Prosta-omics: machine learning-driven TPSA prediction and molecular modeling of ASPM-inhibitors for prostate cancer treatment.

Prostate cancer exhibits complex transcriptional heterogeneity that underlies disease progression and therapeutic resistance. We developed an integrative omics-to-therapy pipeline to identify actionable biomarkers and screen drug candidates using a combination of single-cell transcriptomics, cheminformatics, and machine learning. Single-cell RNA sequencing (scRNA-seq) of prostate cancer tissues enabled fine-grained clustering and differential gene expression analysis across malignant and non-malignant cell populations. ASPM (Abnormal Spindle Microtubule Assembly) emerged as a statistically significant and cell-type-enriched biomarker associated with proliferative tumor phenotypes. We curated a library of drug-like molecules and developed Prosta-Omics, a supervised machine learning tool trained to predict Topological Polar Surface Area (TPSA) from molecular SMILES using a Random Forest model. High-ranking compounds as predicted by Prosta-Omics were docked against a 3D model of ASPM revealing multiple candidates with strong binding affinities and key interaction motifs. The drugs with higher docking score were subjected into molecular dynamics (MD) simulation and ADMET analysis. This integrative strategy highlights ASPM as a viable therapeutic target and introduces Prosta-Omics as a robust predictive platform bridging single-cell analytics with AI-driven drug discovery for precision oncology in prostate cancer.