Comparison of the effect of different inference methods and alignment sources on precision of contact prediction, based on the reduced data set of 384 proteins.

L/5, L/2 and L denote the precisions at respective amounts of contacts considered for evaluation, where L is the length of protein. Runs of gplmDCA with decimation are denoted as decgplmDCA (N), where N indicates the amount of decimation rounds. Difference betwwen individual methods is significantly more perceptible when considering HHblits alignments than Pfam alignments. On this set plmDCA and gplmDCA perform comparably, with plmDCA20 showing slightly higher positive predictive value for the top ranked contacts (0.59 vs 0.58).Comparison of the effect of different inference methods and alignment sources on precision of contact prediction, based on the reduced data set of 384 proteins.