A Multi-Engine Consensus Docking Pipeline for RNA Aptamer Screening Against ACC Oxidase (ACO): Statistical Validation, Machine Learning Analysis, and Pilot Cross-Target Evaluation Against ACC Synthase (ACS)

Postharvest losses in climacteric fruits are largely driven by ethylene, and inhibiting its biosynthesis is an active research goal. RNA aptamers are attractive candidates for modulating the two rate-limiting enzymes—ACC oxidase (ACO) and ACC synthase (ACS), without the toxicity concerns of chemical inhibitors. We built a computational pipeline using three independent docking engines: ITD DOCK (Graphics Processing Unit [GPU]-accelerated, Assisted Model Building with Energy Refinement [AMBER] force fields), HDOCK Lite (knowledge-based geometric potentials), and HADDOCK3 (semi-flexible refinement), screened against large aptamer libraries. Applied to ACC oxidase with 9813 aptamers, engine scores were largely complementary, with all pairwise correlations statistically non-significant on the full dataset (Spearman |ρ|≤0.013, p>0.20 for all pairs, n=9813). A Random Forest on sequence-only features failed to predict docking scores (R2=−0.012, Root Mean Square Error [RMSE] = 21.79 kcal/mol), while a positive-control model on HADDOCK3 energy components achieved R2=0.991, confirming that predictive information only becomes available after docking. A preliminary pilot screening against ACC synthase (n=97 valid complexes) suggested pipeline generalizability to a second structurally distinct target and identified two putative dual-active candidates, pending confirmation against the full library. The complete workflow is automated in AptaDock, a standalone desktop application.