MOTIVATION LETTER The opioid crisis in West Africa is underreported, understudied, and undertreated. Nigeria has no functional methadone maintenance programme, no national naloxone distribution protocol, and zero computational models of addiction tailored to its population. I built one. The Conjunctive Consolidation Threshold model is a tripartite pharmacological framework that couples dopaminergic reward prediction error, NMDAR-dependent long-term potentiation, and affective contrast into a single ODE system. I calibrated it with Bayesian MCMC using priors drawn from a systematic screen of 1,847 records. All five pre-registered hypotheses confirmed. Posterior super-additivity ranged from 13 to 22 percentage points across model versions. Three sole-authored preprints on OSF and Zenodo. One co-authored paper under review at Alcohol. Emergent Ventures funds people who build things that did not exist before. I built a model that predicts drug-resistance mutations from protein sequence alone, covering 100 percent of mutations versus roughly 18 percent for structure-limited tools. TOPOLOGIX uses ESM-2 protein-language-model delta-embeddings combined with Morgan drug fingerprints and a Random Forest classifier. It achieves an AUROC of 0.804 on the Platinum benchmark and 0.634 on SKEMPI 2.0. It beats mCSM-lig by ten points while requiring no crystal structure. I also built neurocascade, a receptor-to-behavior brain-circuit simulation engine that couples pharmacokinetics to Wilson-Cowan dynamics across three calibrated receptor systems. Sixty-two of sixty-two tests pass. I built ergofluids, a Koopman-operator method with a Mori-Zwanzig memory kernel for modeling drug transport through tumor tissue. The first real-data gate did not meet its pre-registered criterion. I reported that directly rather than reframing it. I am a 29-year-old Nigerian pharmacist with a B.Pharm from the University of Ibadan, a 2.1 Upper Division, and a German equivalent of 1.9. I am enrolled in the M.Sc. Digital Health programme at the Hasso Plattner Institute starting winter 2026. I have endorsements from Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU. I have no PhD, no institutional lab, and no grant history. I have seven active research lines, four pre-registered studies, and a track record of shipping code and papers from a laptop in Lagos. Emergent Ventures has a dedicated Africa track. I am based in Lagos. My research targets problems that kill people in my country: addiction, antimicrobial resistance, and cardiotoxicity from unregulated drug markets. The CCT model can guide pharmacotherapy decisions in settings where no psychiatrist is available. TOPOLOGIX can predict resistance mutations for pathogens circulating in West Africa without requiring a structural biology lab. These are not proposals. They are running code with published results. I am applying for 15,000 dollars. This will fund compute time on a university cluster, a GPU workstation for protein-language-model inference, and travel to present the CCT model at a computational neuroscience conference. The output will be a fourth preprint extending the CCT model to opioid-specific parameters calibrated against human behavioral data, and a deployed TOPOLOGIX web interface for African researchers to submit sequences and receive resistance predictions within minutes. SHORT ESSAY The CCT model addresses a specific gap in addiction neuroscience. Existing models treat reward memory encoding as a single process. The CCT model separates it into three coupled axes: dopaminergic reward prediction error, NMDAR-dependent long-term potentiation, and affective contrast. Each axis is a differential equation. The coupling terms are the novel contribution. I derived them from first principles, wrote the ODE system in Python using RK45, and calibrated the 14 free parameters with PyMC using the DEMetropolisZ sampler. The priors came from a systematic literature screen of 1,847 records. The model confirmed all five pre-registered hypotheses. The posterior showed super-additivity of 13 to 22 percentage points across versions, meaning the combined effect of the three axes exceeds the sum of their individual effects. This matters for Nigeria because the country has no computational framework for addiction treatment. Clinicians rely on trial and error. The CCT model can simulate the effect of a given drug combination on reward memory encoding before a single pill is prescribed. It runs on a laptop. It requires no wet lab. It can be deployed in a primary care clinic in Lagos or Kano. The model is open source. The code is on GitHub. The preprints are on OSF and Zenodo. The paper is under review at Alcohol. I am the sole author on three of the four outputs. This is not a group project. This is one person with a laptop, a literature database, and a Bayesian sampler. SHORT ESSAY TOPOLOGIX predicts drug-resistance mutations from protein sequence alone. It uses ESM-2 protein-language-model delta-embeddings, Morgan drug fingerprints, and a Random Forest classifier. It achieves an AUROC of 0.804 on the Platinum benchmark of 553 mutations and 0.634 on SKEMPI 2.0. It beats mCSM-lig, which scores roughly 0.70, while covering 100 percent of mutations versus roughly 18 percent for structure-limited tools. The problem is that most drug-resistance prediction tools require a protein crystal structure. For many pathogens circulating in Africa, no structure exists. TOPOLOGIX bypasses this entirely. It takes a sequence, runs it through ESM-2, computes the delta-embedding relative to the wild type, combines it with the drug fingerprint, and outputs a prediction. No homology modeling. No docking. No waiting for a crystallographer. I built TOPOLOGIX after a failed experiment. I tested whether bipartite persistent homology could predict drug resistance from protein-ligand interface geometry. It could not. The AUROC was 0.425 on the Platinum benchmark. I reported that result directly in a pre-registered, powered replication. Then I pivoted to sequence representations. The pivot worked. TOPOLOGIX is the result. The code is on GitHub. The preprints are on Zenodo. The model runs inference in under a second per mutation. I am currently building a web interface using Supabase and a JavaScript front end so that African researchers can submit sequences and receive predictions without installing anything. RESEARCH STATEMENT My research programme has three pillars: addiction neuroscience, protein-drug machine learning, and dynamical-systems methods for pharmacology. Each pillar produces open-source code, pre-registered studies, and sole-authored preprints. None of them requires a wet lab or a large team. All of them target problems that kill people in low-resource settings. The addiction neuroscience pillar centers on the CCT model. The model is a tripartite pharmacological framework for reward-memory encoding prevention. It couples dopaminergic reward prediction error, NMDAR-dependent long-term potentiation, and affective contrast into a single ODE system. I calibrated it with Bayesian MCMC using literature-elicited priors. All five pre-registered hypotheses confirmed. Posterior super-additivity 13 to 22 percentage points. Three sole-authored preprints. One co-authored paper under review at Alcohol. The next step is to extend the model to opioid-specific parameters and calibrate it against human behavioral data from published studies. I have already identified the datasets and written the data extraction pipeline. The protein-drug machine learning pillar includes TOPOLOGIX and two negative-result studies. The negative-result studies are as important as the positive ones. The hERG cardiotoxicity study tested whether bipartite persistent homology predicts cardiotoxicity from protein-ligand interface geometry. It found that topological features do not beat a plain descriptor baseline. AUROC 0.8426 versus 0.8782. The interface-topology-for-resistance study found that the same constructs carry almost no signal for drug-resistance prediction. AUROC 0.425 and 0.485 on the Platinum benchmark. Both studies were pre-registered and powered. Both results are published on Zenodo. TOPOLOGIX is the positive result that came from learning what does not work. The dynamical-systems pillar includes neurocascade and ergofluids. 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 systems. Sixty-two of sixty-two tests passing. Ergofluids extends Koopman-operator methods with a Mori-Zwanzig memory kernel for modeling drug transport through tumor tissue. The first real-data gate did not meet its pre-registered criterion. I reported that directly. All code is on GitHub. All preprints are on OSF and Zenodo. All studies are pre-registered on OSF. I have no PhD, no institutional lab, and no grant history. I have endorsements from Kent Berridge, Samuel Gershman, Nathaniel Daw, and Marcelo Mattar. I am enrolled in the M.Sc. Digital Health programme at the Hasso Plattner Institute starting winter 2026. I am a licensed pharmacist in Nigeria. I am 29 years old. CHECKLIST - [ ] Motivation letter (500 words, written above) - [ ] Short essay 1: CCT model (350 words, written above) - [ ] Short essay 2: TOPOLOGIX (350 words, written above) - [ ] Research statement (600 words, written above) - [ ] CV or resume (applicant must compile from profile data) - [ ] ORCID profile link: 0009-0001-9272-6735 - [ ] GitHub profile link: github.com/AmunRaPtah - [ ] Personal website link: zyco.org - [ ] Proof of enrollment: M.Sc. Digital Health, Hasso Plattner Institute, Winter Semester 2026/27 - [ ] Two preprint links from OSF or Zenodo (CCT model and TOPOLOGIX) - [ ] One published paper link (Alcohol, under review, provide DOI if available) - [ ] Endorsement letters or contact information for Kent Berridge, Samuel Gershman, Nathaniel Daw, Marcelo Mattar (optional but recommended) EDITOR NOTES - Eligibility risk: Emergent Ventures has no strict age or degree requirement, but the applicant is enrolled in a master's programme starting winter 2026. Confirm that the fellowship does not require full-time commitment or prohibit concurrent enrollment. - Fact verification: The German equivalent GPA of 1.9 for a 5.1/7.0 Nigerian CGPA should be verified against the official conversion table used by German universities. Some conversion formulas produce different results. - Gap: The applicant's employment history shows a transition from clinical pharmacist to national product manager at Synthcare in March 2026. The motivation letter does not explain how this role relates to the research programme. The applicant should add a sentence connecting the product management experience to the goal of deploying TOPOLOGIX as a web service. - Gap: The applicant lists endorsements from four senior researchers but does not specify whether any of them have agreed to write letters. The checklist includes endorsement letters as optional. If the applicant can secure one letter, it should be from Kent Berridge given the direct relevance to the CCT model. - Risk: The ergofluids negative result is reported honestly, but the Emergent Ventures selection criteria include a bias toward action and tangible output. The applicant should emphasize that the negative result was pre-registered and reported directly, which demonstrates scientific integrity and self-correction. The current draft does this adequately but could be more explicit.