A protein calculation has to name three objects: the sequence, the ensemble of folds, and the surface a partner touches. Twenty residues are free at each site, a fold is a population, and a surface charge is not yet a binding free energy.
Tamsin Mansley and Mario Öeren's Optibrium webinar puts a QSAR in a fixed order: curate the assay, match regression or a class to the noise, hold out a test set that is scored once, and read the plot. A clearance label can still be the wrong target, because one number mixes several mechanisms.
The training words are steps in one loop. This note defines each one on a 1,000-molecule example, from the batch count and the loss to message passing and a scaffold split.
A cluster model deletes the protein outside the active-site fragment. QM/MM keeps that protein and couples it to the quantum region. The methods sentence has to say which calculation was run.
BindingDB holds the measurement. PubChem holds the crosswalk. Christopher Southan's BindingDB webinar shows how to follow a patent series from one into the other without treating a repeated deposit as a new experiment.
What EFMC-ISMC 2026 revealed about the changing practice of drug discovery: from generative chemistry and automated DMTA cycles to allosteric pockets, DMPK liabilities, and human decision-making.
Key takeaways from the Optibrium panel discussion featuring Nathan Brown, Charlotte Deane, Pat Walters, Chris Swain, and Paul Czodrowski on AI strategy, data leakage, generative risks, and LLMs in drug discovery.
BindingDB stores measured affinities. Michael Gilson's first BindingDB webinar is a guide to which rows can supervise a docking or machine-learning job, and which lookalikes cannot.
ORCA reports the electrons of a ligand. This note maps each common calculation onto the question it can answer, and onto the binding claim it cannot carry.
What successful incretin drugs, BPC-157, and Epitalon teach us about peptide engineering, translational evidence, and the limits of molecular plausibility.
