Section: New Results

Pairwise distance potential for protein folding

Participants : Maria Kadukova, Guillaume Pages, Alisa Patotskaya, Sergei Grudinin.

We have developed a new knowledge-based pairwise distance-dependent potential using convex optimization. This method uses histogram of distances repartition between each different pair of atom types as feature to feed an SVM-like algorithm. We then obtained a potential for each pair of atom types that can be used to score protein conformations. This method have been extensively used during the CASP12 blind assesement.