Draft — NLnet; NGI Zero Discovery

MOTIVATION LETTER

The NGI Zero Discovery programme funds open infrastructure that puts users in control of their technology. The Conjunctive Consolidation Threshold model, or CCT model, is an open-source computational pharmacology framework that does exactly that for addiction neuroscience. It replaces proprietary, closed-source simulation pipelines with a modular, fully free software stack that any researcher can run, inspect, modify, and redistribute. The CCT model is a tripartite ordinary differential equation system coupling dopaminergic reward prediction error, NMDA-receptor-dependent long-term potentiation, and affective contrast to predict whether a pharmacological intervention can prevent reward-memory encoding. The model has 14 free parameters, was calibrated with Bayesian Markov chain Monte Carlo using literature-elicited priors from an 1,847-record screen, and confirmed all five pre-registered hypotheses with posterior super-additivity of 13 to 22 percentage points across model versions. Three sole-authored preprints are under peer review at the International Journal of Addiction Research and Therapy, Progress in Neuro-Psychopharmacology and Biological Psychiatry, and Neuroscience and Biobehavioral Reviews. A co-authored paper is under review at Alcohol, Elsevier.

The software is written entirely in Python using scipy, numpy, PyMC, and RK45 integrators. It is released under a GNU General Public License v3. The code, documentation, and all pre-registered analysis plans are deposited on Zenodo and GitHub. The architecture is modular: the ODE solver, the Bayesian calibration engine, and the hypothesis-testing pipeline are separate, documented components that can be reused for any three-axis dynamical system. This is infrastructure for discovery, not a one-off simulation.

NGI Zero Discovery funds projects that empower users and decentralize control over technology. The CCT model does this by making pharmacological mechanism discovery accessible to any lab, anywhere, without requiring expensive software licenses or proprietary toolchains. A researcher in Nigeria, India, or Brazil can clone the repository, run the calibration on a standard laptop, and test their own hypotheses about reward-memory encoding. That is direct alignment with the programme's mission.

The budget request is 45,000 euros. This covers 12 months of development time for the applicant at 3,000 euros per month, plus 9,000 euros for cloud compute credits, publication fees for open-access journals, and travel to one international conference to present the software. No equipment, no overhead, no institutional indirect costs. The applicant is an independent researcher with no institutional salary support, so the funding goes entirely to the work.

The CCT model is open source, modular, and designed for community reuse. It enables discovery of pharmacological mechanisms in addiction neuroscience without proprietary dependencies. It aligns with NGI Zero's values of security, privacy, accessibility, and standardization. The applicant has a track record of open-source contributions, peer-reviewed research output, and independent project delivery. This is a frugal, feasible, high-impact proposal.

SHORT ESSAY: RELEVANCE TO SEARCH AND DISCOVERY INFRASTRUCTURE

NGI Zero Discovery defines search and discovery infrastructure broadly, including tools that enable discovery of data, knowledge, or services. The CCT model is a tool for discovering pharmacological mechanisms. It does not index web pages, but it does index the parameter space of a dynamical system to find regions where a drug prevents reward-memory encoding. That is a search problem: given a high-dimensional parameter space, find the subspace where a pre-registered hypothesis holds. The Bayesian MCMC calibration engine performs this search by sampling from the posterior distribution, and the hypothesis-testing pipeline evaluates each sample against the five pre-registered criteria. The output is a ranked list of parameter combinations that satisfy the hypotheses, analogous to a search engine returning relevant results.

The software is designed for reuse. The ODE solver, the calibration engine, and the testing pipeline are separate modules with documented application programming interfaces. A researcher studying a different three-axis dynamical system, such as a predator-prey model or a gene regulatory network, can replace the CCT equations with their own and reuse the calibration and testing infrastructure. This makes the software a general-purpose discovery platform, not a single-purpose simulation.

The applicant has also built four independent DuckDB-based ingest-to-analyze corpus and retrieval-augmented generation pipelines across life sciences, technology and artificial intelligence and security, and social science domains. These pipelines are open source and designed for community reuse. They enable discovery of knowledge from large, heterogeneous text corpora. This directly matches the programme's focus on search and discovery infrastructure.

SHORT ESSAY: TECHNICAL MERIT AND INNOVATION

The CCT model is the first tripartite pharmacological framework that couples dopaminergic reward prediction error, NMDA-receptor-dependent long-term potentiation, and affective contrast in a single dynamical system. Existing models treat these axes separately. The innovation is the coupling: the model predicts how a drug that modulates one axis affects the others, and whether the combined effect prevents reward-memory encoding. The Bayesian calibration with literature-elicited priors from an 1,847-record screen is a methodological innovation that reduces overfitting and increases interpretability. The pre-registration of all five hypotheses and the confirmation of all five with posterior super-additivity of 13 to 22 percentage points is a rigorous validation that is rare in computational pharmacology.

The applicant has also developed neurocascade, a receptor-to-behavior brain-circuit simulation engine that couples pharmacokinetics to receptor-binding to Wilson-Cowan circuit dynamics to behavioral-readout ODE layers. This is a modular, open-source platform for simulating the full chain from drug administration to behavior. The applicant has also developed TOPOLOGIX, a protein-language-model-based predictor of drug-resistance mutations that achieves an area under the receiver operating characteristic curve of 0.804 on the Platinum benchmark, beating structure-based baselines while covering 100 percent of mutations versus approximately 18 percent for structure-limited tools. These are technically innovative, rigorously validated, and openly released.

SHORT ESSAY: FEASIBILITY AND BUDGET

The applicant has delivered all previous research projects independently, on time, and with open-source releases. The CCT model is already built, calibrated, and validated. The three preprints are under review. The software is already released under GNU General Public License v3. The work proposed here is to extend the software with a graphical user interface for non-programmer researchers, write thorough documentation and tutorials, and run a community beta-testing programme with three collaborating labs. The applicant has the technical skills to do this: Python, JavaScript, Node.js, and production systems operations including Linux virtual private server administration, systemd, Caddy TLS, continuous integration and continuous deployment, and automated backup and disaster-recovery.

The budget is 45,000 euros. The applicant requests 3,000 euros per month for 12 months of full-time development work, totaling 36,000 euros. The remaining 9,000 euros cover cloud compute credits for running the calibration engine on large parameter spaces, open-access publication fees for two journal articles describing the software, and travel to one international conference, such as the Organization for Computational Neurosciences annual meeting, to present the software and recruit beta testers. No equipment, no overhead, no institutional indirect costs. The applicant is an independent researcher with no institutional salary support, so the funding goes entirely to the work.

SHORT ESSAY: APPLICANT CAPABILITY AND TRACK RECORD

The applicant is a licensed pharmacist with a Bachelor of Pharmacy from the University of Ibadan, Nigeria, and is enrolled in the Master of Science in Digital Health at the Hasso Plattner Institute and University of Potsdam, Germany. The applicant has published three sole-authored preprints under peer review, one co-authored paper under review at Alcohol, Elsevier, and has delivered four independent open-source research projects: the CCT model, neurocascade, TOPOLOGIX, and ergofluids. The applicant has endorsements from Kent Berridge at the University of Michigan, Samuel Gershman at Harvard University, Nathaniel Daw at Princeton University, and Marcelo Mattar at New York University. The applicant has technical skills across Python, R, topological data analysis, neural simulation, molecular dynamics, and production systems operations. The applicant has built four independent DuckDB-based ingest-to-analyze corpus and retrieval-augmented generation pipelines and a self-hosted local large language model serving infrastructure. The applicant has a track record of independent, rigorous, open-source research delivery.

CHECKLIST

- [ ] Motivation letter, 300-500 words, written as specified
- [ ] Short essay on relevance to search and discovery infrastructure, 200-350 words
- [ ] Short essay on technical merit and innovation, 200-350 words
- [ ] Short essay on feasibility and budget, 200-350 words
- [ ] Short essay on applicant capability and track record, 200-350 words
- [ ] Budget table: 36,000 euros for 12 months salary at 3,000 euros per month, 6,000 euros for cloud compute, 2,000 euros for open-access publication fees, 1,000 euros for conference travel, total 45,000 euros
- [ ] Curriculum vitae with ORCID, GitHub, personal site, publications, and open-source projects
- [ ] Proof of enrollment in M.Sc. Digital Health at Hasso Plattner Institute and University of Potsdam
- [ ] Two letters of recommendation: one from Kent Berridge or Samuel Gershman, one from Nathaniel Daw or Marcelo Mattar
- [ ] Open-source license file for all software repositories: GNU General Public License v3
- [ ] Link to Zenodo deposit with all pre-registered analysis plans and data

EDITOR NOTES

- Eligibility risk: The applicant is enrolled in a master's programme starting winter semester 2026/27. Confirm that NGI Zero Discovery allows applicants who are currently enrolled students. The programme typically funds independent researchers and small organizations, but student status may be acceptable if the applicant is not receiving institutional salary support. The applicant should clarify in the application that the master's programme is part-time and does not provide salary or stipend.
- Facts to verify: The applicant's three preprints are listed as under review at specific journals. Confirm the current status of each review. If any have been accepted or rejected, update the application accordingly. The co-authored paper at Alcohol, Elsevier should also be confirmed.
- Gaps to fill: The applicant's profile does not specify the names of the three collaborating labs for the beta-testing programme. The applicant should insert the names of specific labs that have agreed to participate, or state that discussions are ongoing. The applicant should also specify which international conference they plan to attend and confirm that the budgeted 1,000 euros is sufficient for travel from Germany.

MOTIVATION LETTER

The NGI Zero Discovery programme funds open infrastructure that puts users in control of their technology. The Conjunctive Consolidation Threshold model, or CCT model, is an open-source computational pharmacology framework that does exactly that for addiction neuroscience. It replaces proprietary, closed-source simulation pipelines with a modular, fully free software stack that any researcher can run, inspect, modify, and redistribute. The CCT model is a tripartite ordinary differential equation system coupling dopaminergic reward prediction error, NMDA-receptor-dependent long-term potentiation, and affective contrast to predict whether a pharmacological intervention can prevent reward-memory encoding. The model has 14 free parameters, was calibrated with Bayesian Markov chain Monte Carlo using literature-elicited priors from an 1,847-record screen, and confirmed all five pre-registered hypotheses with posterior super-additivity of 13 to 22 percentage points across model versions. Three sole-authored preprints are under peer review at the International Journal of Addiction Research and Therapy, Progress in Neuro-Psychopharmacology and Biological Psychiatry, and Neuroscience and Biobehavioral Reviews. A co-authored paper is under review at Alcohol, Elsevier.

The software is written entirely in Python using scipy, numpy, PyMC, and RK45 integrators. It is released under a GNU General Public License v3. The code, documentation, and all pre-registered analysis plans are deposited on Zenodo and GitHub. The architecture is modular: the ODE solver, the Bayesian calibration engine, and the hypothesis-testing pipeline are separate, documented components that can be reused for any three-axis dynamical system. This is infrastructure for discovery, not a one-off simulation.

NGI Zero Discovery funds projects that empower users and decentralize control over technology. The CCT model does this by making pharmacological mechanism discovery accessible to any lab, anywhere, without requiring expensive software licenses or proprietary toolchains. A researcher in Nigeria, India, or Brazil can clone the repository, run the calibration on a standard laptop, and test their own hypotheses about reward-memory encoding. That is direct alignment with the programme's mission.

The budget request is 45,000 euros. This covers 12 months of development time for the applicant at 3,000 euros per month, plus 9,000 euros for cloud compute credits, publication fees for open-access journals, and travel to one international conference to present the software. No equipment, no overhead, no institutional indirect costs. The applicant is an independent researcher with no institutional salary support, so the funding goes entirely to the work.

The CCT model is open source, modular, and designed for community reuse. It enables discovery of pharmacological mechanisms in addiction neuroscience without proprietary dependencies. It aligns with NGI Zero's values of security, privacy, accessibility, and standardization. The applicant has a track record of open-source contributions, peer-reviewed research output, and independent project delivery. This is a frugal, feasible, high-impact proposal.

SHORT ESSAY: RELEVANCE TO SEARCH AND DISCOVERY INFRASTRUCTURE

NGI Zero Discovery defines search and discovery infrastructure broadly, including tools that enable discovery of data, knowledge, or services. The CCT model is a tool for discovering pharmacological mechanisms. It does not index web pages, but it does index the parameter space of a dynamical system to find regions where a drug prevents reward-memory encoding. That is a search problem: given a high-dimensional parameter space, find the subspace where a pre-registered hypothesis holds. The Bayesian MCMC calibration engine performs this search by sampling from the posterior distribution, and the hypothesis-testing pipeline evaluates each sample against the five pre-registered criteria. The output is a ranked list of parameter combinations that satisfy the hypotheses, analogous to a search engine returning relevant results.

The software is designed for reuse. The ODE solver, the calibration engine, and the testing pipeline are separate modules with documented application programming interfaces. A researcher studying a different three-axis dynamical system, such as a predator-prey model or a gene regulatory network, can replace the CCT equations with their own and reuse the calibration and testing infrastructure. This makes the software a general-purpose discovery platform, not a single-purpose simulation.

The applicant has also built four independent DuckDB-based ingest-to-analyze corpus and retrieval-augmented generation pipelines across life sciences, technology and artificial intelligence and security, and social science domains. These pipelines are open source and designed for community reuse. They enable discovery of knowledge from large, heterogeneous text corpora. This directly matches the programme's focus on search and discovery infrastructure.

SHORT ESSAY: TECHNICAL MERIT AND INNOVATION

The CCT model is the first tripartite pharmacological framework that couples dopaminergic reward prediction error, NMDA-receptor-dependent long-term potentiation, and affective contrast in a single dynamical system. Existing models treat these axes separately. The innovation is the coupling: the model predicts how a drug that modulates one axis affects the others, and whether the combined effect prevents reward-memory encoding. The Bayesian calibration with literature-elicited priors from an 1,847-record screen is a methodological innovation that reduces overfitting and increases interpretability. The pre-registration of all five hypotheses and the confirmation of all five with posterior super-additivity of 13 to 22 percentage points is a rigorous validation that is rare in computational pharmacology.

The applicant has also developed neurocascade, a receptor-to-behavior brain-circuit simulation engine that couples pharmacokinetics to receptor-binding to Wilson-Cowan circuit dynamics to behavioral-readout ODE layers. This is a modular, open-source platform for simulating the full chain from drug administration to behavior. The applicant has also developed TOPOLOGIX, a protein-language-model-based predictor of drug-resistance mutations that achieves an area under the receiver operating characteristic curve of 0.804 on the Platinum benchmark, beating structure-based baselines while covering 100 percent of mutations versus approximately 18 percent for structure-limited tools. These are technically innovative, rigorously validated, and openly released.

SHORT ESSAY: FEASIBILITY AND BUDGET

The applicant has delivered all previous research projects independently, on time, and with open-source releases. The CCT model is already built, calibrated, and validated. The three preprints are under review. The software is already released under GNU General Public License v3. The work proposed here is to extend the software with a graphical user interface for non-programmer researchers, write thorough documentation and tutorials, and run a community beta-testing programme with three collaborating labs. The applicant has the technical skills to do this: Python, JavaScript, Node.js, and production systems operations including Linux virtual private server administration, systemd, Caddy TLS, continuous integration and continuous deployment, and automated backup and disaster-recovery.

The budget is 45,000 euros. The applicant requests 3,000 euros per month for 12 months of full-time development work, totaling 36,000 euros. The remaining 9,000 euros cover cloud compute credits for running the calibration engine on large parameter spaces, open-access publication fees for two journal articles describing the software, and travel to one international conference, such as the Organization for Computational Neurosciences annual meeting, to present the software and recruit beta testers. No equipment, no overhead, no institutional indirect costs. The applicant is an independent researcher with no institutional salary support, so the funding goes entirely to the work.

SHORT ESSAY: APPLICANT CAPABILITY AND TRACK RECORD

The applicant is a licensed pharmacist with a Bachelor of Pharmacy from the University of Ibadan, Nigeria, and is enrolled in the Master of Science in Digital Health at the Hasso Plattner Institute and University of Potsdam, Germany. The applicant has published three sole-authored preprints under peer review, one co-authored paper under review at Alcohol, Elsevier, and has delivered four independent open-source research projects: the CCT model, neurocascade, TOPOLOGIX, and ergofluids. The applicant has endorsements from Kent Berridge at the University of Michigan, Samuel Gershman at Harvard University, Nathaniel Daw at Princeton University, and Marcelo Mattar at New York University. The applicant has technical skills across Python, R, topological data analysis, neural simulation, molecular dynamics, and production systems operations. The applicant has built four independent DuckDB-based ingest-to-analyze corpus and retrieval-augmented generation pipelines and a self-hosted local large language model serving infrastructure. The applicant has a track record of independent, rigorous, open-source research delivery.

CHECKLIST

- [ ] Motivation letter, 300-500 words, written as specified
- [ ] Short essay on relevance to search and discovery infrastructure, 200-350 words
- [ ] Short essay on technical merit and innovation, 200-350 words
- [ ] Short essay on feasibility and budget, 200-350 words
- [ ] Short essay on applicant capability and track record, 200-350 words
- [ ] Budget table: 36,000 euros for 12 months salary at 3,000 euros per month, 6,000 euros for cloud compute, 2,000 euros for open-access publication fees, 1,000 euros for conference travel, total 45,000 euros
- [ ] Curriculum vitae with ORCID, GitHub, personal site, publications, and open-source projects
- [ ] Proof of enrollment in M.Sc. Digital Health at Hasso Plattner Institute and University of Potsdam
- [ ] Two letters of recommendation: one from Kent Berridge or Samuel Gershman, one from Nathaniel Daw or Marcelo Mattar
- [ ] Open-source license file for all software repositories: GNU General Public License v3
- [ ] Link to Zenodo deposit with all pre-registered analysis plans and data

EDITOR NOTES

- Eligibility risk: The applicant is enrolled in a master's programme starting winter semester 2026/27. Confirm that NGI Zero Discovery allows applicants who are currently enrolled students. The programme typically funds independent researchers and small organizations, but student status may be acceptable if the applicant is not receiving institutional salary support. The applicant should clarify in the application that the master's programme is part-time and does not provide salary or stipend.
- Facts to verify: The applicant's three preprints are listed as under review at specific journals. Confirm the current status of each review. If any have been accepted or rejected, update the application accordingly. The co-authored paper at Alcohol, Elsevier should also be confirmed.
- Gaps to fill: The applicant's profile does not specify the names of the three collaborating labs for the beta-testing programme. The applicant should insert the names of specific labs that have agreed to participate, or state that discussions are ongoing. The applicant should also specify which international conference they plan to attend and confirm that the budgeted 1,000 euros is sufficient for travel from Germany.

AI Draft — NLnet; NGI Zero Discovery