Top Drug Candidates
COMPUTATIONALThis is a unified, computationally ranked list of candidate compounds drawn from ChEMBL bioactivity data, cross-disease repurposing libraries, and DiffDock virtual binding against SMA target pockets. Every entry is a computational prediction — a compound that passed in-silico filters and scored well on an integrated ranking. None of these candidates has been tested in a wet lab; ranking high here is a hypothesis to test, not a measured result.
Each candidate passes through a 6-stage computational filtering pipeline: (1) Drug-likeness — Lipinski, QED, PAINS substructure filters, BBB/CNS-MPO estimates; (2) Structural — DiffDock pose prediction against SMA target binding pockets; (3) Analog search — ChEMBL SAR neighbours; (4) ADMET prediction — rule-based absorption/distribution/metabolism/excretion/toxicity estimates; (5) Literature scan — automated PubMed retrieval; (6) Suggested assays — proposed experimental designs for any future wet-lab follow-up. All six stages are in-silico; none constitutes experimental validation.
Note: ADMET predictions use rule-based heuristics (Lipinski Rule of 5, TPSA-based BBB estimate, QED score, PAINS substructure filters). These are computational filtering tools, not validated pharmacokinetic or toxicology models.
▶How does Candidate Scoring work?
Candidates are scored 0–1 (integrated computational score) and binned into tiers: Tier A (≥ 0.6) — highest computational ranking, the first compounds a wet lab would screen if/when experimental work begins; Tier B (0.4–0.6) — moderate computational signal; Tier C (< 0.4) — weak / computational-only signal.
The integrated score combines target-convergence score, pChEMBL activity, ADMET profile, PAINS/hERG/AMES risk flags, and repurposing evidence where available. These are computational rankings, not measured activities.