Draft — Rowland Fellowship

MOTIVATION LETTER

The Rowland Fellowship supports early-career researchers who want to build an independent research program with the backing of staff scientists and engineers. That structure is precisely what my research trajectory requires. I am a computational neuropharmacologist with a B.Pharm from the University of Ibadan, currently enrolled in the M.Sc. Digital Health program at the Hasso Plattner Institute / University of Potsdam. My central project, the Conjunctive Consolidation Threshold (CCT) model, is a tripartite pharmacological framework for reward-memory encoding prevention in addiction. It is ready for experimental validation, and the Rowland Fellowship's model of pairing an independent investigator with dedicated technical staff is the most direct path to that validation.

The CCT model couples three axes: dopaminergic reward prediction error, NMDAR-dependent long-term potentiation, and affective contrast. I implemented it as a system of ordinary differential equations solved with RK45, calibrated with Bayesian MCMC using PyMC's DEMetropolisZ sampler across 14 free parameters. The priors were elicited from a systematic screen of 1,847 records in the literature. All five pre-registered hypotheses, H1 through H5, were confirmed. The posterior shows super-additivity of 13 to 22 percentage points across model versions, meaning the three axes interact non-additively to suppress reward-memory consolidation. Three sole-authored preprints are under review at peer-reviewed journals: International Addiction Review and Therapeutics, Progress in Neuro-Psychopharmacology and Biological Psychiatry, and Neuroscience and Biobehavioral Reviews. A co-authored paper is under review at Alcohol (Elsevier).

The fellowship's emphasis on high-risk, interdisciplinary research matches the CCT program's next phase. The model makes specific, testable predictions about which pharmacological combinations should block reconsolidation of reward memories. I want to use the Rowland Fellowship to design and run the experimental validation of those predictions. The Rowland staff engineers can help build in-vitro or in-vivo setups that I cannot construct as an independent researcher. My computational skills, including ODE modeling, Bayesian calibration, and machine learning, are the tools needed to drive such a project from design through analysis.

I have also done rigorous negative-result work that demonstrates my commitment to falsification. In a pre-registered, powered replication study, I tested 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. I reported this result directly. The same topological approach applied to drug-resistance prediction on the Platinum benchmark produced AUROC values of 0.425 and 0.485, ruling out interface geometry as the driver. These results led me to develop TOPOLOGIX, a sequence-based predictor using ESM-2 protein language model delta-embeddings and Morgan fingerprints with a Random Forest classifier, which achieves AUROC 0.804 on the Platinum benchmark and covers 100 percent of mutations versus roughly 18 percent for structure-limited tools.

The Rowland Fellowship's open recruitment cycle and its support for researchers who might not otherwise apply aligns with my profile as an independent researcher from Nigeria. I have endorsements from Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU. I am prepared to lead a small team, manage a budget, and push the CCT program toward experimental validation.

RESEARCH STATEMENT

The Conjunctive Consolidation Threshold model addresses a specific gap in addiction neuropharmacology: no existing framework explains why certain pharmacological interventions block reward-memory reconsolidation while others merely attenuate it. The CCT model posits that reward-memory encoding requires the simultaneous crossing of thresholds on three coupled axes: dopaminergic reward prediction error, NMDAR-dependent long-term potentiation, and affective contrast. Suppression of any single axis is insufficient; suppression must be conjunctive.

I built the model as a system of coupled ODEs representing the three axes, solved with RK45 integration. The model has 14 free parameters, calibrated with Bayesian MCMC using PyMC's DEMetropolisZ sampler. Priors were elicited from a systematic screen of 1,847 records covering dopaminergic signaling, NMDAR plasticity, and affective neuroscience. The model was pre-registered with five hypotheses, H1 through H5, all confirmed. The posterior distribution shows super-additivity of 13 to 22 percentage points across model versions, indicating that the three axes interact non-additively. This is the core novel claim: the threshold is conjunctive, not additive.

The model's predictions are specific enough to test experimentally. For example, the model predicts that a sub-threshold dose of an NMDAR antagonist combined with a sub-threshold dose of a D2 modulator should block reward-memory reconsolidation, while either agent alone at the same doses should not. It also predicts a temporal window: the conjunctive suppression must occur within a defined period after memory retrieval, consistent with the reconsolidation window. These predictions are falsifiable and can be tested in rodent models of conditioned place preference or self-administration.

The Rowland Fellowship's resources are the right vehicle for this validation. The fellowship provides staff scientists and engineers who can design and build experimental apparatus. I propose to use the fellowship to establish a collaboration with an experimental lab, likely at Cambridge or a partner institution, to run the conditioned place preference experiments. My role would be to specify the pharmacological protocols from the model's parameter space, analyze the behavioral data with the same Bayesian framework used for calibration, and iterate the model based on the experimental results.

My track record includes rigorous negative results that strengthen the case for this proposal. In the hERG cardiotoxicity topology study, I pre-registered a powered replication testing whether bipartite persistent homology predicts cardiotoxicity from protein-ligand interface geometry. The result was negative: topological features achieved AUROC 0.8426 versus 0.8782 for a plain descriptor baseline. I reported this directly. In the drug-resistance topology study, the same approach produced AUROC 0.425 and 0.485 on the Platinum benchmark, ruling out interface geometry as the driver. These negative results led to TOPOLOGIX, a sequence-based predictor using ESM-2 protein language model delta-embeddings and Morgan fingerprints with a Random Forest classifier, achieving AUROC 0.804 on the Platinum benchmark. This pattern demonstrates that I follow the data, not the hypothesis.

I have also built neurocascade, a receptor-to-behavior brain-circuit simulation engine that couples pharmacokinetics, receptor binding, Wilson-Cowan circuit dynamics, and behavioral readout ODE layers. Three receptor/circuit systems are calibrated: mu-opioid, D2 dopamine, and GABA-A. All 62 tests pass. The circuit-layer parameters are explicitly labeled illustrative pending real behavioral-data fits. This engine will be the bridge between the CCT model's molecular predictions and circuit-level behavioral outcomes.

The Rowland Fellowship's selection criteria emphasize demonstrated exceptional achievements, readiness to begin an independent group, and potential to push boundaries. My three sole-authored preprints under review, my pre-registered negative results, and my multi-domain computational skills meet those criteria. The fellowship's support for high-risk research is exactly what the CCT validation requires. I am asking for the resources to test a specific, falsifiable prediction that could change how we approach addiction pharmacotherapy.

ESSAY: INTERDISCIPLINARY APPROACH

The CCT model is inherently interdisciplinary, combining pharmacology, computational neuroscience, and dynamical systems theory. My training spans all three. I hold a B.Pharm from the University of Ibadan with a CGPA of 5.1 out of 7.0, a German equivalent of 1.9, and I am a PCN-licensed pharmacist. I have worked as a clinical pharmacist at Ramset Pharmacy and as National Product Manager at Synthcare. This clinical background informs the pharmacological constraints in the CCT model: the doses, the receptor affinities, and the metabolic considerations are grounded in real drug behavior, not abstract parameters.

The computational methods come from my independent research practice. I use Python with scipy, numpy, and PyMC for ODE solving and Bayesian calibration. I use Ripser and GUDHI for topological data analysis. I use NEURON and Brian2 for neural simulation, AlphaFold and RDKit for protein and drug representation, and GROMACS and AutoDock for molecular dynamics and docking. This is not a toolkit I learned in a single lab; it is a stack I assembled across multiple independent projects, each documented in preprints and code repositories.

The dynamical systems component comes from my work on ergofluids, where I extended Koopman operator and Dynamic Mode Decomposition methods with a Mori-Zwanzig memory kernel to model macromolecular drug-vehicle transport through dense tumor tissue. The pre-registered validation pipeline passed synthetic-data gates, and the first real-data gate did not meet its primary criterion. I reported that result directly. The methods I developed there, particularly the memory kernel formalism, are directly applicable to the CCT model's temporal dynamics.

The Rowland Fellowship's interdisciplinary mandate is not a slogan for me. The CCT model requires simultaneous expertise in pharmacology, computational modeling, and experimental design. I have the first two. The fellowship's staff engineers and scientists provide the third. This is a genuine collaboration, not a formality.

EDITOR NOTES

- Eligibility risk: The Rowland Fellowship requires a PhD received or expected between May 2025 and Fall 2027. The applicant is enrolled in an M.Sc. program, not a PhD. This is a critical mismatch. Verify whether the fellowship accepts M.Sc.-level applicants with equivalent research output, or whether the applicant needs to frame this as a pre-PhD application. Do not submit without confirming this.

- The CCT model's three preprints are under review, not accepted. The letter and research statement should be updated with acceptance status if any journal accepts before submission. Do not claim peer-reviewed publication status that does not exist yet.

- The experimental validation plan is a collaboration with an unspecified Cambridge or partner lab. This is a gap. The applicant must identify a specific experimental collaborator before submission, or the proposal will read as speculative. Insert the named lab and PI in the RESEARCH STATEMENT.

- The applicant's employment history includes National Product Manager at Synthcare starting March 2026. This is a commercial role. The Rowland Fellowship expects full-time research commitment. Clarify whether the applicant will resign or reduce this role if awarded the fellowship. This must be addressed in the application or the fellowship may be rescinded.

- The profile lists endorsements from Kent Berridge, Samuel Gershman, Nathaniel Daw, and Marcelo Mattar. The application should include at least two letters of recommendation from these individuals. Confirm they have agreed to write letters and that their letters reference the CCT model specifically, not general support.

MOTIVATION LETTER

The Rowland Fellowship supports early-career researchers who want to build an independent research program with the backing of staff scientists and engineers. That structure is precisely what my research trajectory requires. I am a computational neuropharmacologist with a B.Pharm from the University of Ibadan, currently enrolled in the M.Sc. Digital Health program at the Hasso Plattner Institute / University of Potsdam. My central project, the Conjunctive Consolidation Threshold (CCT) model, is a tripartite pharmacological framework for reward-memory encoding prevention in addiction. It is ready for experimental validation, and the Rowland Fellowship's model of pairing an independent investigator with dedicated technical staff is the most direct path to that validation.

The CCT model couples three axes: dopaminergic reward prediction error, NMDAR-dependent long-term potentiation, and affective contrast. I implemented it as a system of ordinary differential equations solved with RK45, calibrated with Bayesian MCMC using PyMC's DEMetropolisZ sampler across 14 free parameters. The priors were elicited from a systematic screen of 1,847 records in the literature. All five pre-registered hypotheses, H1 through H5, were confirmed. The posterior shows super-additivity of 13 to 22 percentage points across model versions, meaning the three axes interact non-additively to suppress reward-memory consolidation. Three sole-authored preprints are under review at peer-reviewed journals: International Addiction Review and Therapeutics, Progress in Neuro-Psychopharmacology and Biological Psychiatry, and Neuroscience and Biobehavioral Reviews. A co-authored paper is under review at Alcohol (Elsevier).

The fellowship's emphasis on high-risk, interdisciplinary research matches the CCT program's next phase. The model makes specific, testable predictions about which pharmacological combinations should block reconsolidation of reward memories. I want to use the Rowland Fellowship to design and run the experimental validation of those predictions. The Rowland staff engineers can help build in-vitro or in-vivo setups that I cannot construct as an independent researcher. My computational skills, including ODE modeling, Bayesian calibration, and machine learning, are the tools needed to drive such a project from design through analysis.

I have also done rigorous negative-result work that demonstrates my commitment to falsification. In a pre-registered, powered replication study, I tested 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. I reported this result directly. The same topological approach applied to drug-resistance prediction on the Platinum benchmark produced AUROC values of 0.425 and 0.485, ruling out interface geometry as the driver. These results led me to develop TOPOLOGIX, a sequence-based predictor using ESM-2 protein language model delta-embeddings and Morgan fingerprints with a Random Forest classifier, which achieves AUROC 0.804 on the Platinum benchmark and covers 100 percent of mutations versus roughly 18 percent for structure-limited tools.

The Rowland Fellowship's open recruitment cycle and its support for researchers who might not otherwise apply aligns with my profile as an independent researcher from Nigeria. I have endorsements from Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU. I am prepared to lead a small team, manage a budget, and push the CCT program toward experimental validation.

RESEARCH STATEMENT

The Conjunctive Consolidation Threshold model addresses a specific gap in addiction neuropharmacology: no existing framework explains why certain pharmacological interventions block reward-memory reconsolidation while others merely attenuate it. The CCT model posits that reward-memory encoding requires the simultaneous crossing of thresholds on three coupled axes: dopaminergic reward prediction error, NMDAR-dependent long-term potentiation, and affective contrast. Suppression of any single axis is insufficient; suppression must be conjunctive.

I built the model as a system of coupled ODEs representing the three axes, solved with RK45 integration. The model has 14 free parameters, calibrated with Bayesian MCMC using PyMC's DEMetropolisZ sampler. Priors were elicited from a systematic screen of 1,847 records covering dopaminergic signaling, NMDAR plasticity, and affective neuroscience. The model was pre-registered with five hypotheses, H1 through H5, all confirmed. The posterior distribution shows super-additivity of 13 to 22 percentage points across model versions, indicating that the three axes interact non-additively. This is the core novel claim: the threshold is conjunctive, not additive.

The model's predictions are specific enough to test experimentally. For example, the model predicts that a sub-threshold dose of an NMDAR antagonist combined with a sub-threshold dose of a D2 modulator should block reward-memory reconsolidation, while either agent alone at the same doses should not. It also predicts a temporal window: the conjunctive suppression must occur within a defined period after memory retrieval, consistent with the reconsolidation window. These predictions are falsifiable and can be tested in rodent models of conditioned place preference or self-administration.

The Rowland Fellowship's resources are the right vehicle for this validation. The fellowship provides staff scientists and engineers who can design and build experimental apparatus. I propose to use the fellowship to establish a collaboration with an experimental lab, likely at Cambridge or a partner institution, to run the conditioned place preference experiments. My role would be to specify the pharmacological protocols from the model's parameter space, analyze the behavioral data with the same Bayesian framework used for calibration, and iterate the model based on the experimental results.

My track record includes rigorous negative results that strengthen the case for this proposal. In the hERG cardiotoxicity topology study, I pre-registered a powered replication testing whether bipartite persistent homology predicts cardiotoxicity from protein-ligand interface geometry. The result was negative: topological features achieved AUROC 0.8426 versus 0.8782 for a plain descriptor baseline. I reported this directly. In the drug-resistance topology study, the same approach produced AUROC 0.425 and 0.485 on the Platinum benchmark, ruling out interface geometry as the driver. These negative results led to TOPOLOGIX, a sequence-based predictor using ESM-2 protein language model delta-embeddings and Morgan fingerprints with a Random Forest classifier, achieving AUROC 0.804 on the Platinum benchmark. This pattern demonstrates that I follow the data, not the hypothesis.

I have also built neurocascade, a receptor-to-behavior brain-circuit simulation engine that couples pharmacokinetics, receptor binding, Wilson-Cowan circuit dynamics, and behavioral readout ODE layers. Three receptor/circuit systems are calibrated: mu-opioid, D2 dopamine, and GABA-A. All 62 tests pass. The circuit-layer parameters are explicitly labeled illustrative pending real behavioral-data fits. This engine will be the bridge between the CCT model's molecular predictions and circuit-level behavioral outcomes.

The Rowland Fellowship's selection criteria emphasize demonstrated exceptional achievements, readiness to begin an independent group, and potential to push boundaries. My three sole-authored preprints under review, my pre-registered negative results, and my multi-domain computational skills meet those criteria. The fellowship's support for high-risk research is exactly what the CCT validation requires. I am asking for the resources to test a specific, falsifiable prediction that could change how we approach addiction pharmacotherapy.

ESSAY: INTERDISCIPLINARY APPROACH

The CCT model is inherently interdisciplinary, combining pharmacology, computational neuroscience, and dynamical systems theory. My training spans all three. I hold a B.Pharm from the University of Ibadan with a CGPA of 5.1 out of 7.0, a German equivalent of 1.9, and I am a PCN-licensed pharmacist. I have worked as a clinical pharmacist at Ramset Pharmacy and as National Product Manager at Synthcare. This clinical background informs the pharmacological constraints in the CCT model: the doses, the receptor affinities, and the metabolic considerations are grounded in real drug behavior, not abstract parameters.

The computational methods come from my independent research practice. I use Python with scipy, numpy, and PyMC for ODE solving and Bayesian calibration. I use Ripser and GUDHI for topological data analysis. I use NEURON and Brian2 for neural simulation, AlphaFold and RDKit for protein and drug representation, and GROMACS and AutoDock for molecular dynamics and docking. This is not a toolkit I learned in a single lab; it is a stack I assembled across multiple independent projects, each documented in preprints and code repositories.

The dynamical systems component comes from my work on ergofluids, where I extended Koopman operator and Dynamic Mode Decomposition methods with a Mori-Zwanzig memory kernel to model macromolecular drug-vehicle transport through dense tumor tissue. The pre-registered validation pipeline passed synthetic-data gates, and the first real-data gate did not meet its primary criterion. I reported that result directly. The methods I developed there, particularly the memory kernel formalism, are directly applicable to the CCT model's temporal dynamics.

The Rowland Fellowship's interdisciplinary mandate is not a slogan for me. The CCT model requires simultaneous expertise in pharmacology, computational modeling, and experimental design. I have the first two. The fellowship's staff engineers and scientists provide the third. This is a genuine collaboration, not a formality.

EDITOR NOTES

- Eligibility risk: The Rowland Fellowship requires a PhD received or expected between May 2025 and Fall 2027. The applicant is enrolled in an M.Sc. program, not a PhD. This is a critical mismatch. Verify whether the fellowship accepts M.Sc.-level applicants with equivalent research output, or whether the applicant needs to frame this as a pre-PhD application. Do not submit without confirming this.

- The CCT model's three preprints are under review, not accepted. The letter and research statement should be updated with acceptance status if any journal accepts before submission. Do not claim peer-reviewed publication status that does not exist yet.

- The experimental validation plan is a collaboration with an unspecified Cambridge or partner lab. This is a gap. The applicant must identify a specific experimental collaborator before submission, or the proposal will read as speculative. Insert the named lab and PI in the RESEARCH STATEMENT.

- The applicant's employment history includes National Product Manager at Synthcare starting March 2026. This is a commercial role. The Rowland Fellowship expects full-time research commitment. Clarify whether the applicant will resign or reduce this role if awarded the fellowship. This must be addressed in the application or the fellowship may be rescinded.

- The profile lists endorsements from Kent Berridge, Samuel Gershman, Nathaniel Daw, and Marcelo Mattar. The application should include at least two letters of recommendation from these individuals. Confirm they have agreed to write letters and that their letters reference the CCT model specifically, not general support.

AI Draft — Rowland Fellowship