In Silico Pharmacological Profiling of MitoCorex Candidate Molecules: ADMET, Target Engagement, Selectivity, and Stability Analysis

Data deposit accompanying the manuscript: In Silico Pharmacological Profiling of MitoCorex Candidate Molecules: ADMET, Target Engagement, Selectivity, and Stability Analysis It reports the complete in silico pharmacological profiling of 11 primary candidate molecules generated by the MitoCorex de novo design pipeline, a component of the DrugSynth AI governed multi-agent computational drug discovery platform targeting mitochondrial diseases. The validation applies a four-filter cascade: (i) Lipinski Rule of Five and PAINS screening, (ii) rescue mechanism compatibility assessment against five mitochondrial disease target classes (DNM1L/DRP1, PINK1, NFE2L2/Keap1, NDUFV1, SDHA), (iii) selectivity threshold evaluation via a 72-rule SMARTS-based structural alert library derived from eight published toxicology frameworks (Baell-Holloway PAINS, Brenk unwanted substructures, Kalgutkar reactive metabolites, Kazius Ames mutagenicity, Dykens-Will mitochondrial toxicity, Aronov hERG pharmacophores, Greene hepatotoxicity, Pelletier phospholipidosis), and (iv) molecular dynamics stability proxy assessment. The 72-rule library includes 8 mitochondria-specific toxicophore rules covering uncoupler pharmacophores, Complex I rotenoid scaffolds, biguanide inhibitors, and mitochondrial permeability transition pore openers. Three mitochondria-specific ADMET endpoints not available in standard tools (pkCSM, SwissADME, ADMETlab 2.0) were evaluated: mitochondrial membrane permeability, Nernst-based accumulation potential, and inner membrane uncoupling risk. All 11 candidates passed all four filter tiers with zero eliminations. Five synthesis priorities were identified: KND-002, PKA-002, DSA-001, SDA-002, and NDS-002. The 0% filter failure rate validates the front-loaded physicochemical property enforcement strategy embedded in the DrugSynth AI design pipeline. This work constitutes Stage 7 of the DrugSynth AI / MitoCorex pipeline and is part of a 10-manuscript series covering computational drug discovery from target identification through platform validation. The 72-rule SMARTS library (toxicity_alert_rules.yaml) and all candidate data (molecule_candidates.yaml, docking_results.yaml, admet_baselines.yaml) are deposited as supplementary files under CC BY 4.0. All molecules described herein are computationally validated hypotheses for experimental testing. Patent pending: US Provisional Application 64/018,624, filed March 27, 2026.