What We can Learn about Protein Folding from Recent Progress in Structure Prediction

Abstract

Protein folding and protein structure prediction are long-standing fundamental problems in biophysics. Inherently they are very close and so their studies should progress vis-a-vis, but in reality, today's structure prediction technique is highly pragmatic so that its relevance to folding problem has been obscured. This paper contributes to filling this gap by focusing on "what we can learn about protein folding from the recent progress in structure prediction." In particular, we analyzed why the fragment assembly method, currently the most successful method in de novo structure prediction, is so powerful by designing a "chimera" protein experiments. In the chimera proteins, local structural preference is specific to the target sequence, while nonlocal interaction is only sequence-independent compaction force. We found that these chimera proteins can find the native fold of the intact sequences with high probability suggesting dominant roles of the local interactions. From these results, we suggest some "principles of protein folding." (1) For small proteins, compact structures that are fully compatible with local structural preference are surprisingly few, one of which is the native fold. (2) These local biases shape up the funnel-like energy landscape of proteins making Go-like model used for folding study a good approximation of real proteins.