Enhancing Antimicrobial Molecule Prediction via Dynamic Routing Capsule Networks and Multi-Source Molecular Embeddings
Fuente:
arXiv
Saved in:
| Main Author: | |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916958945935360 |
|---|---|
| author | He, R. |
| author_facet | He, R. |
| contents | Antibiotics are a vital class of drugs closely associated with the prevention and treatment of bacterial infections. Accurate prediction of molecular antimicrobial activity remains a key challenge in the pursuit of novel antibiotic candidates. However, laboratory-based antimicrobial compounds identification is costly, time-consuming, and prone to rediscovering known antibiotics, highlighting the urgent need for efficient and accurate computational models. Recent advances in machine learning (ML) and deep learning (DL) have significantly enhanced the ability to explore chemical space and identify potential antimicrobial compounds. In this study, we particularly emphasize deep learning models and employ five chemistry language models tailored for chemical data to encode small molecules. Our model incorporates a unique capsule network architecture and introduces innovations in loss function selection and feature processing modules, demonstrating superior performance in predicting inhibitory activities against Escherichia coli and Acinetobacter baumannii. We conducted a series of ablation studies to elucidate the contributions of network design and input features. Case studies validated the usability and effectiveness of our model.To facilitate accessibility, we developed an intuitive web portal to disseminate this novel tool. Our results indicate that the proposed approach offers improved predictive accuracy and enhanced interpretability, underscoring the potential of interpretable artificial intelligence methods in accelerating antibiotic discovery and addressing the urgent challenge of antimicrobial resistance. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_16571 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Enhancing Antimicrobial Molecule Prediction via Dynamic Routing Capsule Networks and Multi-Source Molecular Embeddings He, R. Quantitative Methods Antibiotics are a vital class of drugs closely associated with the prevention and treatment of bacterial infections. Accurate prediction of molecular antimicrobial activity remains a key challenge in the pursuit of novel antibiotic candidates. However, laboratory-based antimicrobial compounds identification is costly, time-consuming, and prone to rediscovering known antibiotics, highlighting the urgent need for efficient and accurate computational models. Recent advances in machine learning (ML) and deep learning (DL) have significantly enhanced the ability to explore chemical space and identify potential antimicrobial compounds. In this study, we particularly emphasize deep learning models and employ five chemistry language models tailored for chemical data to encode small molecules. Our model incorporates a unique capsule network architecture and introduces innovations in loss function selection and feature processing modules, demonstrating superior performance in predicting inhibitory activities against Escherichia coli and Acinetobacter baumannii. We conducted a series of ablation studies to elucidate the contributions of network design and input features. Case studies validated the usability and effectiveness of our model.To facilitate accessibility, we developed an intuitive web portal to disseminate this novel tool. Our results indicate that the proposed approach offers improved predictive accuracy and enhanced interpretability, underscoring the potential of interpretable artificial intelligence methods in accelerating antibiotic discovery and addressing the urgent challenge of antimicrobial resistance. |
| title | Enhancing Antimicrobial Molecule Prediction via Dynamic Routing Capsule Networks and Multi-Source Molecular Embeddings |
| topic | Quantitative Methods |
| url | https://arxiv.org/abs/2509.16571 |