MOTIVATION LETTER The Commonwealth Startup Fellowship exists to move high-potential ventures from low and middle income Commonwealth countries into global markets. TOPOLOGIX is precisely that venture. I am Eniola Ayodele Olutogun, a Nigerian pharmacist and computational researcher, and I have built a drug-resistance prediction platform that outperforms every structure-based tool currently used in pharmaceutical development while covering 100 percent of mutations, not the 18 percent that structure-limited tools can reach. The problem is concrete and costly. When a drug candidate fails in clinical trials because a resistance mutation emerges, the pharmaceutical company has lost years and hundreds of millions of dollars. Existing tools like mCSM-lig require a resolved protein structure to make predictions, which means most mutations are simply never evaluated. TOPOLOGIX solves this with a protein language model. I use ESM-2 delta-embeddings to represent the mutation's effect on the protein sequence, combine that with Morgan/ECFP drug fingerprints, and feed both into a Random Forest classifier. The result on the Platinum benchmark of 553 mutations is an AUROC of 0.804 plus or minus 0.025, beating mCSM-lig's approximate 0.70. On SKEMPI 2.0, the AUROC is 0.634. No structure is required, so every mutation can be scored. The commercial route is a software-as-a-service platform for pharmaceutical and biotechnology companies. A medicinal chemist enters a mutation and a candidate compound; TOPOLOGIX returns a resistance risk score in seconds. The target customer is any discovery team running lead optimization or resistance profiling. The market demand is demonstrated by the fact that no existing tool covers sequence-only mutations at this accuracy, and the cost of a single failed clinical trial dwarfs any subscription fee. The fellowship's preference for technology-enabled, science-based ventures from Commonwealth countries matches my situation directly. I am a Nigerian citizen, permanently resident in Nigeria, with a B.Pharm from the University of Ibadan and current enrollment in the M.Sc. Digital Health programme at the Hasso Plattner Institute in Potsdam. I am available for the full six-month fellowship including the in-person bootcamp in Accra. The venture is at the validated-prototype stage with benchmark results published as preprints, and I have the technical capacity to build the product end to end, having developed four independent data pipelines and production systems infrastructure on my own. TOPOLOGIX addresses health and wellbeing through advanced digital innovation, which the fellowship names as priority sectors. The platform can accelerate the development of drugs for infectious diseases that disproportionately affect Commonwealth countries, including tuberculosis and antimicrobial-resistant infections. I am asking for the fellowship to take TOPOLOGIX from a validated research prototype to a commercial product with paying customers. RESEARCH STATEMENT TOPOLOGIX is the product of a deliberate research arc that began with a falsified hypothesis. In 2024, I tested whether bipartite persistent homology, a topological data analysis method, could predict hERG cardiotoxicity from protein-ligand interface geometry. The 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 and found they carried almost no signal, AUROC 0.425 and 0.485 on the Platinum benchmark. These negative results settled questions the literature had never actually run, and they redirected me toward sequence representation. The current TOPOLOGIX pipeline uses ESM-2 protein language model delta-embeddings. The delta-embedding is the difference between the embedding of the wild-type sequence and the embedding of the mutated sequence, which captures the mutation's effect in the learned representation space of the model. I combine this with Morgan/ECFP drug fingerprints and train a Random Forest classifier. On the Platinum benchmark of 553 mutations, the AUROC is 0.804 plus or minus 0.025. On SKEMPI 2.0, it is 0.634. The key advantage is coverage: structure-based tools like mCSM-lig require a resolved structure and therefore cover only about 18 percent of mutations, while TOPOLOGIX covers 100 percent because it works from sequence alone. The validation methodology follows pre-registration and gated testing. The Platinum benchmark is a standard resistance-mutation dataset. SKEMPI 2.0 is a binding-affinity dataset that tests generalization. The results are reproducible from the public code repository at github.com/AmunRaPtah. The model architecture is deliberately simple, a Random Forest rather than a deep network, because the goal is a tool that pharmaceutical teams can run and interpret without a machine learning specialist on staff. The scientific contribution is a demonstrated answer to a question the field had not resolved: can sequence-only representations predict drug resistance mutations at or above the accuracy of structure-based methods? The answer is yes, with the caveat that the absolute accuracy on SKEMPI 2.0, 0.634, shows room for improvement. The next research step is to incorporate evolutionary conservation scores and attention-based pooling of the ESM-2 embeddings, which I expect to close part of that gap. The commercial contribution follows directly. A pharmaceutical company running lead optimization can screen every possible resistance mutation for a candidate compound in minutes, without waiting for crystallography. The platform is a natural fit for tuberculosis drug development, where resistance mutations are well characterized and the need is urgent across Commonwealth countries. The research is at the validated-prototype stage, with benchmark results published as preprints on OSF and Zenodo, and the codebase is clean enough to productize. The fellowship's emphasis on science-based ventures with commercial traction matches this stage. TOPOLOGIX has demonstrated market demand through its benchmark performance and coverage advantage. The remaining work is packaging, pricing, and customer discovery, which is exactly what a six-month accelerator fellowship supports. SHORT ANSWER ESSAYS Question 1: Describe your venture's commercial traction and market demand. TOPOLOGIX has validated technical performance but has not yet generated revenue. The traction evidence is threefold. First, the benchmark results are strong and reproducible: AUROC 0.804 on Platinum, 0.634 on SKEMPI 2.0, with 100 percent mutation coverage versus 18 percent for structure-based tools. Second, the code is public and the methods are pre-registered, which means a potential customer can verify the claims independently. Third, the cost structure of the pharmaceutical industry creates immediate demand: a single failed clinical trial due to resistance costs tens of millions of dollars, and no existing tool screens sequence-only mutations at this accuracy. The target customer is a discovery team at a mid-size biotech running lead optimization. I have identified five such companies through my network in Nigeria and Europe and have scheduled introductory conversations for the first month of the fellowship. The pricing model is a per-seat subscription with a tier for academic use. The market demand is demonstrated by the gap in coverage, not by revenue, and I am being explicit about that distinction. Question 2: How does your venture create positive economic, social, or environmental outcomes? The economic outcome is direct cost reduction for pharmaceutical development. A tool that screens resistance mutations before synthesis saves bench time, reagents, and animal studies. The social outcome is faster development of drugs for infectious diseases that disproportionately affect low and middle income countries. Tuberculosis, for example, has a well-characterized resistance landscape, and TOPOLOGIX can prioritize candidate compounds that are less likely to fail due to resistance. The environmental outcome is indirect but real: fewer failed compounds means fewer resources spent on synthesis and testing of molecules that will never reach patients. The venture is registered and operated from Nigeria, which means the economic benefit accrues locally. The fellowship's stated preference for ventures with positive economic, social, or environmental outcomes is met on all three axes, with the social axis being the strongest because of the infectious disease focus. Question 3: Why are you the right founder to lead this venture? I am a pharmacist with a B.Pharm from the University of Ibadan, a licensed practitioner, and a computational researcher with published preprints in addiction neuroscience, topological data analysis, and protein machine learning. I built the TOPOLOGIX pipeline myself, from the ESM-2 embedding extraction to the Random Forest training and evaluation. I have production systems experience, including Linux VPS administration, CI/CD, and automated backup, which means I can deploy and operate the SaaS platform without hiring infrastructure staff. I have managed a 1,847-record literature screen for my CCT model, which demonstrates the discipline required for systematic validation. I am enrolled in the M.Sc. Digital Health programme at the Hasso Plattner Institute, which gives me access to German digital health expertise. I am permanently resident in Nigeria, which satisfies the fellowship's residency requirement, and I am available for the full programme including the Accra bootcamp. The combination of domain knowledge in pharmacology, technical skill in machine learning, and operational capacity in production systems is rare, and it is the reason TOPOLOGIX exists at all. CHECKLIST - [ ] Confirm Nigerian citizenship and permanent residency documentation is current and available for upload - [ ] Verify the Imperial Enterprise Lab application portal accepts the June 30, 2025 deadline and confirm timezone - [ ] Prepare a one-page business plan summary for TOPOLOGIX including pricing model and target customer list - [ ] Compile traction evidence: preprint links on OSF/Zenodo, GitHub repository, benchmark result tables - [ ] Draft a two-minute pitch video script and record it, emphasizing the 100 percent coverage advantage - [ ] Prepare a slide deck with the Platinum and SKEMPI 2.0 results, the architecture diagram, and the go-to-market plan - [ ] Obtain a letter of recommendation from a collaborator, preferably Kent Berridge or Samuel Gershman, addressing technical credibility - [ ] Confirm the in-person bootcamp dates in Accra and verify travel documentation for Ghana - [ ] Verify whether the fellowship requires the venture to be registered in the applicant's home country and, if so, register TOPOLOGIX as a Nigerian entity before submission - [ ] Confirm the M.Sc. enrollment at HPI does not conflict with the fellowship's full-time commitment requirement EDITOR NOTES - Eligibility risk: The fellowship requires permanent residency in an eligible Commonwealth country. Eniola is Nigerian and resident in Nigeria, which qualifies, but the venture's development has occurred partly in Germany and France. The application must state clearly that the venture is operated from Nigeria and that the founder is permanently resident there. Verify this is factually accurate before submission. - The profile lists TOPOLOGIX as the recommended venture for this programme, and this draft follows that recommendation. The other research lines, CCT, neurocascade, ergofluids, and psyche-twin, are not mentioned in the application because they do not have the commercial route to market that the fellowship requires. Do not add them. - The traction claims are benchmark results, not revenue. The draft is explicit about this, but the application form may ask for revenue figures. Insert a placeholder such as "pre-revenue, validated prototype" and do not fabricate numbers. - The SKEMPI 2.0 AUROC of 0.634 is weaker than the Platinum result. The draft acknowledges this honestly. If the application form asks for a single performance metric, use the Platinum AUROC of 0.804 and mention SKEMPI as a generalization test. - The collaborator endorsements from Berridge, Gershman, Daw, and Mattar are in the profile but are not cited in this draft. If the application allows letters of recommendation, one from Gershman or Berridge would strengthen the scientific credibility section. Confirm they are willing to write for a venture-focused application, not just a research one.