The Zen of Model Anomalies – Correct Most of Them. Treasure the Meaningful Valid Few. Live Serenely with the Rest!

Abstract

Historically, validation has been considered primarily as a gatekeeping function done at the end of a structure solution. Currently, the most interesting and important part of validation is the opportunity to correct diagnosed errors, provided mainly by local as opposed to global criteria, and available to you throughout the crystallographic process. Elsewhere in this book, you will hear about up-to-date methods in the data and model-to-data aspects of validation. This chapter addresses model validation and model improvement, first about current best-practice methodology (as done on the MolProbity website and elsewhere), and second about some important developments to anticipate in the near future.

Model validation has three primary parts: (a) geometry (bond lengths and angles, planarity, chirality), (b) conformation (rotamers, Ramachandran, ring pucker, RNA backbone conformers), (c) and sterics (clashes, H-bonds, packing). All of these both enhance and must be considered along with the information from electron density. The model criteria are primarily local, but their rate of occurrence can also be summarized as a global score.