this is a particular kind of [[Markov Random Fields (MRF)]] which is the most popular unsupervised method for modeling interaction between amino acids. It is trained to maximize pseudolikelihood on alignments of evolutionarily related proteins
Goal : Capture information about coevolving residues
The parameters of a Potts model directly correspond to the notion of capturing sitewise and pairwise frequencies, thus making them interpretable.
Other articles
https://tianyu-lu.github.io/communication/protein/ml/2021/04/09/Potts-Model-Visualized.html