Predicting the conformation of proteins from sequences. Progress and future progress

Recent progress in structure prediction has allowed bona fide predictions, those made and published before an experimental structure is determined, to be remarkably accurate. The most successful methods rely on an analysis of patterns of conservation and variation within homologous protein sequences, extract tertiary structural information before secondary structure is predicted, and avoid 'three state per residue scores' as a tool for evaluating a prediction, focusing instead on efforts to understand why a prediction is successful when it is successful, and why it fails when it fails.