AI-Driven Discovery of Novel Therapeutic Targets for Biliary Atresia: A Computational Drug Discovery Approach

Keywords: Biliary atresia, Drug discovery, Artificial intelligence, Virtual screening, Molecular docking, ADMET prediction, Molecular dynamics simulation, Structure-activity relationship, Pediatric therapeutics, Computational biology

TL;DR: AI-driven pipeline identified 50 pediatric lead compounds for biliary atresia, with Smoothened receptor targets showing optimal therapeutic potential via virtual screening.

Abstract:

Biliary atresia (BA) is a rare but severe pediatric liver disease characterized by progressive destruction of bile ducts, leading to cholestasis and liver failure if untreated. Current therapeutic options are limited, with liver transplantation being the only definitive treatment for advanced cases. This study presents a comprehensive computational drug discovery pipeline for identifying novel therapeutic targets and candidate compounds for BA treatment. We employed AI-driven approaches including virtual compound library construction, molecular docking, ADMET prediction, molecular dynamics simulation, and structure-activity relationship analysis. Our pipeline screened 492 compounds against three key targets (TNF-α, IL-13, and Smoothened), identifying 50 high-quality hit compounds with favorable drug-like properties. The top candidates demonstrated strong binding affinity (average score 8.2), excellent stability in molecular dynamics simulations (93.8\% classified as highly stable), and pediatric-appropriate ADMET profiles. Notably, compounds targeting the Smoothened receptor showed superior performance across all evaluation metrics. This work establishes a robust computational framework for BA drug discovery and provides a prioritized set of lead compounds for experimental validation.

note: we submit our code as supplementary file in https://github.com/mahongbin970-sketch/Agent4Science-BA

Submission Number: 55