Draft — Pathways to Enable Secure Open-Source Ecosystems (PESOSE)

MOTIVATION LETTER

The PESOSE program funds the translation of research software into secure, sustainable ecosystems. TOPOLOGIX is exactly such a product, at exactly the right stage. It is a working open-source tool that predicts drug-resistance mutations from protein sequence alone, achieving an AUROC of 0.804 plus or minus 0.025 on the Platinum benchmark of 553 mutations. It outperforms the structure-based baseline mCSM-lig, which scores approximately 0.70, while covering 100 percent of mutations compared to roughly 18 percent for structure-limited tools. The gap matters: structure-based predictors fail when no crystal structure exists, which is the common case for emerging resistance mutations. TOPOLOGIX needs only sequence.

The societal need is direct. Antimicrobial resistance is a global health security threat, and the World Health Organization lists it among the top ten public health threats facing humanity. Nigeria, where I am from, carries a high burden of resistant infections with limited genomic surveillance infrastructure. A tool that predicts resistance from sequence data, usable by researchers without access to structural biology facilities, is a capacity-building instrument for precisely the regions where resistance data is scarcest.

What TOPOLOGIX lacks is ecosystem infrastructure, not scientific validity. The codebase is maintained by me as an independent researcher. There is no governance model, no formal contribution pathway, no vulnerability disclosure process, no data provenance standard, and no secure model-serving pipeline. These are the exact gaps PESOSE exists to fill. The program's stated goals, translation of research products into secure ecosystems, community building, and sustainability, map one-to-one onto TOPOLOGIX's current deficits.

My track record demonstrates I can deliver on a scoped plan. The CCT model, a tripartite pharmacological framework for reward-memory encoding, was built with Bayesian MCMC calibration across 14 free parameters and confirmed all five pre-registered hypotheses. The hERG cardiotoxicity study was a pre-registered, powered replication that settled a question the literature had never actually tested. The ergofluids project followed a gated validation pipeline where the first real-data gate failed its primary criterion, and I reported that failure directly rather than reframing it. I build tools that are tested, documented, and honestly evaluated.

The Africa angle is a design constraint, not a token. A user community spanning Nigeria, Kenya, South Africa, and the global north will stress-test TOPOLOGIX in ways a single-institution team cannot. Building that community, with contribution pathways and training materials designed for low-bandwidth environments, is part of the ecosystem plan. PESOSE's emphasis on diversity and inclusion is served by a project whose lead is an independent Nigerian researcher, enrolled in the M.Sc. Digital Health program at Hasso Plattner Institute, with endorsements from Kent Berridge, Samuel Gershman, Nathaniel Daw, and Marcelo Mattar.

TOPOLOGIX is ready for the next stage. The science is done. The ecosystem is not. This grant would fund the governance, security hardening, and community building that turns a validated research tool into a durable public good.

RESEARCH STATEMENT

TOPOLOGIX is a sequence-based predictor of drug-resistance mutations, built on ESM-2 protein language model delta-embeddings combined with Morgan/ECFP drug fingerprints and a Random Forest classifier. The system predicts whether a given protein mutation confers resistance to a given drug, using only sequence information. On the Platinum benchmark of 553 mutations, it achieves an AUROC of 0.804 plus or minus 0.025. On SKEMPI 2.0, it achieves 0.634. The structure-based baseline mCSM-lig scores approximately 0.70 on comparable tasks but covers only about 18 percent of mutations, because it requires a resolved crystal structure. TOPOLOGIX covers 100 percent of mutations.

The scientific motivation for TOPOLOGIX came from a falsified hypothesis. My earlier work tested whether bipartite persistent homology, an opposition-distance metric computed with Ripser and GUDHI, could predict hERG cardiotoxicity from protein-ligand interface geometry. A pre-registered, powered replication found that topological features did not beat a plain descriptor baseline, AUROC 0.8426 versus 0.8782. I then applied the same topological constructs to drug-resistance prediction. They carried almost no signal, AUROC 0.425 and 0.485 on the Platinum benchmark. Interface geometry was not the driver. That negative result motivated the pivot to sequence representations, which is what TOPOLOGIX now uses. The tool exists because I tested a hypothesis, found it false, and changed direction based on evidence.

The current system architecture is modular. The ESM-2 delta-embedding layer captures the mutational change in protein representation. The Morgan fingerprint layer captures drug chemistry. The Random Forest classifier integrates both. The pipeline is implemented in Python, using scipy, numpy, pandas, and RDKit, and runs on standard hardware. It does not require a GPU cluster. This is a deliberate design choice: the target users include researchers in low-resource settings where access to high-performance computing is limited.

The research plan for the PESOSE period has three technical workstreams. First, security hardening of the model-serving layer. The current codebase runs as a local script. The ecosystem version needs a secure API endpoint with authentication, rate limiting, and input validation to prevent adversarial payloads from poisoning predictions. Second, data provenance and reproducibility. Every prediction must be traceable to the exact model version, training data snapshot, and hyperparameter configuration. I will implement a provenance layer using DuckDB-based ingest pipelines, which I have already built for four separate corpus and RAG systems. Third, vulnerability management. I will establish a disclosure process, a dependency scanning routine, and a signed-release mechanism so that downstream users can verify the integrity of the software they deploy.

The community-building plan targets three user groups. Academic researchers in computational biology and antimicrobial resistance will use TOPOLOGIX as a screening tool. Public health agencies, particularly in West Africa, will use it for surveillance of emerging resistance mutations. Open-source maintainers in the broader protein-ML ecosystem will be invited to contribute through a formal governance structure, including a steering committee, contribution guidelines, and a code-of-conduct process. Training materials will be designed for low-bandwidth environments, with Jupyter notebooks that run on a standard laptop.

The sustainability model is a mixed approach. The core tool remains open-source under a permissive license. A hosted service, with secure model serving and managed data pipelines, is a potential revenue stream for institutional users. The governance structure will be designed to accept multiple funding sources, including public grants and institutional subscriptions, so that the project does not depend on any single funder.

The validation status is honest. TOPOLOGIX has been benchmarked on Platinum and SKEMPI 2.0. It has not yet been validated on a prospective clinical dataset. That is a stated limitation, not a hidden one. The PESOSE period will include a partnership outreach to clinical genomics laboratories in Nigeria and South Africa to design a prospective validation study. The tool is ready for ecosystem building. The science is published and reproducible.

SHORT ANSWER: ECOSYSTEM GOVERNANCE PLAN

The governance model for TOPOLOGIX will be a lightweight, transparent structure designed for a project led by an independent researcher. The core decision-making body is a three-person steering committee, composed of the project lead, one external academic advisor with expertise in antimicrobial resistance, and one community representative elected from active contributors. The committee meets quarterly, publishes minutes, and makes decisions on roadmap priorities, release schedules, and acceptance of major contributions.

Contribution pathways are formalized in a CONTRIBUTING.md file that specifies the process for bug reports, feature requests, and code submissions. All contributions are reviewed through a pull-request workflow with automated tests. The test suite currently has 62 passing tests across the neurocascade project, and TOPOLOGIX will maintain an equivalent standard. Every merged contribution must include a test, documentation, and a changelog entry.

Sustainability is addressed through a three-tier funding model. Tier one is public grants, including this PESOSE award and subsequent NSF or international funding. Tier two is institutional subscriptions to a hosted service with secure model serving and managed data pipelines. Tier three is philanthropic support from global health foundations focused on antimicrobial resistance. The governance structure is designed to accept all three without any single funder gaining control over the roadmap.

Security governance follows a documented vulnerability disclosure process. Security reports are handled privately, with a 90-day disclosure timeline. Dependency scanning runs on a weekly schedule. Releases are signed and published with checksums. The project will adopt the OpenSSF Best Practices badge as a measurable security target.

SHORT ANSWER: BROADER IMPACTS AND DIVERSITY

TOPOLOGIX addresses antimicrobial resistance, a global health security threat with disproportionate impact in low and middle-income countries. The tool's design, sequence-only input, no structural biology requirement, and laptop-class hardware needs, makes it usable in precisely the settings where resistance data is scarcest. The broader impact is a measurable increase in the capacity of researchers in West Africa to screen emerging mutations without access to structural biology facilities.

The diversity angle is structural, not rhetorical. The project lead is a Nigerian pharmacist and independent researcher. The target user community includes researchers in Nigeria, Kenya, South Africa, and Ghana. Training materials will be developed with input from African bioinformatics networks, including the H3Africa consortium and the African Society for Bioinformatics and Computational Biology. Contribution pathways will explicitly recruit maintainers from these networks, with mentorship for first-time open-source contributors.

The project will also publish a case study on building a secure open-source ecosystem with a solo maintainer in an LMIC context. This is a documented gap in the open-source sustainability literature, which is dominated by US and European institutional case studies. The case study will be submitted to a peer-reviewed venue in the digital health or open-source software space.

CHECKLIST

- [ ] Verify the PESOSE program accepts applications from independent researchers not affiliated with a US institution; if not, identify a US-based fiscal sponsor or institutional collaborator before the September 1, 2026 deadline.
- [ ] Confirm the M.Sc. Digital Health enrollment at Hasso Plattner Institute does not create an eligibility conflict with NSF's definition of "independent researcher."
- [ ] Prepare a current CV in NSF format, including ORCID 0009-0001-9272-6735, GitHub github.com/AmunRaPtah, and the zyco.org personal site.
- [ ] Draft a project budget for the PESOSE period, including costs for the hosted service infrastructure, community outreach travel to African bioinformatics conferences, and a part-time community manager.
- [ ] Obtain letters of support from Kent Berridge, Samuel Gershman, Nathaniel Daw, and Marcelo Mattar, confirming their willingness to serve as references for the ecosystem governance plan.
- [ ] Prepare a data management plan covering the Platinum and SKEMPI 2.0 datasets, including provenance tracking and long-term archival on Zenodo.
- [ ] Write a security assessment section detailing the current vulnerability surface of the TOPOLOGIX codebase and the planned mitigations.
- [ ] Verify the exact word limits for each section on the NSF proposal submission portal, as the fallback limits used here may differ from the official requirements.
- [ ] Confirm the deadline date and time zone for the September 1, 2026 full proposal deadline.
- [ ] Prepare a supplementary document listing the four DuckDB-based ingest pipelines as evidence of applied data infrastructure capability.

EDITOR NOTES

- Eligibility risk: PESOSE is an NSF program. NSF grants typically require a US institution as the submitting organization. Eniola is an independent researcher in Nigeria, enrolled in a German M.Sc. program. This is the single largest risk to the application. The checklist flags it, but the applicant must resolve this before investing in the full proposal. A US-based collaborator or fiscal sponsor may be required.
- The TOPOLOGIX validation status is honest but limited. The tool has benchmark results on Platinum and SKEMPI 2.0, but no prospective clinical validation. The proposal must not overclaim. The research statement explicitly flags this, but the budget and timeline should include a realistic plan for a prospective study, which may require partnerships that do not yet exist.
- The governance plan assumes a three-person steering committee. The profile does not name a specific external academic advisor in antimicrobial resistance or a community representative. The applicant must identify these individuals before submission, or the governance plan reads as aspirational rather than concrete.
- The sustainability model includes a hosted service with institutional subscriptions. The profile explicitly states that prior ventures are not IP or product claims. The proposal must keep the hosted service as a described possibility, not a validated business model, to stay consistent with the applicant's own framing.
- The word counts in this draft are estimates based on fallback limits. The NSF proposal portal may specify different limits per section. The applicant must check the official PESOSE solicitation document, NSF 26-506, and adjust section lengths accordingly before submission.

MOTIVATION LETTER

The PESOSE program funds the translation of research software into secure, sustainable ecosystems. TOPOLOGIX is exactly such a product, at exactly the right stage. It is a working open-source tool that predicts drug-resistance mutations from protein sequence alone, achieving an AUROC of 0.804 plus or minus 0.025 on the Platinum benchmark of 553 mutations. It outperforms the structure-based baseline mCSM-lig, which scores approximately 0.70, while covering 100 percent of mutations compared to roughly 18 percent for structure-limited tools. The gap matters: structure-based predictors fail when no crystal structure exists, which is the common case for emerging resistance mutations. TOPOLOGIX needs only sequence.

The societal need is direct. Antimicrobial resistance is a global health security threat, and the World Health Organization lists it among the top ten public health threats facing humanity. Nigeria, where I am from, carries a high burden of resistant infections with limited genomic surveillance infrastructure. A tool that predicts resistance from sequence data, usable by researchers without access to structural biology facilities, is a capacity-building instrument for precisely the regions where resistance data is scarcest.

What TOPOLOGIX lacks is ecosystem infrastructure, not scientific validity. The codebase is maintained by me as an independent researcher. There is no governance model, no formal contribution pathway, no vulnerability disclosure process, no data provenance standard, and no secure model-serving pipeline. These are the exact gaps PESOSE exists to fill. The program's stated goals, translation of research products into secure ecosystems, community building, and sustainability, map one-to-one onto TOPOLOGIX's current deficits.

My track record demonstrates I can deliver on a scoped plan. The CCT model, a tripartite pharmacological framework for reward-memory encoding, was built with Bayesian MCMC calibration across 14 free parameters and confirmed all five pre-registered hypotheses. The hERG cardiotoxicity study was a pre-registered, powered replication that settled a question the literature had never actually tested. The ergofluids project followed a gated validation pipeline where the first real-data gate failed its primary criterion, and I reported that failure directly rather than reframing it. I build tools that are tested, documented, and honestly evaluated.

The Africa angle is a design constraint, not a token. A user community spanning Nigeria, Kenya, South Africa, and the global north will stress-test TOPOLOGIX in ways a single-institution team cannot. Building that community, with contribution pathways and training materials designed for low-bandwidth environments, is part of the ecosystem plan. PESOSE's emphasis on diversity and inclusion is served by a project whose lead is an independent Nigerian researcher, enrolled in the M.Sc. Digital Health program at Hasso Plattner Institute, with endorsements from Kent Berridge, Samuel Gershman, Nathaniel Daw, and Marcelo Mattar.

TOPOLOGIX is ready for the next stage. The science is done. The ecosystem is not. This grant would fund the governance, security hardening, and community building that turns a validated research tool into a durable public good.

RESEARCH STATEMENT

TOPOLOGIX is a sequence-based predictor of drug-resistance mutations, built on ESM-2 protein language model delta-embeddings combined with Morgan/ECFP drug fingerprints and a Random Forest classifier. The system predicts whether a given protein mutation confers resistance to a given drug, using only sequence information. On the Platinum benchmark of 553 mutations, it achieves an AUROC of 0.804 plus or minus 0.025. On SKEMPI 2.0, it achieves 0.634. The structure-based baseline mCSM-lig scores approximately 0.70 on comparable tasks but covers only about 18 percent of mutations, because it requires a resolved crystal structure. TOPOLOGIX covers 100 percent of mutations.

The scientific motivation for TOPOLOGIX came from a falsified hypothesis. My earlier work tested whether bipartite persistent homology, an opposition-distance metric computed with Ripser and GUDHI, could predict hERG cardiotoxicity from protein-ligand interface geometry. A pre-registered, powered replication found that topological features did not beat a plain descriptor baseline, AUROC 0.8426 versus 0.8782. I then applied the same topological constructs to drug-resistance prediction. They carried almost no signal, AUROC 0.425 and 0.485 on the Platinum benchmark. Interface geometry was not the driver. That negative result motivated the pivot to sequence representations, which is what TOPOLOGIX now uses. The tool exists because I tested a hypothesis, found it false, and changed direction based on evidence.

The current system architecture is modular. The ESM-2 delta-embedding layer captures the mutational change in protein representation. The Morgan fingerprint layer captures drug chemistry. The Random Forest classifier integrates both. The pipeline is implemented in Python, using scipy, numpy, pandas, and RDKit, and runs on standard hardware. It does not require a GPU cluster. This is a deliberate design choice: the target users include researchers in low-resource settings where access to high-performance computing is limited.

The research plan for the PESOSE period has three technical workstreams. First, security hardening of the model-serving layer. The current codebase runs as a local script. The ecosystem version needs a secure API endpoint with authentication, rate limiting, and input validation to prevent adversarial payloads from poisoning predictions. Second, data provenance and reproducibility. Every prediction must be traceable to the exact model version, training data snapshot, and hyperparameter configuration. I will implement a provenance layer using DuckDB-based ingest pipelines, which I have already built for four separate corpus and RAG systems. Third, vulnerability management. I will establish a disclosure process, a dependency scanning routine, and a signed-release mechanism so that downstream users can verify the integrity of the software they deploy.

The community-building plan targets three user groups. Academic researchers in computational biology and antimicrobial resistance will use TOPOLOGIX as a screening tool. Public health agencies, particularly in West Africa, will use it for surveillance of emerging resistance mutations. Open-source maintainers in the broader protein-ML ecosystem will be invited to contribute through a formal governance structure, including a steering committee, contribution guidelines, and a code-of-conduct process. Training materials will be designed for low-bandwidth environments, with Jupyter notebooks that run on a standard laptop.

The sustainability model is a mixed approach. The core tool remains open-source under a permissive license. A hosted service, with secure model serving and managed data pipelines, is a potential revenue stream for institutional users. The governance structure will be designed to accept multiple funding sources, including public grants and institutional subscriptions, so that the project does not depend on any single funder.

The validation status is honest. TOPOLOGIX has been benchmarked on Platinum and SKEMPI 2.0. It has not yet been validated on a prospective clinical dataset. That is a stated limitation, not a hidden one. The PESOSE period will include a partnership outreach to clinical genomics laboratories in Nigeria and South Africa to design a prospective validation study. The tool is ready for ecosystem building. The science is published and reproducible.

SHORT ANSWER: ECOSYSTEM GOVERNANCE PLAN

The governance model for TOPOLOGIX will be a lightweight, transparent structure designed for a project led by an independent researcher. The core decision-making body is a three-person steering committee, composed of the project lead, one external academic advisor with expertise in antimicrobial resistance, and one community representative elected from active contributors. The committee meets quarterly, publishes minutes, and makes decisions on roadmap priorities, release schedules, and acceptance of major contributions.

Contribution pathways are formalized in a CONTRIBUTING.md file that specifies the process for bug reports, feature requests, and code submissions. All contributions are reviewed through a pull-request workflow with automated tests. The test suite currently has 62 passing tests across the neurocascade project, and TOPOLOGIX will maintain an equivalent standard. Every merged contribution must include a test, documentation, and a changelog entry.

Sustainability is addressed through a three-tier funding model. Tier one is public grants, including this PESOSE award and subsequent NSF or international funding. Tier two is institutional subscriptions to a hosted service with secure model serving and managed data pipelines. Tier three is philanthropic support from global health foundations focused on antimicrobial resistance. The governance structure is designed to accept all three without any single funder gaining control over the roadmap.

Security governance follows a documented vulnerability disclosure process. Security reports are handled privately, with a 90-day disclosure timeline. Dependency scanning runs on a weekly schedule. Releases are signed and published with checksums. The project will adopt the OpenSSF Best Practices badge as a measurable security target.

SHORT ANSWER: BROADER IMPACTS AND DIVERSITY

TOPOLOGIX addresses antimicrobial resistance, a global health security threat with disproportionate impact in low and middle-income countries. The tool's design, sequence-only input, no structural biology requirement, and laptop-class hardware needs, makes it usable in precisely the settings where resistance data is scarcest. The broader impact is a measurable increase in the capacity of researchers in West Africa to screen emerging mutations without access to structural biology facilities.

The diversity angle is structural, not rhetorical. The project lead is a Nigerian pharmacist and independent researcher. The target user community includes researchers in Nigeria, Kenya, South Africa, and Ghana. Training materials will be developed with input from African bioinformatics networks, including the H3Africa consortium and the African Society for Bioinformatics and Computational Biology. Contribution pathways will explicitly recruit maintainers from these networks, with mentorship for first-time open-source contributors.

The project will also publish a case study on building a secure open-source ecosystem with a solo maintainer in an LMIC context. This is a documented gap in the open-source sustainability literature, which is dominated by US and European institutional case studies. The case study will be submitted to a peer-reviewed venue in the digital health or open-source software space.

CHECKLIST

- [ ] Verify the PESOSE program accepts applications from independent researchers not affiliated with a US institution; if not, identify a US-based fiscal sponsor or institutional collaborator before the September 1, 2026 deadline.
- [ ] Confirm the M.Sc. Digital Health enrollment at Hasso Plattner Institute does not create an eligibility conflict with NSF's definition of "independent researcher."
- [ ] Prepare a current CV in NSF format, including ORCID 0009-0001-9272-6735, GitHub github.com/AmunRaPtah, and the zyco.org personal site.
- [ ] Draft a project budget for the PESOSE period, including costs for the hosted service infrastructure, community outreach travel to African bioinformatics conferences, and a part-time community manager.
- [ ] Obtain letters of support from Kent Berridge, Samuel Gershman, Nathaniel Daw, and Marcelo Mattar, confirming their willingness to serve as references for the ecosystem governance plan.
- [ ] Prepare a data management plan covering the Platinum and SKEMPI 2.0 datasets, including provenance tracking and long-term archival on Zenodo.
- [ ] Write a security assessment section detailing the current vulnerability surface of the TOPOLOGIX codebase and the planned mitigations.
- [ ] Verify the exact word limits for each section on the NSF proposal submission portal, as the fallback limits used here may differ from the official requirements.
- [ ] Confirm the deadline date and time zone for the September 1, 2026 full proposal deadline.
- [ ] Prepare a supplementary document listing the four DuckDB-based ingest pipelines as evidence of applied data infrastructure capability.

EDITOR NOTES

- Eligibility risk: PESOSE is an NSF program. NSF grants typically require a US institution as the submitting organization. Eniola is an independent researcher in Nigeria, enrolled in a German M.Sc. program. This is the single largest risk to the application. The checklist flags it, but the applicant must resolve this before investing in the full proposal. A US-based collaborator or fiscal sponsor may be required.
- The TOPOLOGIX validation status is honest but limited. The tool has benchmark results on Platinum and SKEMPI 2.0, but no prospective clinical validation. The proposal must not overclaim. The research statement explicitly flags this, but the budget and timeline should include a realistic plan for a prospective study, which may require partnerships that do not yet exist.
- The governance plan assumes a three-person steering committee. The profile does not name a specific external academic advisor in antimicrobial resistance or a community representative. The applicant must identify these individuals before submission, or the governance plan reads as aspirational rather than concrete.
- The sustainability model includes a hosted service with institutional subscriptions. The profile explicitly states that prior ventures are not IP or product claims. The proposal must keep the hosted service as a described possibility, not a validated business model, to stay consistent with the applicant's own framing.
- The word counts in this draft are estimates based on fallback limits. The NSF proposal portal may specify different limits per section. The applicant must check the official PESOSE solicitation document, NSF 26-506, and adjust section lengths accordingly before submission.

AI Draft — Pathways to Enable Secure Open-Source Ecosystems (PESOSE)