Fig. 3
From: COUSCOus: improved protein contact prediction using an empirical Bayes covariance estimator
Dependence of the performance of PSICOV and COUSCOus on the effective number of sequences (N eff ) in the MSAs. The performance is evaluated using accuracies for the top L/10long, medium and short contacts. The solid line represents the averaged accuracies of the test set binned into five different categories of Neff (ln(Neff): [4, 5), [5, 6), [6, 7), [7, 8), [8, 10)). COUSCOus outperforms PSICOV independent of the l n(N eff ) in the test set