Avoiding pitfalls when modelling ligands in macromolecular crystallography

The determination of protein-ligand complex structures by X-ray crystallography is a cornerstone of modern structure-guided drug discovery. However, the process is complex and fraught with potential pitfalls at every stage, from data collection to final model deposition. The presence of flawed or misinterpreted ligand models in the Protein Data Bank (PDB) can misdirect scientific efforts that rely on them as the basis for new hypotheses and experiments. This article outlines a practical approach for ligand validation during structure determination. We discuss the application of validation tools, such as those in Coot, MolProbity, Mogul and Buster-report, to avoid errors. By re-examining several deposited PDB entries, we illustrate key pitfalls, including (i) modelling a ligand into absent or ambiguous electron density, (ii) incorrect chemical definitions (e.g. chirality, tautomers), (iii) poor fit of parts of a ligand to the electron density and (iv) data/model mismatches during deposition. We emphasize the importance of a `null-hypothesis' approach and continuous critical assessment throughout the modelling process to improve the reliability of deposited structures.