Draft — Grand Challenges Africa

MOTIVATION LETTER

Antimicrobial resistance is a present, measurable crisis in Nigeria. At the GHRU-GSAR bioinformatics unit, I built AMR surveillance pipelines that track resistant bacterial lineages from clinical samples. Those pipelines tell us which resistance genes are spreading, but they cannot tell us which new mutations will defeat the drugs we are about to deploy. That predictive gap is what TOPOLOGIX closes.

TOPOLOGIX predicts drug-resistance mutations from protein sequence alone. It uses ESM-2 protein language model delta-embeddings combined with Morgan/ECFP drug fingerprints, classified by a Random Forest model. On the Platinum benchmark of 553 mutations, it achieves an AUROC of 0.804 plus or minus 0.025. On SKEMPI 2.0, it reaches 0.634. It covers 100 percent of mutations, where structure-based tools like mCSM-lig cover only about 18 percent because they require a resolved crystal structure. Most resistance-relevant proteins in African clinical isolates have no such structure. Sequence-only prediction is the only option that scales.

Grand Challenges Africa funds Africa-led scientific innovations with a path to real-world impact. TOPOLOGIX is exactly that. It is a deployable software product, not a paper. It can be integrated into existing AMR surveillance pipelines, including the ones I helped build at GHRU-GSAR. It can be run on a laptop, requires no wet-lab infrastructure, and produces results in minutes. The target users are public health agencies, clinical microbiology labs, and drug development programs across the continent.

The science is rigorous and the negative results are reported honestly. My earlier work on interface topology for resistance prediction found that bipartite persistent homology carries almost no signal for this task, with AUROC values of 0.425 and 0.485 on the Platinum benchmark. That falsified result is published in my preprints and motivated the pivot to sequence representations. TOPOLOGIX is the product of that pivot, and it beats structure-based baselines by a wide margin.

I am a Nigerian pharmacist and computational researcher, currently enrolled in the M.Sc. Digital Health program at the Hasso Plattner Institute in Potsdam. I hold a B.Pharm from the University of Ibadan and am licensed by the Pharmacists Council of Nigeria. My collaborators include Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU. I am applying as an independent researcher with a strong Africa-based track record.

TOPOLOGIX is pre-revenue and pre-incorporation. What it has is a validated model, a clear deployment pathway, and a health crisis that demands a software answer. I am seeking a seed grant to build the deployment layer, validate on Nigerian clinical isolate data, and pilot with a public health partner. Grand Challenges Africa is the right home for that work.

RESEARCH STATEMENT

TOPOLOGIX is a sequence-based machine learning system that predicts drug-resistance mutations in proteins. The problem it solves is structural: existing resistance prediction tools require a resolved three-dimensional protein structure to compute features. In clinical settings across Africa, that structure is almost never available. The result is that the tools cover less than one fifth of the mutations that matter. TOPOLOGIX removes the structure requirement entirely.

The method has three components. First, ESM-2, a protein language model, generates delta-embeddings that capture the mutational change in the protein's learned representation space. Second, the drug is encoded as a Morgan/ECFP fingerprint, capturing its chemical identity. Third, a Random Forest classifier combines these inputs to predict whether a given mutation confers resistance to a given drug. The model was trained and evaluated on the Platinum benchmark of 553 mutations, achieving an AUROC of 0.804 plus or minus 0.025. On SKEMPI 2.0, a binding-affinity benchmark, it achieves 0.634. The structure-based baseline mCSM-lig scores approximately 0.70 on comparable tasks but covers only 18 percent of mutations. TOPOLOGIX covers all of them.

The development path was not linear. My initial hypothesis was that interface topology, specifically bipartite persistent homology using an opposition-distance metric, would predict resistance from protein-ligand interface geometry. I pre-registered that study and ran a powered replication. The result was negative: topological features scored an AUROC of 0.425 and 0.485 on the Platinum benchmark, below the plain descriptor baseline. I reported that result directly in my preprints rather than reframing it. That falsification is what redirected the project toward sequence representations, and TOPOLOGIX is the outcome.

The deployment pathway has three stages. Stage one is integration into existing AMR surveillance pipelines. I built such pipelines at GHRU-GSAR, which sequenced and analyzed resistant bacterial genomes across Nigeria. TOPOLOGIX can be added as a prediction layer that flags emerging resistance mutations before they appear in clinical isolates. Stage two is validation on Nigerian clinical data, which requires a partnership with a sequencing lab and ethics approval. Stage three is a pilot with a public health agency or clinical network, delivering predictions through a simple web interface.

The technical infrastructure is already in place. The model runs in Python using scipy, numpy, and scikit-learn. The pipeline is containerized and can run on a standard laptop or a cloud instance. I have built four independent DuckDB-based data pipelines across life-sciences and other domains, and I self-host local LLM serving with llama.cpp for documentation and query interfaces. The software engineering is not a risk; the data partnerships are.

The grant request is for seed funding to complete stage one and begin stage two. Specific deliverables are: a packaged TOPOLOGIX module with a documented API, a validation report on Nigerian clinical isolate sequences, and a partnership agreement with at least one public health or clinical laboratory. The budget is modest because the compute is modest. The impact is not modest: a tool that predicts resistance from sequence alone, usable in any lab in Africa that can sequence a genome.

ESSAY: AFRICA-LED INNOVATION AND LOCAL IMPACT

Grand Challenges Africa requires that the Principal Investigator be African and that the innovation be developed in Africa. I am Nigerian, born and educated in Ibadan, and licensed as a pharmacist by the Pharmacists Council of Nigeria. My research career began at the University of Ibadan and continued at the GHRU-GSAR bioinformatics unit, where I worked on AMR genomics and surveillance pipelines for Nigerian bacterial isolates. The problem TOPOLOGIX addresses, the gap between genomic surveillance data and actionable resistance prediction, is one I observed directly in that work. The solution is being built by me, as an independent researcher, with the African deployment context as the design constraint from day one.

The local impact is concrete. Nigerian clinical microbiology labs increasingly have sequencing capacity but lack the computational tools to turn sequence data into resistance predictions. TOPOLOGIX fills that gap with a tool that requires no structural biology expertise and no high-performance computing. A lab that can produce a FASTA file can get a resistance prediction. That is the difference between surveillance that documents a crisis and surveillance that anticipates it.

The scale pathway is continental. AMR surveillance networks exist across Africa, coordinated through the African Union and the WHO Regional Office for Africa. TOPOLOGIX is designed to plug into those networks as a software layer. The model is drug-agnostic and protein-agnostic; it can be retrained on new resistance data as it emerges. The barrier to scale is not technical; it is the cost of packaging, validation, and partnership development. That is what this grant would fund.

ESSAY: SCIENCE-BASED SOLUTION WITH A PATHWAY TO DEPLOYMENT

The selection criteria for Grand Challenges Africa emphasize science-based solutions with real-world application, not academic novelty alone. TOPOLOGIX meets the science bar through its benchmark results and its honest reporting of negative findings. The Platinum benchmark AUROC of 0.804 plus or minus 0.025 is a strong, reproducible result. The SKEMPI 2.0 score of 0.634 is lower, and I report it as such. The falsified topology hypothesis is documented in my preprints. This is a research program that reports what it finds.

The deployment pathway is the differentiator. Most resistance prediction tools are published as papers and never shipped. TOPOLOGIX is being built as a product. The codebase is modular, the API is documented, and the compute requirements are minimal. The first deployment target is the Nigerian AMR surveillance ecosystem, where I have existing relationships from my GHRU-GSAR work. The second target is the broader African network of sequencing labs. The third is integration into drug development pipelines that need to screen candidate compounds against known and predicted resistance mutations.

The sustainability model is a subscription or service fee paid by public health agencies and clinical networks, with a tiered pricing structure that keeps the cost low for Nigerian public-sector users. The software is open-source under a permissive license, with commercial support and customization as the revenue stream. This is a standard open-core model, and it is appropriate for a tool that addresses a public health crisis.

CHECKLIST

- [ ] Confirm current Grand Challenges Africa application form and submission portal access
- [ ] Verify seed grant amount and any co-funding or matching requirements
- [ ] Prepare CV in the required format, including ORCID 0009-0001-9272-6735 and GitHub github.com/AmunRaPtah
- [ ] Obtain or confirm letters of support from Kent Berridge, Samuel Gershman, Nathaniel Daw, or Marcelo Mattar
- [ ] Prepare a one-page budget table for seed funding, itemizing compute, data partnerships, and software packaging costs
- [ ] Prepare a two-page project timeline covering stage one integration, stage two validation, and stage three pilot
- [ ] Confirm eligibility for independent researcher status without institutional affiliation, or identify a Nigerian host institution if required
- [ ] Prepare a data-sharing and ethics plan for validation on Nigerian clinical isolate sequences
- [ ] Confirm the current status of the Platinum benchmark and SKEMPI 2.0 results, and update if newer versions exist
- [ ] Prepare a short technical appendix describing the ESM-2 delta-embedding method, Morgan fingerprint parameters, and Random Forest hyperparameters
- [ ] Verify the exact word limits for each essay and letter on the application form and trim to fit
- [ ] Confirm the deadline and submission method, and submit at least five business days before the deadline

EDITOR NOTES

- Eligibility risk: The programme may require affiliation with an African institution or organization. The applicant is independent. Confirm whether independent researchers are accepted, or identify a Nigerian host institution such as the University of Ibadan or a public health agency before submission.
- The TOPOLOGIX results on Platinum and SKEMPI 2.0 are from the profile and should be verified against the current codebase and any updated benchmark versions before submission. The AUROC values must match the latest run.
- The applicant must insert specific details not in the profile: the name of the Nigerian sequencing lab or public health partner for stage two validation, the exact budget figures for the seed grant request, and the names and affiliations of any additional collaborators or endorsers beyond the four listed.
- The motivation letter references GHRU-GSAR experience. Confirm the exact dates and scope of that role, and whether any confidentiality or data-sharing agreements restrict the use of surveillance pipeline details in a public grant application.
- The essays describe a subscription or service fee model. This is a claim about future business structure. Confirm that the applicant is comfortable stating this in a grant application, and soften the language if the venture is pre-incorporation and pre-revenue, as the profile states.

MOTIVATION LETTER

Antimicrobial resistance is a present, measurable crisis in Nigeria. At the GHRU-GSAR bioinformatics unit, I built AMR surveillance pipelines that track resistant bacterial lineages from clinical samples. Those pipelines tell us which resistance genes are spreading, but they cannot tell us which new mutations will defeat the drugs we are about to deploy. That predictive gap is what TOPOLOGIX closes.

TOPOLOGIX predicts drug-resistance mutations from protein sequence alone. It uses ESM-2 protein language model delta-embeddings combined with Morgan/ECFP drug fingerprints, classified by a Random Forest model. On the Platinum benchmark of 553 mutations, it achieves an AUROC of 0.804 plus or minus 0.025. On SKEMPI 2.0, it reaches 0.634. It covers 100 percent of mutations, where structure-based tools like mCSM-lig cover only about 18 percent because they require a resolved crystal structure. Most resistance-relevant proteins in African clinical isolates have no such structure. Sequence-only prediction is the only option that scales.

Grand Challenges Africa funds Africa-led scientific innovations with a path to real-world impact. TOPOLOGIX is exactly that. It is a deployable software product, not a paper. It can be integrated into existing AMR surveillance pipelines, including the ones I helped build at GHRU-GSAR. It can be run on a laptop, requires no wet-lab infrastructure, and produces results in minutes. The target users are public health agencies, clinical microbiology labs, and drug development programs across the continent.

The science is rigorous and the negative results are reported honestly. My earlier work on interface topology for resistance prediction found that bipartite persistent homology carries almost no signal for this task, with AUROC values of 0.425 and 0.485 on the Platinum benchmark. That falsified result is published in my preprints and motivated the pivot to sequence representations. TOPOLOGIX is the product of that pivot, and it beats structure-based baselines by a wide margin.

I am a Nigerian pharmacist and computational researcher, currently enrolled in the M.Sc. Digital Health program at the Hasso Plattner Institute in Potsdam. I hold a B.Pharm from the University of Ibadan and am licensed by the Pharmacists Council of Nigeria. My collaborators include Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU. I am applying as an independent researcher with a strong Africa-based track record.

TOPOLOGIX is pre-revenue and pre-incorporation. What it has is a validated model, a clear deployment pathway, and a health crisis that demands a software answer. I am seeking a seed grant to build the deployment layer, validate on Nigerian clinical isolate data, and pilot with a public health partner. Grand Challenges Africa is the right home for that work.

RESEARCH STATEMENT

TOPOLOGIX is a sequence-based machine learning system that predicts drug-resistance mutations in proteins. The problem it solves is structural: existing resistance prediction tools require a resolved three-dimensional protein structure to compute features. In clinical settings across Africa, that structure is almost never available. The result is that the tools cover less than one fifth of the mutations that matter. TOPOLOGIX removes the structure requirement entirely.

The method has three components. First, ESM-2, a protein language model, generates delta-embeddings that capture the mutational change in the protein's learned representation space. Second, the drug is encoded as a Morgan/ECFP fingerprint, capturing its chemical identity. Third, a Random Forest classifier combines these inputs to predict whether a given mutation confers resistance to a given drug. The model was trained and evaluated on the Platinum benchmark of 553 mutations, achieving an AUROC of 0.804 plus or minus 0.025. On SKEMPI 2.0, a binding-affinity benchmark, it achieves 0.634. The structure-based baseline mCSM-lig scores approximately 0.70 on comparable tasks but covers only 18 percent of mutations. TOPOLOGIX covers all of them.

The development path was not linear. My initial hypothesis was that interface topology, specifically bipartite persistent homology using an opposition-distance metric, would predict resistance from protein-ligand interface geometry. I pre-registered that study and ran a powered replication. The result was negative: topological features scored an AUROC of 0.425 and 0.485 on the Platinum benchmark, below the plain descriptor baseline. I reported that result directly in my preprints rather than reframing it. That falsification is what redirected the project toward sequence representations, and TOPOLOGIX is the outcome.

The deployment pathway has three stages. Stage one is integration into existing AMR surveillance pipelines. I built such pipelines at GHRU-GSAR, which sequenced and analyzed resistant bacterial genomes across Nigeria. TOPOLOGIX can be added as a prediction layer that flags emerging resistance mutations before they appear in clinical isolates. Stage two is validation on Nigerian clinical data, which requires a partnership with a sequencing lab and ethics approval. Stage three is a pilot with a public health agency or clinical network, delivering predictions through a simple web interface.

The technical infrastructure is already in place. The model runs in Python using scipy, numpy, and scikit-learn. The pipeline is containerized and can run on a standard laptop or a cloud instance. I have built four independent DuckDB-based data pipelines across life-sciences and other domains, and I self-host local LLM serving with llama.cpp for documentation and query interfaces. The software engineering is not a risk; the data partnerships are.

The grant request is for seed funding to complete stage one and begin stage two. Specific deliverables are: a packaged TOPOLOGIX module with a documented API, a validation report on Nigerian clinical isolate sequences, and a partnership agreement with at least one public health or clinical laboratory. The budget is modest because the compute is modest. The impact is not modest: a tool that predicts resistance from sequence alone, usable in any lab in Africa that can sequence a genome.

ESSAY: AFRICA-LED INNOVATION AND LOCAL IMPACT

Grand Challenges Africa requires that the Principal Investigator be African and that the innovation be developed in Africa. I am Nigerian, born and educated in Ibadan, and licensed as a pharmacist by the Pharmacists Council of Nigeria. My research career began at the University of Ibadan and continued at the GHRU-GSAR bioinformatics unit, where I worked on AMR genomics and surveillance pipelines for Nigerian bacterial isolates. The problem TOPOLOGIX addresses, the gap between genomic surveillance data and actionable resistance prediction, is one I observed directly in that work. The solution is being built by me, as an independent researcher, with the African deployment context as the design constraint from day one.

The local impact is concrete. Nigerian clinical microbiology labs increasingly have sequencing capacity but lack the computational tools to turn sequence data into resistance predictions. TOPOLOGIX fills that gap with a tool that requires no structural biology expertise and no high-performance computing. A lab that can produce a FASTA file can get a resistance prediction. That is the difference between surveillance that documents a crisis and surveillance that anticipates it.

The scale pathway is continental. AMR surveillance networks exist across Africa, coordinated through the African Union and the WHO Regional Office for Africa. TOPOLOGIX is designed to plug into those networks as a software layer. The model is drug-agnostic and protein-agnostic; it can be retrained on new resistance data as it emerges. The barrier to scale is not technical; it is the cost of packaging, validation, and partnership development. That is what this grant would fund.

ESSAY: SCIENCE-BASED SOLUTION WITH A PATHWAY TO DEPLOYMENT

The selection criteria for Grand Challenges Africa emphasize science-based solutions with real-world application, not academic novelty alone. TOPOLOGIX meets the science bar through its benchmark results and its honest reporting of negative findings. The Platinum benchmark AUROC of 0.804 plus or minus 0.025 is a strong, reproducible result. The SKEMPI 2.0 score of 0.634 is lower, and I report it as such. The falsified topology hypothesis is documented in my preprints. This is a research program that reports what it finds.

The deployment pathway is the differentiator. Most resistance prediction tools are published as papers and never shipped. TOPOLOGIX is being built as a product. The codebase is modular, the API is documented, and the compute requirements are minimal. The first deployment target is the Nigerian AMR surveillance ecosystem, where I have existing relationships from my GHRU-GSAR work. The second target is the broader African network of sequencing labs. The third is integration into drug development pipelines that need to screen candidate compounds against known and predicted resistance mutations.

The sustainability model is a subscription or service fee paid by public health agencies and clinical networks, with a tiered pricing structure that keeps the cost low for Nigerian public-sector users. The software is open-source under a permissive license, with commercial support and customization as the revenue stream. This is a standard open-core model, and it is appropriate for a tool that addresses a public health crisis.

CHECKLIST

- [ ] Confirm current Grand Challenges Africa application form and submission portal access
- [ ] Verify seed grant amount and any co-funding or matching requirements
- [ ] Prepare CV in the required format, including ORCID 0009-0001-9272-6735 and GitHub github.com/AmunRaPtah
- [ ] Obtain or confirm letters of support from Kent Berridge, Samuel Gershman, Nathaniel Daw, or Marcelo Mattar
- [ ] Prepare a one-page budget table for seed funding, itemizing compute, data partnerships, and software packaging costs
- [ ] Prepare a two-page project timeline covering stage one integration, stage two validation, and stage three pilot
- [ ] Confirm eligibility for independent researcher status without institutional affiliation, or identify a Nigerian host institution if required
- [ ] Prepare a data-sharing and ethics plan for validation on Nigerian clinical isolate sequences
- [ ] Confirm the current status of the Platinum benchmark and SKEMPI 2.0 results, and update if newer versions exist
- [ ] Prepare a short technical appendix describing the ESM-2 delta-embedding method, Morgan fingerprint parameters, and Random Forest hyperparameters
- [ ] Verify the exact word limits for each essay and letter on the application form and trim to fit
- [ ] Confirm the deadline and submission method, and submit at least five business days before the deadline

EDITOR NOTES

- Eligibility risk: The programme may require affiliation with an African institution or organization. The applicant is independent. Confirm whether independent researchers are accepted, or identify a Nigerian host institution such as the University of Ibadan or a public health agency before submission.
- The TOPOLOGIX results on Platinum and SKEMPI 2.0 are from the profile and should be verified against the current codebase and any updated benchmark versions before submission. The AUROC values must match the latest run.
- The applicant must insert specific details not in the profile: the name of the Nigerian sequencing lab or public health partner for stage two validation, the exact budget figures for the seed grant request, and the names and affiliations of any additional collaborators or endorsers beyond the four listed.
- The motivation letter references GHRU-GSAR experience. Confirm the exact dates and scope of that role, and whether any confidentiality or data-sharing agreements restrict the use of surveillance pipeline details in a public grant application.
- The essays describe a subscription or service fee model. This is a claim about future business structure. Confirm that the applicant is comfortable stating this in a grant application, and soften the language if the venture is pre-incorporation and pre-revenue, as the profile states.

AI Draft — Grand Challenges Africa