Running times in seconds for a representative sample of proteins with varying length () and sequences in alignment (), using different algorithms.

Since the Gaussian DCA code is parallelized, we show two series of results, one in which we used 8 cores and one in which we forced the code to run on a single core, for the sake of comparing with the non-parallel code of PSICOV and plmDCA. These benchmarks were taken on a -core cluster of MHz AMD Opteron 6172 processors running Linux 3.5.0; PSICOV version 1.11 was used, compiled with gcc 4.7.2 at -O3 optimization level; plmDCA was run with MATLAB version r2011b. Gaussian DCA timings shown are taken using the Julia version of the code, using Julia version 0.2.