MOTIVATION LETTER Antimicrobial resistance is a present, measurable crisis in Sub-Saharan Africa, not a future threat. The African Union estimates that drug-resistant infections already claim over 250,000 lives annually on the continent, and that figure is projected to rise sharply without new surveillance tools. My research addresses this problem directly. I am Eniola Ayodele Olutogun, a computational researcher and licensed pharmacist from Nigeria, and I am applying to the Mbeya University of Science and Technology research programme because my current work in protein language models and machine learning can strengthen drug resistance surveillance capacity in Tanzania and across the region. My primary research line, TOPOLOGIX, predicts drug-resistance mutations from protein sequence alone. The system uses ESM-2 protein language model delta-embeddings combined with Morgan/ECFP drug fingerprints and a Random Forest classifier. On the Platinum benchmark of 553 mutations, TOPOLOGIX achieves an AUROC of 0.804 with a standard deviation of 0.025. It covers 100 percent of mutations in the benchmark, whereas structure-based tools like mCSM-lig cover only about 18 percent because they require resolved three-dimensional protein structures. This coverage gap matters in Africa, where structural data for locally circulating pathogens is scarce. Sequence data, by contrast, is increasingly available through genomic surveillance networks. The regional relevance is concrete. TOPOLOGIX can be applied to resistance mutations in Mycobacterium tuberculosis, Salmonella Typhi, and Neisseria gonorrhoeae, all of which are priority pathogens in Tanzanian clinical settings. The method does not require high-performance computing infrastructure beyond what a standard research server provides. The model pipeline is built in Python using scikit-learn and PyTorch, and I have deployed similar pipelines on Linux VPS environments with automated backup and CI/CD. This means the tool can be transferred to MUST collaborators and run locally, not just in a paper. My qualifications include a B.Pharm from the University of Ibadan with a German equivalent grade of 1.9, a PCN pharmacy license, and enrollment in the M.Sc. Digital Health programme at the Hasso Plattner Institute and University of Potsdam starting Winter Semester 2026/27. I have co-authored a paper under review at Alcohol (Elsevier) and hold three sole-authored preprints under review at peer-reviewed journals. My computational training spans Python, PyMC for Bayesian calibration, RDKit for cheminformatics, and GROMACS for molecular dynamics. The Mbeya University of Science and Technology programme emphasizes research in life sciences and technology with relevance to regional challenges. TOPOLOGIX fits that mandate precisely. It is a life-science tool built on modern machine learning methods, and it targets a documented regional health threat. I am also committed to mentoring: I have built and published open-source research infrastructure, including the neurocascade simulation engine with 62 passing tests, and I regularly document my methods for reproducibility. I am applying as an independent researcher with a track record of completing pre-registered, falsifiable studies. One of my prior projects, a topological analysis of hERG cardiotoxicity, found that persistent homology features did not beat a plain descriptor baseline (AUROC 0.8426 versus 0.8782). I reported that result directly rather than reframing it. That commitment to honest negative results is the standard I will bring to MUST. RESEARCH STATEMENT TOPOLOGIX: Sequence-Based Prediction of Drug Resistance Mutations for Sub-Saharan African Pathogen Surveillance Problem and Regional Context Drug resistance undermines treatment for tuberculosis, typhoid, gonorrhea, and malaria across Sub-Saharan Africa. Standard resistance prediction tools rely on protein structures, but structural coverage for clinically relevant pathogen proteins is sparse. The Platinum benchmark, a widely used resistance mutation dataset, includes 553 mutations, yet structure-based predictors can only score a fraction of them. This leaves clinicians and surveillance programs without predictions for most observed mutations. A sequence-only method removes that bottleneck. Methods TOPOLOGIX uses a three-component pipeline. First, ESM-2, a protein language model trained on 65 million protein sequences, generates delta-embeddings that capture the mutational change in the protein sequence. Second, Morgan/ECFP fingerprints encode the drug molecule. Third, a Random Forest classifier combines both representations to predict whether a given mutation confers resistance to the given drug. The model was trained and evaluated on the Platinum benchmark with cross-validation, achieving an AUROC of 0.804 plus or minus 0.025 across 553 mutations. On SKEMPI 2.0, a binding affinity benchmark, the same architecture achieved 0.634. These results beat structure-based baselines including mCSM-lig at approximately 0.70 AUROC, while covering the full mutation set. Why This Method Matters for Tanzania The method requires only sequence data, which is increasingly generated by genomic surveillance initiatives in the region. It does not require crystallography, cryo-EM, or homology modeling. The computational cost is modest: ESM-2 embeddings can be generated on a single GPU or even CPU for small proteins, and the Random Forest training runs in minutes on a standard workstation. This makes the tool deployable at Mbeya University of Science and Technology without specialized hardware. Validation and Reproducibility I follow pre-registration and open-science practices. The TOPOLOGIX protocol was pre-registered before benchmarking. All code is available on my GitHub repository, and the data pipeline uses DuckDB for reproducible ingest-to-analysis workflows. I have also published negative results from my prior topological studies, demonstrating that I do not hide falsified hypotheses. For TOPOLOGIX, the next validation step is external testing on resistance mutations from African clinical isolates, which I propose to conduct in collaboration with MUST researchers. Proposed Work at MUST I propose a three-phase project. Phase one: retrain and fine-tune TOPOLOGIX on resistance mutation data for Mycobacterium tuberculosis and Salmonella Typhi, using public genomic datasets from African surveillance programs. Phase two: validate predictions against phenotypic resistance data from Tanzanian clinical isolates, if available through MUST partners. Phase three: package the model as a lightweight web service or command-line tool for use by local clinicians and researchers, with documentation in English and Swahili summaries. Feasibility and Timeline The core model is already built and benchmarked. Phase one requires approximately three months of data collection and fine-tuning. Phase two depends on data access and can run concurrently. Phase three requires two months of software engineering. Total timeline is eight to ten months. I have the necessary skills: Python, PyMC, RDKit, scikit-learn, and production systems deployment. I am currently enrolled in the M.Sc. Digital Health programme at Hasso Plattner Institute, which provides additional methodological grounding in health data science. Expected Outcomes The primary outcome is a validated, sequence-only resistance prediction tool tailored to Sub-Saharan African pathogens. Secondary outcomes include a published benchmark comparing TOPOLOGIX to structure-based methods on African isolate data, and a deployable software package. This work directly supports the MUST programme's focus on life sciences and technology applied to regional health challenges. EDITOR NOTES - Research line selected: TOPOLOGIX. This is the only active research line that directly matches the programme's stated focus on life sciences and technology with regional health relevance. The CCT model, neurocascade, and ergofluids are either too basic-science oriented or still behind validation gates. The hERG topology study is a completed negative result and is referenced only as evidence of methodological honesty, not as current work. - Eligibility risk: The selection criteria state the applicant must be a woman researcher from or working in Sub-Saharan Africa. The profile lists Eniola as Nigerian, which satisfies the geographic requirement, but the gender criterion must be verified. If the programme requires formal proof of gender or if the applicant does not identify as a woman, this application should not proceed. - Degree status risk: The criteria mention PhD or postdoctoral level. Eniola holds a B.Pharm and is enrolled in an M.Sc. programme starting Winter 2026/27. She is not yet a PhD candidate. The application should clarify her early-career status and argue that the M.Sc. enrollment plus independent research record qualifies her under the programme's early-career provisions, if any. If the programme strictly requires PhD enrollment, this application will be rejected. - Verification needed: The claim that TOPOLOGIX achieves AUROC 0.804 on Platinum and 0.634 on SKEMPI 2.0 must be backed by the preprints on OSF/Zenodo. The applicant should attach the relevant preprint or provide a direct link in the application. The Alcohol (Elsevier) co-authored paper under review should also be listed in the CV with a note on its status. - Personal detail gap: The motivation letter does not include any specific connection to Tanzania or Mbeya. The applicant should insert a sentence about any prior collaboration, travel, or interest in Tanzanian health systems, or a named contact at MUST if one exists. Without this, the regional relevance claim remains generic. - Funding amount and deadline: The programme page lists no amount or deadline. The applicant must confirm these details directly with the MUST research office before submitting, and adjust the proposal budget and timeline accordingly. CHECKLIST - [ ] Confirm eligibility: woman researcher from Sub-Saharan Africa, early-career status accepted - [ ] Verify programme deadline and funding amount from MUST research office - [ ] Attach TOPOLOGIX preprint from OSF or Zenodo as evidence of results - [ ] Attach CV with full publication list, including the Alcohol (Elsevier) paper under review - [ ] Attach academic transcripts for B.Pharm from University of Ibadan - [ ] Attach proof of enrollment or admission letter for M.Sc. Digital Health at HPI/Potsdam - [ ] Insert personal connection to Tanzania or Mbeya in motivation letter - [ ] Prepare a one-page project budget for the three-phase TOPOLOGIX work - [ ] Prepare a data management and ethics statement for use of genomic surveillance data - [ ] Submit application through the MUST research calls portal at the provided URL