Cecilia Lövkvist: Improved prediction of direct contact in protein structure with Potts model

Protein structure prediction (PSP) has become an active field in structural biology and aims to determine 3D-structures of proteins. This is a time-consuming and expensive task and various methods for PSP are therefore developed to solve this problem. In statistical physics, probabilistic models are built to accurately determine the PSP from the multiple sequence alignment (MSA). Two methods are recently used: the naive mean-field inversion and the pseudolikelihood maximization. The methods are tested for 17 small protein families and in this work they are applied to 53 families including very large ones. Both methods show effectiveness on the extended data set. The pseudolikehood method seems to have better performance than the naive mean field inversion.