AI-Driven Herbal Drug Discovery

Abstract

Artificial intelligence has transformed drug discovery over the past decade, with machine-learning methodology now central to target prediction, virtual screening, bioactivity prediction, and lead optimisation workflows across pharmaceutical industry and academic drug discovery. For herbal drug discovery directed at traditional-medicine phytochemical libraries including Ayurvedic, Chinese, and other traditional-medicine plant sources, AI-driven methodology has particular value because the library scale (tens of thousands of characterised phytochemicals across traditional-medicine sources) is well suited to high-throughput computational screening, and the classical-literature bioactivity annotation provides training signal for supervised-learning methodology. This study reviews 52 primary computational and applied AI-drug-discovery studies published between 2015 and 2022 on AI-driven herbal drug discovery relevant to Ayurvedic and related traditional-medicine phytochemical libraries, grouped under five frames: AD1 (target prediction and molecular property ML models), AD2 (virtual screening of phytochemical libraries), AD3 (bioactivity prediction using deep learning and graph neural networks), AD4 (generative models for phytochemistry-inspired lead optimisation), and AD5 (translational integrated AI platforms for herbal drug discovery). Each frame was scored with the AI Herbal Drug Discovery Index (AHDDI). Translational integrated AI platforms (AD5) scored highest at AHDDI 0.796, with bioactivity prediction (AD3) at 0.782 and virtual screening (AD2) at 0.768.