Draft — FDA–AACR Oncology Educational Fellowship 2026–2027

MOTIVATION LETTER

The resistance of targeted cancer therapies is a regulatory problem before it is a clinical one. When a drug receives FDA approval, the review rests on evidence about who will respond and who will not. My computational work predicts resistance mutations from protein sequence alone, and it does so with coverage that structure-based tools cannot match. TOPOLOGIX, my current research line, 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 an AUROC of 0.804 plus or minus 0.025. It covers 100 percent of mutations, where structure-limited tools like mCSM-lig cover roughly 18 percent and score near 0.70. This is the kind of evidence that could inform how regulators evaluate targeted therapies across a patient population, not just the subset with solved crystal structures.

My path to this work is unconventional. I hold a B.Pharm from the University of Ibadan with a German equivalent grade of 1.9, and I am enrolled in the M.Sc. Digital Health program at the Hasso Plattner Institute and University of Potsdam for Winter Semester 2026/27. I am a licensed pharmacist in Nigeria. Before TOPOLOGIX, I ran a pre-registered, powered replication testing whether bipartite persistent homology predicts hERG cardiotoxicity from protein-ligand interface geometry. The topological features did not beat a plain descriptor baseline, AUROC 0.8426 versus 0.8782. That negative result, which the published literature had never actually tested, taught me to trust falsification over publication pressure. I then applied the same topological constructs to drug-resistance prediction and found they carried almost no signal, AUROC 0.425 and 0.485 on the Platinum benchmark. That ruling-out motivated the sequence-representation approach that became TOPOLOGIX.

The FDA-AACR Oncology Educational Fellowship is the right venue for this pivot. The program connects oncology drug development to regulatory science, and my work sits exactly at that intersection. I am applying as an independent researcher with a strong Africa angle: drug resistance in oncology is a global problem, and computational tools that do not depend on expensive structural data are more accessible to researchers and regulators in low-resource settings. I meet the eligibility criteria as a PharmD graduate within the past ten years. I will submit my AACR membership application before applying. I am prepared to attend the monthly virtual sessions, the AACR Annual Meeting 2027, and the Project ODAC Odyssey at FDA. I am not seeking a stipend; I understand the fellowship covers travel and lodging for in-person events. I am seeking the training, the network, and the regulatory context that this program provides.

RESEARCH STATEMENT

My research program asks a single question: can we predict how drugs fail before they fail in patients? The answer, for resistance mutations, is yes, if we use the right representation of the protein. My current project, TOPOLOGIX, demonstrates this. It predicts drug-resistance mutations from sequence alone using ESM-2 protein-language-model delta-embeddings, Morgan/ECFP drug fingerprints, and a Random Forest classifier. 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 structure-based baselines such as mCSM-lig, which scores near 0.70, while covering 100 percent of mutations compared to roughly 18 percent for structure-limited tools. The implication for oncology is direct: resistance to targeted cancer therapies often arises from mutations that are not captured in solved structures, and sequence-based prediction can fill that gap.

The path to TOPOLOGIX ran through two falsified hypotheses. First, I tested whether bipartite persistent homology, using an opposition-distance metric implemented with Ripser and GUDHI, could predict hERG cardiotoxicity from protein-ligand interface geometry. In a pre-registered, powered replication, the topological features did not beat a plain descriptor baseline, AUROC 0.8426 versus 0.8782. Second, I applied the same topological constructs to drug-resistance prediction on the Platinum benchmark. The AUROC values were 0.425 and 0.485, essentially no signal. These results ruled out interface geometry as the driver of resistance and motivated the sequence-representation approach. I report these negative results directly because they define the boundary of what topological methods can do, and they are the reason TOPOLOGIX exists.

My broader computational pharmacology toolkit supports this work. I have built neurocascade, a receptor-to-behavior brain-circuit simulation engine that couples pharmacokinetics to receptor binding to Wilson-Cowan circuit dynamics, with 62 of 62 tests passing. I have developed the CCT model, a tripartite pharmacological framework for reward-memory encoding prevention in addiction, with all five pre-registered hypotheses confirmed and posterior super-additivity of 13 to 22 percentage points across model versions. I have extended Koopman-operator and Dynamic Mode Decomposition methods with a Mori-Zwanzig memory kernel in the ergofluids project, testing macromolecular drug-vehicle transport through tumor tissue. The ergofluids project is gated: synthetic-data gates passed, but the first real-data gate did not meet its primary pre-registered criterion, and I reported that result directly. These projects share a methodological core: Bayesian calibration with PyMC, rigorous pre-registration, and honest reporting of what does not work.

For this fellowship, I propose to extend TOPOLOGIX toward oncology-specific resistance prediction. The Platinum benchmark is a general drug-resistance dataset; the next step is to train and validate on oncology-specific mutation databases, including tyrosine kinase inhibitor resistance in lung cancer and BCR-ABL mutations in chronic myeloid leukemia. The regulatory science angle is central. If sequence-based resistance prediction can be validated across cancer types, it could inform how the FDA evaluates the breadth of a targeted therapy's coverage during review. I am not proposing a product. I am proposing a method and a validation pathway that regulatory scientists can assess on its merits.

The Africa angle is practical, not rhetorical. Structural data for resistance mutations is scarce in African research contexts, and sequence-based tools run on standard hardware. My background includes building four independent DuckDB-based ingest-to-analyze pipelines across life-sciences and other domains, and self-hosted local LLM serving with llama.cpp. These skills mean the methods I develop can be deployed in low-resource settings without cloud dependencies. I am an independent researcher, which has taught me to design studies that do not require a large lab: pre-registration, public benchmarks, and reproducible pipelines are my infrastructure.

ESSAY RESPONSE: PROFESSIONAL BIOGRAPHY

I am a pharmacist and computational researcher based in Nigeria, currently enrolled in the M.Sc. Digital Health program at the Hasso Plattner Institute and University of Potsdam. I hold a B.Pharm from the University of Ibadan with a German equivalent grade of 1.9 and am licensed by the Pharmacists Council of Nigeria. My research spans addiction neuroscience, protein machine learning, and dynamical-systems methods, unified by a commitment to pre-registration and honest reporting of negative results.

My current project, TOPOLOGIX, predicts drug-resistance mutations from protein sequence alone using ESM-2 protein-language-model delta-embeddings and drug fingerprints, achieving an AUROC of 0.804 on the Platinum benchmark while covering 100 percent of mutations. This work grew out of two falsified hypotheses about topological data analysis, which I reported directly rather than reframed. I have also built neurocascade, a receptor-to-behavior brain-circuit simulation engine with 62 passing tests, and developed the CCT model for reward-memory encoding prevention in addiction, with all five pre-registered hypotheses confirmed.

My professional experience includes work as a National Product Manager at Synthcare, a Clinical Pharmacist at Ramset Pharmacy, and research roles in antimicrobial resistance genomics and NMDA/insulin docking. I have collaborated with or received endorsements from Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU. My technical skills include Python, PyMC for Bayesian calibration, topological data analysis with Ripser and GUDHI, and production systems operations. I am applying to the FDA-AACR Oncology Educational Fellowship to pivot my computational methods toward oncology drug development and regulatory science, where sequence-based resistance prediction can inform how targeted therapies are evaluated.

ESSAY RESPONSE: STATEMENT OF INTEREST

I am applying to the FDA-AACR Oncology Educational Fellowship because my computational work on drug-resistance prediction has reached the point where it needs regulatory context. TOPOLOGIX, my current project, predicts resistance mutations from protein sequence alone with an AUROC of 0.804 on the Platinum benchmark and 100 percent coverage of mutations. The method is validated on a general benchmark. The next step is to understand how such predictions could inform FDA review of targeted cancer therapies, and this fellowship is the venue where that question is asked seriously.

My interest in oncology is recent but grounded. My PharmD training covered oncology pharmacology, and my computational work on hERG cardiotoxicity and drug resistance has been in the service of drug safety and efficacy. The negative results I obtained with topological methods, which I reported directly, taught me that the field needs rigorous falsification more than it needs another promising method. TOPOLOGIX is the positive result that emerged from that rigor. I want to bring that same standard to regulatory science.

The fellowship's structure fits my situation. I am an independent researcher, so the monthly virtual sessions and the two in-person events, the AACR Annual Meeting 2027 and the Project ODAC Odyssey at FDA, are accessible to me. I am not seeking a stipend. I will submit my AACR membership application before applying. I meet the eligibility criteria as a PharmD graduate within the past ten years. The Africa angle matters to me: sequence-based resistance prediction is more accessible to researchers and regulators in low-resource settings than structure-based tools, and I have the infrastructure skills to deploy such methods without cloud dependencies. I am asking for training and context, not funding, and I will bring a validated computational method and a track record of honest reporting to the cohort.

CHECKLIST

- [ ] Confirm AACR membership application is submitted before the fellowship application deadline
- [ ] Verify that B.Pharm from University of Ibadan qualifies as terminal advanced degree equivalent to PharmD for eligibility
- [ ] Draft professional biography, 250 words maximum, incorporating the text above
- [ ] Draft statement of interest, 300 words maximum, incorporating the text above
- [ ] Request two recommendation letters, each maximum 2 pages, from individuals who can assess scientific or clinical experience
- [ ] Confirm availability for monthly virtual sessions, AACR Annual Meeting 2027, and Project ODAC Odyssey at FDA
- [ ] Verify all dates and deadlines on the oncodaily.com listing and the AACR website
- [ ] Prepare a current CV listing all research lines, publications, and preprints with DOIs or OSF/Zenodo links
- [ ] Confirm enrollment status at Hasso Plattner Institute and University of Potsdam for Winter Semester 2026/27
- [ ] Submit application before the 2026-08-03 deadline

EDITOR NOTES

- Eligibility risk: the fellowship requires a terminal advanced degree completed within the past 10 years. The B.Pharm from University of Ibadan is listed as equivalent to PharmD, but the applicant should verify that the Nigerian B.Pharm is accepted as equivalent by AACR. The M.Sc. enrollment at HPI/Potsdam is not yet started as of the application date, so the B.Pharm is the only qualifying degree.
- The applicant must confirm AACR membership status. The profile does not state whether membership has been applied for or granted. This must be resolved before the application is submitted.
- The statement of interest pivots from addiction neuroscience to oncology. The CCT model and neurocascade are not mentioned in the application materials because they are not relevant to this fellowship. The applicant should be prepared to explain the pivot in interviews or recommendation letters, and the recommenders should be briefed on the oncology framing.
- The TOPOLOGIX results are reported as AUROC values on the Platinum benchmark and SKEMPI 2.0. The applicant should verify these numbers against the current preprints and be prepared to provide the benchmark versions and train/test splits used.
- The fellowship provides no salary or stipend. The applicant should confirm that travel and lodging for the two in-person events are covered, and that the monthly virtual sessions are compatible with their time zone and work schedule at Synthcare.

MOTIVATION LETTER

The resistance of targeted cancer therapies is a regulatory problem before it is a clinical one. When a drug receives FDA approval, the review rests on evidence about who will respond and who will not. My computational work predicts resistance mutations from protein sequence alone, and it does so with coverage that structure-based tools cannot match. TOPOLOGIX, my current research line, 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 an AUROC of 0.804 plus or minus 0.025. It covers 100 percent of mutations, where structure-limited tools like mCSM-lig cover roughly 18 percent and score near 0.70. This is the kind of evidence that could inform how regulators evaluate targeted therapies across a patient population, not just the subset with solved crystal structures.

My path to this work is unconventional. I hold a B.Pharm from the University of Ibadan with a German equivalent grade of 1.9, and I am enrolled in the M.Sc. Digital Health program at the Hasso Plattner Institute and University of Potsdam for Winter Semester 2026/27. I am a licensed pharmacist in Nigeria. Before TOPOLOGIX, I ran a pre-registered, powered replication testing whether bipartite persistent homology predicts hERG cardiotoxicity from protein-ligand interface geometry. The topological features did not beat a plain descriptor baseline, AUROC 0.8426 versus 0.8782. That negative result, which the published literature had never actually tested, taught me to trust falsification over publication pressure. I then applied the same topological constructs to drug-resistance prediction and found they carried almost no signal, AUROC 0.425 and 0.485 on the Platinum benchmark. That ruling-out motivated the sequence-representation approach that became TOPOLOGIX.

The FDA-AACR Oncology Educational Fellowship is the right venue for this pivot. The program connects oncology drug development to regulatory science, and my work sits exactly at that intersection. I am applying as an independent researcher with a strong Africa angle: drug resistance in oncology is a global problem, and computational tools that do not depend on expensive structural data are more accessible to researchers and regulators in low-resource settings. I meet the eligibility criteria as a PharmD graduate within the past ten years. I will submit my AACR membership application before applying. I am prepared to attend the monthly virtual sessions, the AACR Annual Meeting 2027, and the Project ODAC Odyssey at FDA. I am not seeking a stipend; I understand the fellowship covers travel and lodging for in-person events. I am seeking the training, the network, and the regulatory context that this program provides.

RESEARCH STATEMENT

My research program asks a single question: can we predict how drugs fail before they fail in patients? The answer, for resistance mutations, is yes, if we use the right representation of the protein. My current project, TOPOLOGIX, demonstrates this. It predicts drug-resistance mutations from sequence alone using ESM-2 protein-language-model delta-embeddings, Morgan/ECFP drug fingerprints, and a Random Forest classifier. 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 structure-based baselines such as mCSM-lig, which scores near 0.70, while covering 100 percent of mutations compared to roughly 18 percent for structure-limited tools. The implication for oncology is direct: resistance to targeted cancer therapies often arises from mutations that are not captured in solved structures, and sequence-based prediction can fill that gap.

The path to TOPOLOGIX ran through two falsified hypotheses. First, I tested whether bipartite persistent homology, using an opposition-distance metric implemented with Ripser and GUDHI, could predict hERG cardiotoxicity from protein-ligand interface geometry. In a pre-registered, powered replication, the topological features did not beat a plain descriptor baseline, AUROC 0.8426 versus 0.8782. Second, I applied the same topological constructs to drug-resistance prediction on the Platinum benchmark. The AUROC values were 0.425 and 0.485, essentially no signal. These results ruled out interface geometry as the driver of resistance and motivated the sequence-representation approach. I report these negative results directly because they define the boundary of what topological methods can do, and they are the reason TOPOLOGIX exists.

My broader computational pharmacology toolkit supports this work. I have built neurocascade, a receptor-to-behavior brain-circuit simulation engine that couples pharmacokinetics to receptor binding to Wilson-Cowan circuit dynamics, with 62 of 62 tests passing. I have developed the CCT model, a tripartite pharmacological framework for reward-memory encoding prevention in addiction, with all five pre-registered hypotheses confirmed and posterior super-additivity of 13 to 22 percentage points across model versions. I have extended Koopman-operator and Dynamic Mode Decomposition methods with a Mori-Zwanzig memory kernel in the ergofluids project, testing macromolecular drug-vehicle transport through tumor tissue. The ergofluids project is gated: synthetic-data gates passed, but the first real-data gate did not meet its primary pre-registered criterion, and I reported that result directly. These projects share a methodological core: Bayesian calibration with PyMC, rigorous pre-registration, and honest reporting of what does not work.

For this fellowship, I propose to extend TOPOLOGIX toward oncology-specific resistance prediction. The Platinum benchmark is a general drug-resistance dataset; the next step is to train and validate on oncology-specific mutation databases, including tyrosine kinase inhibitor resistance in lung cancer and BCR-ABL mutations in chronic myeloid leukemia. The regulatory science angle is central. If sequence-based resistance prediction can be validated across cancer types, it could inform how the FDA evaluates the breadth of a targeted therapy's coverage during review. I am not proposing a product. I am proposing a method and a validation pathway that regulatory scientists can assess on its merits.

The Africa angle is practical, not rhetorical. Structural data for resistance mutations is scarce in African research contexts, and sequence-based tools run on standard hardware. My background includes building four independent DuckDB-based ingest-to-analyze pipelines across life-sciences and other domains, and self-hosted local LLM serving with llama.cpp. These skills mean the methods I develop can be deployed in low-resource settings without cloud dependencies. I am an independent researcher, which has taught me to design studies that do not require a large lab: pre-registration, public benchmarks, and reproducible pipelines are my infrastructure.

ESSAY RESPONSE: PROFESSIONAL BIOGRAPHY

I am a pharmacist and computational researcher based in Nigeria, currently enrolled in the M.Sc. Digital Health program at the Hasso Plattner Institute and University of Potsdam. I hold a B.Pharm from the University of Ibadan with a German equivalent grade of 1.9 and am licensed by the Pharmacists Council of Nigeria. My research spans addiction neuroscience, protein machine learning, and dynamical-systems methods, unified by a commitment to pre-registration and honest reporting of negative results.

My current project, TOPOLOGIX, predicts drug-resistance mutations from protein sequence alone using ESM-2 protein-language-model delta-embeddings and drug fingerprints, achieving an AUROC of 0.804 on the Platinum benchmark while covering 100 percent of mutations. This work grew out of two falsified hypotheses about topological data analysis, which I reported directly rather than reframed. I have also built neurocascade, a receptor-to-behavior brain-circuit simulation engine with 62 passing tests, and developed the CCT model for reward-memory encoding prevention in addiction, with all five pre-registered hypotheses confirmed.

My professional experience includes work as a National Product Manager at Synthcare, a Clinical Pharmacist at Ramset Pharmacy, and research roles in antimicrobial resistance genomics and NMDA/insulin docking. I have collaborated with or received endorsements from Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU. My technical skills include Python, PyMC for Bayesian calibration, topological data analysis with Ripser and GUDHI, and production systems operations. I am applying to the FDA-AACR Oncology Educational Fellowship to pivot my computational methods toward oncology drug development and regulatory science, where sequence-based resistance prediction can inform how targeted therapies are evaluated.

ESSAY RESPONSE: STATEMENT OF INTEREST

I am applying to the FDA-AACR Oncology Educational Fellowship because my computational work on drug-resistance prediction has reached the point where it needs regulatory context. TOPOLOGIX, my current project, predicts resistance mutations from protein sequence alone with an AUROC of 0.804 on the Platinum benchmark and 100 percent coverage of mutations. The method is validated on a general benchmark. The next step is to understand how such predictions could inform FDA review of targeted cancer therapies, and this fellowship is the venue where that question is asked seriously.

My interest in oncology is recent but grounded. My PharmD training covered oncology pharmacology, and my computational work on hERG cardiotoxicity and drug resistance has been in the service of drug safety and efficacy. The negative results I obtained with topological methods, which I reported directly, taught me that the field needs rigorous falsification more than it needs another promising method. TOPOLOGIX is the positive result that emerged from that rigor. I want to bring that same standard to regulatory science.

The fellowship's structure fits my situation. I am an independent researcher, so the monthly virtual sessions and the two in-person events, the AACR Annual Meeting 2027 and the Project ODAC Odyssey at FDA, are accessible to me. I am not seeking a stipend. I will submit my AACR membership application before applying. I meet the eligibility criteria as a PharmD graduate within the past ten years. The Africa angle matters to me: sequence-based resistance prediction is more accessible to researchers and regulators in low-resource settings than structure-based tools, and I have the infrastructure skills to deploy such methods without cloud dependencies. I am asking for training and context, not funding, and I will bring a validated computational method and a track record of honest reporting to the cohort.

CHECKLIST

- [ ] Confirm AACR membership application is submitted before the fellowship application deadline
- [ ] Verify that B.Pharm from University of Ibadan qualifies as terminal advanced degree equivalent to PharmD for eligibility
- [ ] Draft professional biography, 250 words maximum, incorporating the text above
- [ ] Draft statement of interest, 300 words maximum, incorporating the text above
- [ ] Request two recommendation letters, each maximum 2 pages, from individuals who can assess scientific or clinical experience
- [ ] Confirm availability for monthly virtual sessions, AACR Annual Meeting 2027, and Project ODAC Odyssey at FDA
- [ ] Verify all dates and deadlines on the oncodaily.com listing and the AACR website
- [ ] Prepare a current CV listing all research lines, publications, and preprints with DOIs or OSF/Zenodo links
- [ ] Confirm enrollment status at Hasso Plattner Institute and University of Potsdam for Winter Semester 2026/27
- [ ] Submit application before the 2026-08-03 deadline

EDITOR NOTES

- Eligibility risk: the fellowship requires a terminal advanced degree completed within the past 10 years. The B.Pharm from University of Ibadan is listed as equivalent to PharmD, but the applicant should verify that the Nigerian B.Pharm is accepted as equivalent by AACR. The M.Sc. enrollment at HPI/Potsdam is not yet started as of the application date, so the B.Pharm is the only qualifying degree.
- The applicant must confirm AACR membership status. The profile does not state whether membership has been applied for or granted. This must be resolved before the application is submitted.
- The statement of interest pivots from addiction neuroscience to oncology. The CCT model and neurocascade are not mentioned in the application materials because they are not relevant to this fellowship. The applicant should be prepared to explain the pivot in interviews or recommendation letters, and the recommenders should be briefed on the oncology framing.
- The TOPOLOGIX results are reported as AUROC values on the Platinum benchmark and SKEMPI 2.0. The applicant should verify these numbers against the current preprints and be prepared to provide the benchmark versions and train/test splits used.
- The fellowship provides no salary or stipend. The applicant should confirm that travel and lodging for the two in-person events are covered, and that the monthly virtual sessions are compatible with their time zone and work schedule at Synthcare.

AI Draft — FDA–AACR Oncology Educational Fellowship 2026–2027