MOTIVATION LETTER A computational model of reward-memory encoding that predicts addiction treatment outcomes from pharmacology alone, and a protein-language model that predicts drug resistance mutations across 100 percent of the genome instead of the 18 percent reachable by structure-based tools: these are the two research lines I bring to the Microsoft Africa Research Institute Fellowship 2026. I am Eniola Ayodele Olutogun, a Nigerian pharmacist and independent computational researcher, and I am applying to the MicRise programme to scale both projects with Microsoft Research Africa's computational infrastructure and domain expertise. The CCT model, Conjunctive Consolidation Threshold, is a tripartite pharmacological framework I built from first principles. It couples dopaminergic reward-prediction error, NMDAR-dependent long-term potentiation, and affective contrast into a single ODE system solved with RK45. I calibrated its 14 free parameters using Bayesian MCMC (PyMC DEMetropolisZ) against priors elicited from a systematic screen of 1,847 records. All five pre-registered hypotheses confirmed. Posterior super-additivity ranges from 13 to 22 percentage points across model versions. Three sole-authored preprints are under review at peer-reviewed journals. This model, if validated against clinical data, could guide pharmacotherapy selection for addiction in African populations where genetic and environmental factors differ from the Western cohorts that dominate the literature. The TOPOLOGIX pipeline addresses a different problem with the same computational rigor. Drug resistance mutations in pathogens render standard treatments ineffective. Structure-based tools cover only mutations in proteins with solved crystal structures, roughly 18 percent of the total. TOPOLOGIX 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, it achieves AUROC 0.804 plus or minus 0.025. On SKEMPI 2.0, AUROC 0.634. It covers every mutation, not just those with structures. For African health systems where antimicrobial resistance is a growing crisis, a sequence-only predictor that works without waiting for crystallography has immediate practical value. Microsoft Research Africa's focus on AI-driven health technology aligns directly with both projects. The fellowship would allow me to port the CCT model's Bayesian calibration pipeline to Azure Machine Learning, running parallel MCMC chains at scale. TOPOLOGIX could be retrained on African pathogen genomic data available through public repositories, then deployed as a lightweight API for clinical decision support. I hold a B.Pharm from the University of Ibadan, am enrolled in the M.Sc. Digital Health at Hasso Plattner Institute / University of Potsdam starting Winter 2026/27, and have endorsements from Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU. I am ready to collaborate with Microsoft scientists, publish in top venues, and build tools that serve African patients directly. RESEARCH STATEMENT My research programme spans three domains: addiction neuroscience, protein-drug machine learning, and dynamical-systems methods. The connecting thread is computational modeling of biological systems at the interface of pharmacology and data science, with a consistent emphasis on Bayesian calibration, pre-registered hypothesis testing, and direct reporting of negative results. The CCT model addresses a fundamental question in addiction neuroscience: can a pharmacological intervention prevent the encoding of reward-memory associations that drive relapse? I formulated the model as three coupled ordinary differential equations representing dopaminergic reward-prediction error, NMDAR-dependent long-term potentiation, and affective contrast. The system was solved with RK45 and calibrated using Bayesian MCMC with 14 free parameters. Priors came from a systematic literature screen of 1,847 records. All five pre-registered hypotheses were confirmed. Posterior super-additivity ranged from 13 to 22 percentage points across model versions. Three sole-authored preprints are under review at IART, PNPBP, and NBR. A co-authored paper is under review at Alcohol (Elsevier). The next step is to fit the model to human behavioral data from published clinical trials, which requires access to individual-level data and computational resources for large-scale Bayesian inference. The TOPOLOGIX project emerged from a systematic failure. I tested whether bipartite persistent homology, a topological data analysis method, could predict hERG cardiotoxicity from protein-ligand interface geometry. It could not. A pre-registered, powered replication showed topological features underperformed a plain descriptor baseline: AUROC 0.8426 versus 0.8782. I then applied the same method to drug resistance prediction. Again, it failed: AUROC 0.425 and 0.485 on the Platinum benchmark. These negative results, reported directly rather than reframed, motivated a shift to sequence-based methods. TOPOLOGIX uses ESM-2 protein-language-model delta-embeddings plus Morgan/ECFP fingerprints and a Random Forest classifier. It achieves AUROC 0.804 on Platinum and 0.634 on SKEMPI 2.0, covering 100 percent of mutations versus roughly 18 percent for structure-limited tools. The pipeline is implemented in Python with RDKit, scikit-learn, and Hugging Face transformers, and runs on standard HPC infrastructure. A third project, neurocascade, is a receptor-to-behavior brain-circuit simulation engine. It couples pharmacokinetics to receptor binding to Wilson-Cowan circuit dynamics to behavioral-readout ODE layers. Three literature-calibrated receptor/circuit systems, mu-opioid, D2 dopamine, GABA-A, have been Bayesian-calibrated with PyMC. All 62 tests pass. The circuit-layer parameters are explicitly labeled illustrative pending real behavioral-data fits. This engine could eventually simulate how a given drug dose, in a given patient genotype, produces a given behavioral outcome, a tool for personalized pharmacotherapy in psychiatry. My technical stack includes Python (scipy, numpy, PyMC, pandas), R, topological data analysis (Ripser, Gudhi), neural simulation (NEURON, Brian2), protein modeling (AlphaFold, RDKit, GROMACS, AutoDock), and HPC workflow management (Nextflow, SLURM). I have built four independent DuckDB-based ingest-to-analyze corpus pipelines across life sciences, tech/AI/security, and social science domains. I self-host local LLM serving with llama.cpp and manage production systems on Linux VPS with systemd, Caddy TLS, CI/CD, and automated backup/disaster-recovery. The MicRise fellowship would enable three specific advances. First, scaling the CCT model's Bayesian calibration to Azure Machine Learning for parallel MCMC across hundreds of chains. Second, retraining TOPOLOGIX on African pathogen genomic data and deploying it as a clinical decision support API. Third, establishing a collaboration with Microsoft Research Africa scientists to publish the CCT model in a high-impact neuroscience journal and TOPOLOGIX in a machine learning for health venue. My Nigerian nationality, current enrollment in a German digital health MSc, and track record of independent research make me well-suited for this early-career programme. SHORT ESSAY: RELEVANCE TO MICROSOFT RESEARCH AFRICA Microsoft Research Africa's mission is to conduct research that addresses African challenges using technology. My work fits this mission directly. The CCT model targets addiction treatment, a major public health issue in Nigeria where substance use disorders are underdiagnosed and undertreated. A computational model that predicts which pharmacotherapy regimen works for which patient, calibrated on African genetic and environmental data, would be a practical tool for clinicians. TOPOLOGIX addresses antimicrobial resistance, which the Africa CDC has identified as a silent pandemic. A sequence-only predictor that works without crystal structures is immediately deployable in African laboratories where structural biology infrastructure is limited. Both projects use AI and machine learning: the CCT model uses Bayesian MCMC, TOPOLOGIX uses protein language models. This aligns with Microsoft's AI/ML research priority. The fellowship would allow me to scale both projects using Azure infrastructure and collaborate with Microsoft scientists who have domain expertise in health AI and deployment in low-resource settings. SHORT ESSAY: CAREER IMPACT The MicRise fellowship would be the institutional home I currently lack. I am an independent researcher with a B.Pharm, enrolled in a digital health MSc, but not yet in a PhD programme. This fellowship would provide the research environment, computational resources, and mentorship needed to transition from independent work to collaborative, institutionally-supported research. Specifically, it would enable me to publish the CCT model in a top neuroscience journal: the Bayesian calibration and pre-registered confirmation of all five hypotheses is publishable work that needs a strong institutional affiliation to be taken seriously by reviewers. It would also allow me to deploy TOPOLOGIX as a working clinical tool rather than a preprint. Long-term, I aim to pursue a PhD in computational neuroscience or machine learning for health. The MicRise fellowship would strengthen my application by demonstrating the ability to produce high-impact research in a collaborative setting. CHECKLIST - [ ] Complete MicRise online application form at micrise.com.ng - [ ] Upload motivation letter (this document) - [ ] Upload research statement (this document) - [ ] Upload short essays (this document) - [ ] Upload CV with ORCID, GitHub, and publication list - [ ] Upload academic transcripts (B.Pharm, University of Ibadan) - [ ] Upload proof of enrollment (M.Sc. Digital Health, HPI/Potsdam) - [ ] Upload two letters of recommendation (Kent Berridge, Samuel Gershman) - [ ] Verify Nigerian citizenship documentation - [ ] Confirm eligibility as early-career researcher not yet in PhD programme EDITOR NOTES - Eligibility risk: The programme description states preference for PhD students, postdocs, or early-career faculty, with exceptional final-year Master's students considered. Eniola is enrolled in an MSc starting Winter 2026/27, not yet a student. Confirm with programme administrators whether enrollment counts as "final-year Master's student" or whether a different category applies. - Facts to verify: The three preprints under review, confirm journal names (IART, PNPBP, NBR) are correct and that the review status is current as of application date. The co-authored paper in Alcohol (Elsevier), confirm status. - Gap: The profile does not specify Eniola's current location. The programme is in Lagos, Nigeria. If Eniola is currently in Germany for the MSc, clarify availability for in-person components. If still in Nigeria, state that explicitly. - Gap: The profile mentions "OIQB" in the strategy notes but this term does not appear in the research profile. Remove or clarify before submission. - Gap: The profile lists "startup/founder programmes" in the task instructions but Eniola's profile does not mention a startup. The TOPOLOGIX work is described as a research pipeline, not a venture. Ensure no language implies a commercial product unless Eniola has a registered company.