MOTIVATION LETTER The Conjunctive Consolidation Threshold model proposes a tripartite pharmacological framework for preventing reward-memory encoding in addiction. Three coupled axes—dopaminergic reward prediction error, NMDAR-dependent long-term potentiation, and affective contrast—form an ordinary differential equation system solved with RK45. Bayesian MCMC calibration using PyMC DEMetropolisZ with 14 free parameters and literature-elicited priors from an 1,847-record screen confirmed all five pre-registered hypotheses. Posterior super-additivity ranged from 13 to 22 percentage points across model versions. Three sole-authored preprints are deposited on OSF and Zenodo. A co-authored paper is under review at Alcohol (Elsevier). This application to the ResearchHub Preregistration Grant seeks funding to conduct a preregistered computational replication and extension of the CCT model. The replication will use an independent Bayesian calibration pipeline with refined priors drawn from a broader literature screen. The extension will add a sensitivity analysis across parameter regimes and a formal model comparison against a simpler two-factor alternative. All code, data, and analysis scripts are already public on GitHub and Zenodo. The preregistration will be deposited on OSF before any new analyses begin. The ResearchHub Foundation’s mission—open, community-driven science with no institutional gatekeeping—aligns directly with my independent research practice. I hold no faculty position. I am a licensed pharmacist, an enrolled M.Sc. Digital Health student at Hasso Plattner Institute, and an independent computational researcher. My work is funded out of pocket. A grant of 8,000 USD would cover 400 hours of HPC compute time on a cloud cluster for the MCMC sampling, publication fees for an open-access preprint server, and a stipend for two months of full-time work. No institutional overhead applies. Overhead is capped at 10 percent and I have no institution to charge it. Addiction is a growing burden in Nigeria and across sub-Saharan Africa. Pharmacological interventions remain limited. The CCT model offers a mechanistic framework for designing combination therapies that prevent memory consolidation without blocking acute reward. A validated, open-source computational model can guide preclinical experiments at a fraction of the cost of wet-lab screening. This grant would accelerate that validation. The preregistration will be posted on the ResearchHub platform for community review and crowdfunding. I will respond to all comments and revise the analysis plan if warranted. Results will be shared as a preprint with a DOI, and all data and code will remain open. RESEARCH STATEMENT The CCT model addresses a specific gap in addiction neuroscience: no existing pharmacological framework predicts which drug combinations can prevent reward-memory consolidation while preserving natural reward processing. Existing models focus on either dopamine signaling or NMDA receptor function in isolation. The CCT model couples three axes—dopaminergic reward prediction error, NMDAR-dependent long-term potentiation, and affective contrast—into a single ODE system. This coupling allows the model to predict super-additive effects of combination therapies that target multiple axes simultaneously. The proposed project has three aims. First, replicate the original Bayesian calibration using an independent MCMC chain with 100,000 draws and 50,000 warmup iterations. The original calibration used 14 free parameters with priors from an 1,847-record literature screen. The replication will use the same priors but a different random seed and a longer chain to assess convergence stability. Second, conduct a global sensitivity analysis using Sobol indices to identify which parameters drive the super-additive predictions. Third, compare the CCT model against a simpler two-factor model that omits the affective contrast axis. The comparison will use leave-one-out cross-validation on simulated data from the original parameter posterior. The preregistration will specify all analysis steps, including exclusion criteria, stopping rules, and the exact metrics for model comparison. No exploratory analyses will be conducted. If the replication fails to confirm the original results, that will be reported directly. The project is designed to produce a definitive answer about the CCT model’s strength, not to confirm a preferred outcome. Feasibility is high. I have already built and calibrated the CCT model once. The codebase is modular, documented, and version-controlled. The MCMC pipeline uses PyMC with DEMetropolisZ, which I have run on HPC clusters for previous projects. The sensitivity analysis uses SALib, which I have used for the neurocascade project. The timeline is three months: one month for the replication calibration, one month for the sensitivity analysis and model comparison, and one month for writing and depositing the preprint. BUDGET HPC compute time: 4,000 USD. 400 hours on a cloud cluster with 32 cores and 128 GB RAM. Sufficient for 100,000 MCMC draws with 50,000 warmup iterations across four chains. Open-access publication fee: 1,500 USD. Covers preprint DOI registration and journal submission fee if the paper is accepted. Stipend: 2,500 USD. Two months of full-time work at 1,250 USD per month. Below the average salary for a computational researcher in Germany. Total: 8,000 USD. No overhead. No equipment costs. No travel costs. TIMELINE Month 1: Replication calibration. Run MCMC chains. Assess convergence with R-hat and effective sample size. Compare posterior distributions to original. Month 2: Sensitivity analysis and model comparison. Compute Sobol indices. Fit two-factor model. Compute leave-one-out cross-validation scores. Month 3: Write preprint. Deposit on OSF and Zenodo. Submit to a peer-reviewed journal. Respond to ResearchHub community comments. OPEN SCIENCE COMMITMENT All code is public on GitHub under a BSD-3 license. All data are public on Zenodo with a DOI. The preregistration will be deposited on OSF before any new analyses begin. The preprint will be posted on a public server. No results will be withheld regardless of outcome. The ResearchHub platform will host the preregistration and the final report. CHECKLIST - [ ] ORCID account verified and linked to ResearchHub profile - [ ] Identity verification submitted to ResearchHub Foundation - [ ] Preregistration document written (2-4 pages) and uploaded to ResearchHub platform - [ ] Budget table included in preregistration - [ ] Timeline included in preregistration - [ ] Open science statement included in preregistration - [ ] GitHub repository URL included in application - [ ] Zenodo DOI for original CCT preprints included in application - [ ] ORCID iD included in application - [ ] Personal website (zyco.org) included in application - [ ] Proof of enrollment at Hasso Plattner Institute (M.Sc. Digital Health) attached - [ ] Proof of PCN pharmacist license attached EDITOR NOTES - Eligibility risk: ResearchHub Preregistration Grants are open to independent researchers, but the applicant is also enrolled in an M.Sc. program. Confirm that enrollment does not disqualify the applicant as an independent researcher. The programme language says "no institutional affiliation required," but enrollment may be interpreted as affiliation. Clarify with the foundation before submitting. - Fact verification: The budget states 400 hours of HPC compute at 4,000 USD. Verify that this rate (10 USD per hour) is realistic for a cloud cluster. The applicant may need to specify the provider (e.g., AWS, Google Cloud, or a university cluster) and confirm the pricing. - Gap: The applicant's profile lists a co-authored paper under review at Alcohol (Elsevier). The preregistration should cite this paper and explain how the proposed replication extends beyond the submitted manuscript. If the paper is accepted before the grant is awarded, the replication may need to be reframed as a follow-up study. - Personal detail needed: The applicant should insert a brief statement about their connection to Nigeria and the addiction burden in sub-Saharan Africa. The profile mentions this angle but the draft does not include a specific statistic or reference. A single sentence with a concrete number (e.g., "Nigeria has an estimated 3 million people with substance use disorders, according to the 2023 WHO report") would strengthen the LMIC framing. - Timeline risk: The three-month timeline assumes no interruptions from the applicant's current employment as National Product Manager at Synthcare. The applicant should confirm that they can dedicate two months of full-time work to this project while employed. If not, the timeline should be extended to six months with a reduced stipend.