Accelerate Your Lead Discovery: ADMET-AI Now Integrated into Vecura

What is ADMET-AI?

ADMET-AI is an open-source machine learning platform designed to predict over 30 absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties for small molecules. It utilizes powerful Chemprop v2 message-passing neural networks and RDKit-derived physicochemical descriptors to analyze molecular structures provided as SMILES strings. By leveraging data from the Therapeutics Data Commons (TDC) benchmarks, the tool provides rapid, accurate, and insightful pharmacokinetic analysis in approximately 200 milliseconds per compound.

What can users do with ADMET-AI on Vecura?

With ADMET-AI on Vecura, users can:

• Instantly predict a comprehensive panel of over 30 ADMET properties for a single molecule or a batch of compounds.
• Contextualize candidate molecules by comparing their predicted values against the DrugBank approved drug set.
• Access actionable lead-optimization guidance through percentile ranking scores.
• Screen chemical libraries efficiently without managing complex, independent technical infrastructure.

What the output means

The output provides a structured JSON object containing predicted values for each ADMET endpoint, including probability scores for classification tasks and real-valued outputs for regression tasks. Furthermore, the model provides a 0–100 percentile rank for each endpoint relative to approved drugs, helping researchers instantly understand the drug-likeness of their candidates.

This output should be used to support scientific decision making. It does not replace experimental validation.

Why this matters

The early-stage evaluation of pharmacokinetic and toxicity profiles is a significant bottleneck in drug discovery. Traditionally, assessing these properties requires extensive and costly in-vitro or in-vivo assays. ADMET-AI helps streamline this process by acting as an "in-silico filter," allowing researchers to prioritize promising chemical scaffolds and eliminate compounds with unfavorable profiles early in the development cycle.

By providing rapid, contextualized feedback, ADMET-AI enables more informed decision-making, ultimately increasing the efficiency of the lead optimization process and reducing the risk of late-stage failure due to ADMET-related issues.

• Developed by: Swanson K et al.
• Source: GitHub repository
• Reference: Swanson K et al. "ADMET-AI: a machine learning ADMET platform for evaluation of large-scale chemical libraries." Bioinformatics 40(7):btae416, 2024. https://academic.oup.com/bioinformatics/article/40/7/btae416/7698030