Computational drug discovery under auditable governance
DrugSynthAI is the computational drug discovery research platform of the clinician scientist research and development program. It applies machine learning models to target identification, protein ligand modeling, ADMET prediction, and generative chemistry, within a governance layer that separates learning authority from decision authority.
Scientific context
Contemporary drug discovery underwent a methodological transformation over the past decade. Protein language models, graph neural networks applied to small molecules, diffusion architectures for de novo ligand generation, and multi agent systems for virtual screening reshaped the cost, scale, and nature of therapeutic hypothesis generation.
DrugSynthAI operates at the intersection of these computational methods and the epistemological rigor required by translational medicine. The starting point is not a demonstration that a model can propose a molecule, but the prior question: under what governance conditions can an AI system participate in the generation of therapeutic hypotheses so that its proposals remain auditable, reproducible, and explainable against the ethical and regulatory standards of biomedical research.
A recent translational review from the program addresses this in detail: AI in drug discovery: from target to clinical signal.
Platform scope
Research and development modules, with distinct maturity stages. No module is authorized for clinical or regulatory use.
| Module | Scope | Methodological references |
|---|---|---|
| Target identification | Integrated analysis of omics data and literature for hierarchizing candidate molecular targets by pathophysiological relevance. | Vamathevan et al., Nat Rev Drug Discov 2019 |
| Structural modeling | Prediction of protein and protein ligand complex structures at atomic resolution to guide virtual screening. | Jumper et al., Nature 2021; Abramson et al., Nature 2024 |
| Generative chemistry | Conditioned generation of small molecules with pre specified pharmacokinetic properties via diffusion architectures. | Watson et al., Nature 2023 |
| ADMET prediction | Predictive models of absorption, distribution, metabolism, excretion, and toxicity over candidate compounds. | Peer reviewed pharmacological ML literature |
| Antimicrobial screening | Application of deep neural networks for identification of novel antimicrobial scaffolds, with in vitro validation required. | Stokes et al., Cell 2020 (halicin) |
| Repositioning | Systematic mining of literature and knowledge bases for therapeutic repositioning hypotheses. | Xu, Ren et al., Nat Med 2025 (rentosertib) |
Governance layer · AIDD-GOV
DrugSynthAI operates under an AI assisted drug discovery governance standard called AIDD-GOV. The central architectural principle is the separation between learning authority and decision authority: a model can classify, predict, and propose, but the decision to synthesize a compound, prioritize a target, or advance to experimental characterization remains under explicit, traceable, and reversible human review.
This governance layer is a specific implementation of the broader Externally Governed Learning Systems (EGLS) standard, subject of USPTO provisional patent 63/975,551 from the same research program. AIDD-GOV adapts the EGLS principles to the drug discovery lifecycle, with particular attention to training data provenance, traceability of model generated suggestions, and pre established rejection criteria that are independent of computational confidence.
Governance is not a downstream regulatory envelope. It is an architectural condition prior to predictive validation.
Intellectual property
Publications and academic identifiers
Related academic output and scientific identity profiles
Translational review on academic blog
AI in drug discovery: from target to clinical signal · June 2026
Scientific identity profiles
Development state
PRE SEED RESEARCH AND DEVELOPMENT PHASE
DrugSynthAI is in pre seed research and development phase. The platform is not authorized for clinical, regulatory, or commercial use. No suggestion generated by the modules should be interpreted as a therapeutic recommendation.
The objective of the current stage is technical characterization and documentation of the AIDD-GOV governance architecture, with associated academic output and USPTO intellectual property filings in progress.
Institutional research collaborations are considered case by case. Contact via academic contact page.