Maximum Entropy Approach for Deducing Amino Acid Interactions in Proteins

We present a maximum entropy approach for inferring amino acid interactions in proteins subject to constraints pertaining to the mean numbers of various types of equilibrium contacts for a given sequence or a set of sequences. We have carried out several kinds of tests for a two-dimensional lattice model with just two types of amino acids with very promising results. We also show that the method works very well even when the mean numbers of contacts are not known and therefore can be applied to real proteins.

Figure 1

Fraction of the 17 021 sequences, with a unique ground state (${ϵ}_1=5$, ${ϵ}_2=3$, ${ϵ}_1=1$) in Eq. (13), for which the maximum entropy method correctly predicts the true ground state as a function of the probability $P_0$. The different curves correspond to different sequence sets considered.