cidalsDB: an AI-empowered platform for anti-pathogen therapeutics research

Area of Science:

• Computational chemistry and cheminformatics.
• Infectious disease research.
• Artificial intelligence in drug discovery.

Background:

• Advances in big data analytics and AI are transforming drug discovery (DD).
• Reliable datasets are crucial for effective AI-driven drug discovery.
• Existing resources may lack integrated datasets and predictive models for specific pathogens.

Purpose of the Study:

• To present CidalsDB, a novel web server for AI-assisted drug discovery against infectious pathogens.
• To provide accessible, ready-to-use datasets and optimized AI models for predicting anti-pathogen activity.
• To foster innovation and collaboration in drug discovery through a no-code platform.

Main Methods:

• Literature search for molecules with validated anti-pathogen effects.
• Consolidation of data with bioassays from PubChem.
• Construction of a web-accessible database (CidalsDB).
• Implementation and optimization of machine learning (ML) and deep learning (DL) algorithms for activity prediction.

Main Results:

• CidalsDB provides curated datasets for Leishmania parasites and Coronaviruses.
• Optimized ML/DL models demonstrated high performance in predicting biological activity.
• Random Forests, MLP, and ChemBERTa excelled for anti-Leishmania prediction.
• Gradient Boosting, GCN, and ChemBERTa showed best performance for Coronavirus prediction.

Conclusions:

• CidalsDB is an open-access, web-based tool for AI-based drug discovery.
• It offers integrated datasets and predictive AI models, democratizing CADD.
• The platform facilitates research on anti-pathogen molecules and accelerates drug discovery efforts.