Assay2Mol: Large Language Model-based Drug Design Using BioAssay Context

Area of Science:

• Biochemistry
• Computational drug discovery
• Bioinformatics

Background:

• Scientific databases contain vast quantitative and text data.
• Biochemical assays screen molecules against disease targets.
• Unstructured text in assays holds valuable drug discovery information.
• This information is largely untapped due to its format.

Purpose of the Study:

• To present Assay2Mol, a large language model-based workflow.
• To leverage existing biochemical screening assays for early-stage drug discovery.
• To unlock the potential of unstructured assay data.

Main Methods:

• Assay2Mol utilizes a large language model.
• It retrieves existing assay records for similar targets.
• It generates candidate molecules using in-context learning from retrieved data.

Main Results:

• Assay2Mol outperforms recent machine learning approaches.
• It effectively generates candidate ligand molecules for target protein structures.
• The workflow promotes the generation of more synthesizable molecules.

Conclusions:

• Assay2Mol capitalizes on existing biochemical screening assays.
• It offers a novel approach for early-stage drug discovery.
• The method enhances the generation of viable drug candidates.