DrugSynq Blog
Practical guides on FEP, ADMET prediction, and structure-based design from the team building DrugSynq.
Featured
FEP is no longer confined to methodology papers. This guide covers what it predicts, where it fails, how to design perturbations for a lead series, and what accuracy you should realistically expect for your target class.
AUROC 0.90 sounds impressive, but which training split? What prevalence assumption? This post unpacks the benchmarking choices that determine whether published ADMET accuracy translates to your program.
A step-by-step walkthrough of the SBDD pipeline: target preparation, hit generation, FEP-guided lead optimization, and ADMET gating — with practical decision points at each stage.
Neural network potentials achieve QM accuracy at a fraction of the cost. But for relative binding free energies in drug discovery, physics-based FEP with ML correction terms currently outperforms end-to-end ML. Here's why.
Computational FBDD covers initial fragment docking, fragment growing into the binding pocket, and fragment merging strategies. FEP validates each elaboration before synthesis. How the pieces connect.
A detailed walkthrough of a kinase inhibitor optimization program using DrugSynq: how FEP ΔΔG predictions ranked congeneric series, where ADMET flags redirected synthesis, and what accuracy looked like in practice.
