer essays. If the application portal specifies different questions or word limits, the essays must be rewritten to match exactly.
- The profile lists multiple research lines. This application deliberately features TOPOLOGIX only, because it is the line whose sequence-based, infrastructure-light approach directly matches the grant's emerging-markets healthcare access mission. The CCT model, neurocascade, ergofluids, and psyche-twin are not mentioned in the application materials; do not add them.

AI Draft — Healthcare Access Initiative Grant 2026