Abstract
Isovaleric acidemia (IVA; OMIM 243500) is caused by biallelic mutations in isovaleryl-CoA dehydrogenase (IVD), a mitochondrial flavoenzyme belonging to the acyl-CoA dehydrogenase (ACAD) family. Yet, currently, no pharmacological treatment exists that targets the underlying protein defect. In this study, we assembled a library of 15,474 compounds with recorded activity against ACAD-family enzymes from ChEMBL and trained binary classifiers (using Random Forest and XGBoost) on three sets of descriptors. The best-performing model (Random Forest with hybrid descriptors) achieved a cross-validated AUC-ROC of 0.709, a result we confirmed through Y-randomization (p < 0.001) and hold-out testing (AUC-ROC 0.705). Our virtual screening identified 1,606 hits. We then docked these hits, along with 21 repurposed drugs, into the FAD/substrate cavity of the human IVD crystal structure (PDB 1IVH) using AutoDock Vina. Notably, three CFTR modulators already approved for cystic fibrosis—lumacaftor (−9.71 kcal/mol), tezacaftor (−9.29 kcal/mol), and ivacaftor (−8.75 kcal/mol)—ranked among the strongest binders. They stood alongside our QSAR-derived hits, such as CHEMBL1513643 (−9.33 kcal/mol) and CHEMBL1367030 (−9.31 kcal/mol). We then applied a multi-parameter ranking that combined docking affinity, drug-likeness, predicted blood–brain barrier permeability, and Lipinski compliance. This process placed CHEMBL1338958 and CHEMBL1365696 at the very top of our candidate list. By mapping interactions across the 15 highest-scoring compounds, we identified 22 conserved contact residues, with Trp166, Gly141, and Gly376 acting as the principal pharmacophoric anchors. To the best of our knowledge, this work represents the first QSAR-driven virtual screening campaign directed at IVD and the first report to suggest CFTR correctors as candidate IVD chaperones.