Fig. 4From: DeepECA: an end-to-end learning framework for protein contact prediction from a multiple sequence alignmentAccuracy improvement depends on the number of sequences in an MSA. The mean differences of prediction accuracy, between the Feature Added model and the Weighted MSA Model, against the number of sequences in an MSA, are shown for (a) top L/5, (b) top L/2, and (c) top L contacts of prediction at long range. The number of proteins in each bin is equal (i.e., 32 protein domains in each bin)Back to article page