Tom Terwilliger Publishes Work on AlphaFold Predictions
Tom Terwilliger, a New Mexico Consortium Senior Research Scientist and affiliated Los Alamos National Laboratory Scientist, recently published his work, AlphaFold predictions are valuable hypotheses, and accelerate but do not replace experimental structure determination, in bioRxiv, The Preprint Server for Biology.
Artificial Intelligence (AI)-based methods such as AlphaFold have revolutionized structural biology, because now we are often able to predict protein structures with high accuracy. What is an AlphaFold prediction model? It is an AI system that predicts a protein’s 3D structure from its amino acid sequence The accuracy of a prediction is typically assessed by how closely it matches a structure in the Protein Data Bank (PDB) with the same sequence. Although these predictions are fairly good, they unfortunately are not always completely accurate as they do not include ligands, covalent modifications or other environmental factors.
In this research, the scientists look at the accuracies of AlphaFold predictions by assessing how well they agree with experimental data. In particular, they focus on very-high-confidence parts of AlphaFold predictions, evaluating how well they can be expected to describe the structure of a protein in a particular environment.
They put these results into context by examining how closely one crystal structure in the PDB can typically be reproduced by another crystal structure containing the same components, but crystallized in a different space group. They compared predictions with experimental crystallographic maps of the same proteins for 102 crystal structures.
In many cases, those parts of AlphaFold predictions that were predicted with very high confidence matched experimental maps remarkably closely. In other cases, these predictions differed from experimental maps on a global scale through distortion and domain orientation, and on a local scale in backbone and side-chain conformation.
This research suggests considering AlphaFold predictions as exceptionally useful hypotheses. They further suggest that it is important to consider the confidence in prediction when interpreting AlphaFold predictions and to carry out experimental structure determination to verify structural details, particularly those that involve interactions not included in the prediction. Although not perfect, AlphaFold predictions are increasingly useful in structural biology and forming useful structural hypotheses.
To read more see the entire article at: AlphaFold predictions are valuable hypotheses, and accelerate but do not replace experimental structure determination.