MOTIVATION LETTER Drug resistance is the reason a child with sepsis in Lagos dies from a pathogen that a colleague in Berlin can still treat. The tools we use to predict resistance have been structurally limited: they require a resolved protein structure, which exists for only a fraction of clinically relevant mutations, and they fail to generalize across protein families. My research removes that limitation. TOPOLOGIX, the project I lead as an independent computational researcher, predicts drug-resistance mutations from amino acid sequence alone using ESM-2 protein language model embeddings and Morgan fingerprints in a Random Forest classifier. On the Platinum benchmark of 553 mutations, it achieves an AUROC of 0.804 plus or minus 0.025, and it covers 100 percent of mutations, where structure-based tools like mCSM-lig cover roughly 18 percent and score near 0.70. This is a change in what is possible. The L'Oréal-UNESCO Women in Science Programme, through AndPurpose, funds women researchers from Sub-Saharan Africa whose work advances life sciences and technology. My work sits precisely at that intersection. I am a Nigerian pharmacist, licensed by the Pharmacists Council of Nigeria, with a B.Pharm from the University of Ibadan. I am enrolled in the M.Sc. Digital Health program at the Hasso Plattner Institute and the University of Potsdam, beginning Winter Semester 2026/27. My research practice is built on pre-registration, Bayesian calibration, and open reporting. When my cardiotoxicity topology study found that persistent homology features did not beat a plain descriptor baseline (AUROC 0.8426 versus 0.8782), I reported the negative result directly. That finding redirected my work toward sequence-based methods, which became TOPOLOGIX. When my ergofluids project failed its first real-data validation gate, I reported the failure rather than reframing it. This is how science should work, and it is the standard I hold myself to. TOPOLOGIX addresses a global health problem with direct relevance to Africa. Antimicrobial and antimalarial resistance are urgent threats across the continent, yet most predictive tools are built and validated on datasets dominated by non-African pathogens. A sequence-only method removes the structural bottleneck and makes resistance prediction feasible for organisms where no crystal structure exists. The method is also fast and lightweight, which matters for research environments with limited compute. I have published three sole-authored preprints on my addiction neuroscience work, with a co-authored paper under review at Alcohol (Elsevier), and I maintain an active GitHub and ORCID record. I am applying to this programme because it recognizes that rigorous, independent researchers from Sub-Saharan Africa can lead scientific innovation. I am one of them. RESEARCH STATEMENT TOPOLOGIX: Sequence-Based Prediction of Drug-Resistance Mutations Using Protein Language Model Embeddings The problem. Drug resistance emerges through mutations that alter how proteins interact with therapeutic compounds. Predicting which mutations confer resistance is a core task in drug development, clinical genomics, and surveillance. The dominant approach, structure-based prediction, requires a resolved three-dimensional protein structure. This is a severe constraint. On the Platinum benchmark, structure-based tools such as mCSM-lig cover only about 18 percent of mutations because the remaining 82 percent lack a usable structure. This means the majority of clinically relevant mutations are simply never scored. The field has accepted this limitation for years. The approach. TOPOLOGIX eliminates the structure requirement entirely. The method uses ESM-2 protein language model delta-embeddings, which capture the functional effect of a mutation by comparing the embedding of the wild-type sequence to the embedding of the mutant sequence. These delta-embeddings are concatenated with Morgan/ECFP drug fingerprints, which encode the chemical structure of the ligand. A Random Forest classifier then learns the mapping from this combined representation to a resistance label. The entire pipeline runs on sequence and SMILES strings. No structure is needed. Results. On the Platinum benchmark of 553 mutations, TOPOLOGIX achieves an AUROC of 0.804 with a standard deviation of 0.025. On SKEMPI 2.0, it achieves 0.634. It outperforms structure-based baselines, including mCSM-lig at approximately 0.70, while covering 100 percent of mutations. The performance gap is largest precisely where structure-based methods fail: mutations in proteins without resolved structures. This is the population of mutations that matters most for real-world surveillance, because we rarely have structures for the pathogens that are actually evolving resistance. Why this project fits the programme. The L'Oréal-UNESCO Women in Science Programme funds research in life sciences and technology with clear societal relevance. Drug resistance is one of the most pressing global health threats, and it is disproportionately severe in Sub-Saharan Africa. A method that predicts resistance from sequence alone is directly applicable to African pathogens, many of which lack structural data. The project is also methodologically innovative: it applies protein language models, a recent advance in machine learning, to a problem that has been dominated by physics-based and structure-based approaches. This is a technology project with a life-sciences mission. My role and track record. I am the sole researcher on TOPOLOGIX. I designed the architecture, wrote the code, ran the benchmarks, and analyzed the results. This is consistent with my broader research practice. I have built the CCT model, a tripartite pharmacological framework for reward-memory encoding in addiction, with all five pre-registered hypotheses confirmed and posterior super-additivity of 13 to 22 percentage points across model versions. I have built neurocascade, a receptor-to-behavior brain-circuit simulation engine with 62 of 62 tests passing. I have conducted a pre-registered, powered replication in cardiotoxicity topology that settled a comparison the literature had never actually run, and I reported the negative result honestly. My work is characterized by pre-registration, Bayesian calibration, and open reporting. I do not hide failures, and I do not overclaim successes. Next steps. The immediate next step for TOPOLOGIX is to expand validation beyond Platinum and SKEMPI 2.0 to include resistance datasets from African pathogens, particularly Mycobacterium tuberculosis and Plasmodium falciparum. I will also test whether the method generalizes to resistance prediction for antiviral drugs, where structural data is often sparse. The M.Sc. Digital Health program at Hasso Plattner Institute will provide access to computational resources and collaborators. This fellowship would support the compute, dataset acquisition, and travel needed to move TOPOLOGIX from a benchmark result to a deployed tool for resistance surveillance. ESSAY: RESEARCH IMPACT AND INNOVATION The impact of TOPOLOGIX is best understood by what it changes in practice. Today, a clinical genomics lab that identifies a novel mutation in a drug target faces a choice: either a structure exists and a structure-based predictor can score it, or no structure exists and the mutation is classified as a variant of unknown significance. The second case is the majority. TOPOLOGIX changes that by scoring every mutation, regardless of structural coverage. This is not a theoretical advantage. On the Platinum benchmark, structure-based tools cover 18 percent of mutations. TOPOLOGIX covers 100 percent. For a surveillance program tracking resistance emergence in a pathogen with no crystal structure, this is the difference between having a prediction and having nothing. The innovation is methodological. Protein language models such as ESM-2 learn representations of amino acid sequences from large-scale unsupervised training. The delta-embedding, the difference between the wild-type and mutant representations, encodes the functional effect of the mutation in a way that is transferable across protein families. This is why TOPOLOGIX generalizes beyond the training distribution. It does not memorize structures; it learns the language of protein function. Combining this with drug fingerprints creates a joint representation of the mutation and the drug, which is necessary because resistance is always a mutation-drug pair property, not a mutation property alone. The innovation is also practical. The method runs on a standard laptop. It does not require a GPU cluster, a cryo-EM facility, or a crystallography lab. This matters for researchers in Nigeria and across Sub-Saharan Africa, where computational infrastructure is limited but the need for resistance prediction is acute. A method that democratizes access to resistance prediction is a method that empowers African researchers to study African pathogens with tools that were previously available only to well-funded laboratories in Europe and North America. I measure impact by adoption and by clinical relevance. The benchmark results are the first step. The second step is validation on African pathogen datasets, which I will pursue during my M.Sc. at Hasso Plattner Institute. The third step is deployment as an open-source tool, which is consistent with my practice of publishing preprints and code openly. I am applying to the L'Oréal-UNESCO Women in Science Programme because it funds exactly this kind of work: rigorous, innovative, and aimed at problems that matter for the Global South. EDITOR NOTES - Framing choice: TOPOLOGIX is the selected research line because it is the only project in the profile that directly matches the programme's life sciences and technology focus, has a clear global health relevance (drug resistance), and has quantitative benchmark results that demonstrate innovation and impact. The CCT model, neurocascade, ergofluids, and psyche-twin are either in different domains (addiction neuroscience, brain simulation, fluid dynamics, knowledge graphs) or at earlier validation stages. The cardiotoxicity topology study is a negative result and is used only as evidence of rigorous reporting, not as current work. - Eligibility risk: The programme targets women researchers from Sub-Saharan Africa affiliated with a recognized institution in the region. Eniola is Nigerian (eligible by nationality) but is enrolled in a German M.Sc. program and employed by Synthcare (location unspecified). The application must clarify her institutional affiliation at the time of application. If the programme requires current affiliation in Sub-Saharan Africa, this is a potential disqualifier. Verify before submission. - Eligibility risk: The programme requires doctoral candidates in at least their 2nd year or post-doctoral applicants who completed their thesis after February 2021. Eniola is enrolled in an M.Sc. program and has no PhD. She may not meet the degree-level requirement. This must be verified against the full programme guidelines. If she is ineligible, do not submit. - Fact verification: The profile lists "National Product Manager, Synthcare (Mar 2026-present)" and "Clinical Pharmacist, Ramset Pharmacy (Jan-Mar 2026)". These dates are in the future relative to the current date. Confirm the actual employment timeline and adjust the application accordingly. Also confirm the exact title of the M.Sc. program (Digital Health) and the start date (Winter Semester 2026/27). - Gap: The motivation letter and research statement do not include a personal narrative about why drug resistance matters to Eniola specifically, beyond the general framing. The applicant should insert one concrete personal or professional experience (e.g., a clinical case from her pharmacy work, a moment in her bioinformatics research at GHRU-GSAR) that connects her to the problem. This is required for a competitive application. - Gap: The application does not name a specific mentor or host institution for the fellowship period. The programme may require a letter of support or a host confirmation. Eniola should identify a faculty member at Hasso Plattner Institute or a collaborator (e.g., Kent Berridge, Samuel Gershman, Nathaniel Daw, Marcelo Mattar) who can serve as a reference or host. - Compliance: The profile lists "Application materials" as the required submission. The programme URL is https://andpurpose.world/loreal-unesco-women-in-science-programme-2026/. The applicant must check the programme website for the actual application form, document upload requirements, and any additional essays or letters of recommendation. The deadline is listed as "see programme website" and must be confirmed. CHECKLIST - [ ] Verify eligibility: confirm the programme accepts M.Sc. students or requires PhD candidacy (2nd year or later) or post-doctoral status - [ ] Verify institutional affiliation requirement: confirm whether current affiliation must be in Sub-Saharan Africa or if Nigerian nationality suffices - [ ] Confirm the application deadline from the programme website - [ ] Confirm the exact list of required documents from the programme website - [ ] Update employment timeline: confirm dates for Synthcare (National Product Manager) and Ramset Pharmacy (Clinical Pharmacist) - [ ] Confirm M.Sc. enrollment details: Hasso Plattner Institute, University of Potsdam, Winter Semester 2026/27, program name - [ ] Insert one personal narrative paragraph connecting Eniola's clinical or research experience to the drug-resistance problem - [ ] Identify and contact one referee or host institution for a letter of support - [ ] Prepare TOPOLOGIX benchmark results as a supplementary document or figure (AUROC 0.804 on Platinum, 0.634 on SKEMPI 2.0, 100% coverage vs 18% for structure-based tools) - [ ] Prepare a list of publications and preprints, including the three sole-authored preprints (OSF/Zenodo) and the co-authored paper under review at Alcohol (Elsevier) - [ ] Prepare ORCID (0009-0001-9272-6735) and GitHub (github.com/AmunRaPtah) links for the application form - [ ] Confirm the programme does not require a prior L'Oréal-UNESCO Women in Science award (Eniola has not received one, which satisfies this condition) - [ ] Draft and attach the motivation letter (300-500 words) as written above - [ ] Draft and attach the research statement (400-600 words) as written above - [ ] Draft and attach the essay response (200-350 words) as written above - [ ] Review all documents for compliance with the programme's formatting and submission requirements