Alzheimer’s disease (AD) is a progressive neurodegenerative disorder categorized by cognitive decline, synaptic dysfunction and complex multifactorial pathology. Among its molecular contributors, monoamine oxidase-B (MAO-B) plays a significant role in oxidative stress and neurotransmitter imbalance. Given the limited efficacy, off-target risks and pharmacokinetic constraints of existing MAO-B inhibitors such as safinamide (SAF), this study aimed to design improved analogues through AI-driven bioisosteric optimization. Using five FDA-approved MAO-B inhibitors as chemical templates, a total of 14,552 Bioisosteres (BIOS) were generated via the MolOpt AI platform. These structures were refined through Lipinski filters and TOPKAT toxicity models, yielding 440 viable candidates. Their inhibitory potential was assessed using an ML-based MAO-B predictor employing PubChem fingerprints, substructure keys and 1D/2D descriptors, with all BIOS displaying predicted nanomolar (nM) activity. Docking simulations against MAO-B (PDB ID: 2V5Z) using the CDOCKER algorithm identified ten high-affinity hits with superior interaction energies and enriched residue interactions compared to SAF. Subsequent ADMET profiling through ADMETlab 3.0 prioritized two leads BIOS 5261 and BIOS 7673 exhibiting favourable pharmacokinetics, low predicted toxicity and enhanced blood brain barrier permeability. Complementary AI-based target prediction (MolPredictX) indicated strong affinity toward NADPH and COX-2, suggesting potential multitarget relevance in AD modulation. Collectively, this integrated AI/ML-guided framework enabled the efficient discovery of structurally novel, drug-like MAO-B inhibitors with improved computational profiles. The findings underscore the power of bioisosteric design combined with computational intelligence for developing next-generation therapeutic candidates for AD.
Kumar et al. (Wed,) studied this question.