MOTIVATION LETTER Drug resistance is a present emergency in Africa, not a future threat. At the GHRU-GSAR programme in Nigeria, I built AMR genomics surveillance pipelines and watched the same resistance mutations appear across bacterial isolates with no computational tool fast enough or complete enough to catch them early. That experience drives my current research: TOPOLOGIX, a sequence-based prediction system for drug-resistance mutations that covers 100 percent of mutations in the Platinum benchmark, compared to roughly 18 percent for structure-limited tools, while achieving an AUROC of 0.804 plus or minus 0.025. It outperforms the structure-based baseline mCSM-lig, which sits near 0.70, and it does so from sequence alone using ESM-2 protein language model delta-embeddings combined with Morgan fingerprints and a Random Forest classifier. I am applying to the L'Oréal-UNESCO For Women in Science programme because its dual commitment to scientific excellence and to advancing women researchers in Sub-Saharan Africa matches both what I do and where I do it. I am a Nigerian citizen, a licensed pharmacist, and an independent computational researcher. My work sits at the intersection of protein machine learning, pharmacology, and dynamical systems. TOPOLOGIX is the current focus because it addresses a concrete global health problem with direct African relevance: predicting which mutations confer drug resistance before they spread, using data that exists for every sequenced pathogen, not just the minority with resolved crystal structures. The L'Oréal-UNESCO fellowship would fund the next validation stage: expanding TOPOLOGIX from the Platinum benchmark of 553 mutations to a prospective test on clinical resistance data from African sequencing efforts, including the surveillance pipelines I worked on at GHRU-GSAR. That extension requires compute for ESM-2 embedding generation and for retraining on larger mutation sets, plus time to write and publish the validation protocol. The fellowship's recognition would also open institutional doors that remain closed to independent researchers in Nigeria, where I currently work without a university laboratory affiliation. My track record demonstrates I can deliver rigorous results independently. I have three sole-authored preprints under peer review at the International Addiction Review, Progress in Neuro-Psychopharmacology and Biological Psychiatry, and Neuroscience and Biobehavioral Reviews, plus a co-authored paper under review at Alcohol. My hERG cardiotoxicity study, a pre-registered replication, settled a question the literature had never actually tested: topological features did not beat a plain descriptor baseline, AUROC 0.8426 versus 0.8782. I report negative results as clearly as positive ones. That honesty is the foundation of credible science. I am enrolled in the M.Sc. Digital Health programme at the Hasso Plattner Institute and University of Potsdam starting Winter Semester 2026/27, which will formalize my computational training and connect me to European research networks. The L'Oréal-UNESCO award would bridge my independent work in Nigeria and my formal graduate training in Germany, and it would signal to young African women that rigorous computational research is possible without leaving the continent. I ask for your consideration. RESEARCH STATEMENT TOPOLOGIX: Sequence-Based Prediction of Drug-Resistance Mutations Using Protein Language Model Embeddings Problem and Motivation Drug resistance kills hundreds of thousands of people in Sub-Saharan Africa each year, yet the standard computational tools for predicting resistance mutations require a resolved protein structure. mCSM-lig, a widely used structure-based predictor, covers only about 18 percent of mutations in the Platinum benchmark because most resistance mutations occur in proteins without crystal structures. This structural bottleneck is the reason resistance prediction fails precisely where sequencing data is most abundant and structural data is most scarce, which is the African clinical context. My prior work on AMR genomics at GHRU-GSAR in Nigeria made this concrete: we could sequence resistance genes quickly, but we could not predict which new mutations would matter. Method and Innovation TOPOLOGIX replaces the structure requirement with protein language model embeddings. The system uses ESM-2 delta-embeddings, which capture the change in the protein's learned representation when a mutation is introduced, combined with Morgan/ECFP drug fingerprints and a Random Forest classifier. The model predicts whether a given mutation confers drug resistance directly from sequence. This design choice is deliberate: sequence data is cheap, universal, and available for every sequenced pathogen, while structural data is expensive, incomplete, and often impossible to obtain for membrane proteins and other difficult targets. Results to Date On the Platinum benchmark of 553 mutations, TOPOLOGIX achieves an AUROC of 0.804 plus or minus 0.025. On SKEMPI 2.0, it achieves 0.634. It beats the structure-based baseline mCSM-lig, which sits near 0.70, while covering 100 percent of mutations versus roughly 18 percent for structure-limited tools. These results are pre-registered and reproducible. The code is available on my GitHub under the TOPOLOGIX repository, and the benchmark protocols follow the original Platinum evaluation design. The path to TOPOLOGIX included a falsified hypothesis. My earlier work tested whether bipartite persistent homology of protein-ligand interfaces could predict hERG cardiotoxicity. A pre-registered, powered replication found that topological features did not beat a plain descriptor baseline, AUROC 0.8426 versus 0.8782. A follow-up study applying the same topological constructs to drug-resistance prediction found they carried almost no signal, AUROC 0.425 and 0.485 on the Platinum benchmark. These negative results ruled out interface geometry as the driver of resistance and motivated the sequence-representation approach that became TOPOLOGIX. I report them here because they define the boundary of what does not work, which is as important as what does. Proposed Work The next phase has three aims. First, expand the training set beyond Platinum to include resistance mutations from clinical isolates collected through African sequencing networks, including the GHRU-GSAR surveillance pipeline I helped build. Second, benchmark TOPOLOGIX against additional structure-based and sequence-based baselines on a held-out clinical set to establish whether the coverage advantage translates to predictive advantage in real clinical data. Third, publish the full validation protocol and results, including negative findings, in an open-access venue. Relevance to Africa and to L'Oréal-UNESCO Drug resistance is a leading cause of treatment failure for tuberculosis, malaria, and bacterial infections across Sub-Saharan Africa. A sequence-based predictor that works without structures is directly applicable to the data African laboratories actually generate. My Nigerian citizenship, my prior work on AMR surveillance in Nigeria, and my current independent research status mean this work is rooted in the region, not merely about it. The L'Oréal-UNESCO For Women in Science programme's commitment to supporting women researchers in Sub-Saharan Africa is the right home for this project, which combines computational rigor, public health relevance, and a demonstrated record of honest, reproducible science. EDITOR NOTES - Research line selected: TOPOLOGIX, because it directly addresses drug resistance, a critical global health challenge with high African relevance, and it is the current active project with the strongest results. The hERG and interface-topology studies are mentioned only as falsified prior work that motivated TOPOLOGIX, never as current claims. - Eligibility risk: The programme requires citizenship of a Sub-Saharan African country and research conducted at a recognized institution within the region. Eniola is Nigerian, which satisfies citizenship, but she is currently an independent researcher without a university affiliation in Nigeria. Verify whether the programme accepts independent researchers or requires institutional sponsorship. If required, identify a Nigerian university or research institute willing to host the project. - Fact verification needed: Confirm the Platinum benchmark size (553 mutations), the SKEMPI 2.0 AUROC (0.634), the mCSM-lig baseline (approximately 0.70), and the coverage figures (100 percent versus 18 percent). Confirm the AUROC values for the hERG study (0.8426 versus 0.8782) and the Platinum topology study (0.425 and 0.485). These numbers come from the applicant profile but must be checked against the actual preprints before submission. - Gap: The motivation letter and research statement do not include a specific budget or timeline. The programme may require a budget breakdown. Insert a one-page budget covering compute costs for ESM-2 embedding generation, open-access publication fees, and travel to a partner laboratory or conference. - Gap: The applicant's M.Sc. enrollment at HPI/Potsdam begins Winter Semester 2026/27. Confirm whether the programme considers her a student or an independent researcher at the time of application, and whether the fellowship can be held while enrolled in a degree programme in Germany. - Gap: No letters of recommendation are mentioned in the profile. The programme likely requires at least two. Kent Berridge, Samuel Gershman, Nathaniel Daw, and Marcelo Mattar are named as endorsers; confirm which of them can provide letters and whether their letters address the TOPOLOGIX project specifically. - Tone check: The motivation letter opens with the problem, not with "I," per the formatting rules. No banned phrases are used. All claims are tied to specific numbers or named institutions. The negative results are presented as deliberate methodological choices, not failures. CHECKLIST - [ ] Confirm eligibility as an independent researcher under the L'Oréal-UNESCO FWIS Sub-Saharan Africa programme - [ ] Verify all benchmark numbers against preprints (Platinum, SKEMPI 2.0, mCSM-lig, hERG, topology studies) - [ ] Obtain or confirm institutional affiliation or host institution in Nigeria, if required - [ ] Secure two letters of recommendation, preferably from Kent Berridge, Samuel Gershman, Nathaniel Daw, or Marcelo Mattar - [ ] Prepare one-page budget for compute, publication fees, and travel - [ ] Confirm M.Sc. enrollment status does not conflict with fellowship eligibility - [ ] Prepare CV in the programme's required format, including ORCID 0009-0001-9272-6735 and GitHub github.com/AmunRaPtah - [ ] Prepare copies of preprints for TOPOLOGIX, hERG study, and interface-topology study - [ ] Submit application through the official L'Oréal-UNESCO FWIS portal before the deadline listed on the programme website - [ ] Confirm the application deadline and required document format from the official programme page, not the aggregator site