Reading ADMET Results: A Practical Guide for Drug Discovery Teams

Running an ADMET screen is straightforward. Knowing what to do when your compound has a predicted hERG IC₅₀ of 1.4 µM, a DILI liability flag, and Caco-2 permeability at the low end of acceptable — all at the same time — is not. ADMET interpretation requires domain knowledge, therapeutic-area context, and an understanding of how individual liabilities interact. BioMate is designed to make that interpretation explicit and actionable.

What the Numbers Actually Mean

Predicted ADMET values are not pass/fail thresholds to be read in isolation. A hERG IC₅₀ matters relative to the intended therapeutic exposure — a compound with 1 µM hERG inhibition and a predicted Cmax of 10 nM carries a very different cardiac risk than the same compound dosed to achieve µM plasma concentrations. Caco-2 permeability matters relative to the intended route of administration and whether the drug class generally relies on active transport. A DILI liability flag should be evaluated against whether the predicted structural alert is a well-validated mechanism or a statistical association in a noisy dataset.

BioMate contextualizes every ADMET prediction against these dimensions. Predicted values are presented alongside the relevant therapeutic-area thresholds from the literature, the expected therapeutic dose range, and the confidence of the underlying model for the specific structural class of the compound.

Reading Multiple Liabilities Together

The harder problem in ADMET interpretation is co-occurring liabilities. A compound that is marginally acceptable on three metrics simultaneously may have a combined risk profile that is unacceptable — metabolic instability that shortens half-life, combined with poor solubility that reduces the achievable dose, combined with moderate hERG liability, can together produce a compound that is not viable at any dose. BioMate surfaces co-occurring liabilities with a structured summary that assesses their interaction, not just their individual values.

"A single ADMET flag is often manageable with structural modification. Three co-occurring flags pointing in different structural directions may signal a scaffold that should be abandoned."

Remediation Suggestions

When the auto-QC loop identifies a compound with ADMET liabilities that fall below the target profile, BioMate does not just flag the failure — it suggests the class of structural modifications known in the medicinal chemistry literature to address each liability. Reducing hERG liability typically involves reducing lipophilicity and adding basic nitrogen. Improving solubility typically involves adding polar functionality or reducing aromaticity. These suggestions are not guaranteed to work, but they give the medicinal chemist a principled starting point rather than an open-ended optimization problem.

Further reading: FDA novel drug approvals database, ChEMBL bioactivity database (EBI), and RDKit open-source cheminformatics toolkit.