Molecular Image-Based Explainable AI Framework Identifies Novel Candidate Antibiotics
AJAS · 2025 Computational Biology and Bioinformatics (inferred)
Overview
The continued growth of antibiotic resistance and the slowing of antibiotic discovery poses a large challenge in modern medicine. Recent advances in Artificial intelligence (AI) technologies offer a time- and cost-effective solution for rapid development of effective antibiotics. However, most AI models are 'black-boxes' and are difficult to interpret. We previously developed an explainable AI framework from a pre-trained model using 10 million drug-like molecular images. In this study, we developed a finetuned ImageMol from experimental S. aureus inhibition assays which contained 24,521 molecules consisting of 516 with demonstrated antibacterial behavior and 24,005 chemicals without. Our optimized AI model achieved a strong AUROC of 0.926 and was then used to predict the potential antibiotic activity of 10,247 molecules from the DrugBank database. After further filtering, 340 molecules were identified as candidate antibiotics that were dissimilar to known antibiotics. Finally, 76 of those candidates were identified as FDA-approved drugs for other applications. We also illustrated explainable molecular images for top predicted candidate drugs via Gradient-weighted Class Activation Mapping (Grad-CAM) heatmap analysis. In summary, this project provides a proof of concept suggesting that image-based AI models like ImageMol could be highly favorable in drug discovery due to its performance, speed, and explainability.
Competition history
- AJAS 2025
Related projects
CSEF · 2023
SMILES Encoding and Deep Learning to Predict Quantitative Efficacy of Structurally Distinct Molecules for Antibiotics
ISEF · 2022
Synthesis and Motif Discovery of Novel Antibacterial Molecules and Orally Active Drugs via Sequence-Based Machine Learning
ISEF · 2026
Beyond the Mutation: Evolution-Aware Antibiotic Discovery Through a Novel Resistance-Robust Computational Framework
ISEF · 2018
Battling Superbugs: A Machine Learning Framework for Identifying Novel Antibiotic Resistance Factors
ISEF · 2026
Explainable AI-Driven Enhancement of in silico Drug Screening for Chronic Toxoplasma gondii Infection
ISEF · 2025
AI Driven Drug Discovery: Designing Cancer Inhibiting Molecules With Genetic Algorithm
AJAS · 2020
Development of an Antimicrobial Peptide Activity Detector Using Machine Learning
CWSF · 2026
Computer Vision For Bacterial Gram Stain Classification
Closest projects by meaning, across every fair and year in the corpus.
Browse more like this
Source: AAAS Annual Meeting (Confex) / American Junior Academy of Science