Deep Learning Pipeline for Accelerating Virtual Screening in Drug Discovery
Problem
Drug discovery is slow and costly, and diseases like Marburg virus lack treatments, so faster identification of candidate compounds is needed.
Why it matters
High-cost, time-consuming screening delays therapies for deadly diseases with limited treatment options.
What the AI did
VirtuDockDL, a Python-based web platform using a Graph Neural Network, predicts compound effectiveness and identified non-covalent inhibitors against the Marburg virus VP35 protein during validation.
Results
99% accuracy, F1 score 0.992, AUC 0.99 on HER2 dataset, surpassing DeepChem (89%) and AutoDock Vina (82%)
Impact
Benchmarked against multiple existing tools with stated superior performance; inhibitor candidates identified computationally only.
Record
| Venue | Methods in molecular biology |
|---|---|
| Published | 2024 |
| Work type | book-chapter |
| Authors | Fatima Noor, Muhammad Junaid, Atiah H. Almalki, Mohammed Almaghrabi, Shakira Ghazanfar, Muhammad Tahir ul Qamar |
| Citations | 54 (OpenAlex, at import time) |
| References | 34 |
| Model or method | VirtuDockDL (Graph Neural Network) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1007/978-1-0716-5190-2_18 |
| Open access | gold |
| Keywords | virtual screening, graph neural network, drug discovery, Marburg virus, VP35 inhibitors |
Importance rating
major — millions of lives, an industry or a field
Drug discovery is described as high-cost and time-consuming, and Marburg virus has high fatality with limited treatment options; no quantified scale given, but the problem spans an entire industry and deadly diseases.