FreeContact: fast and free software for protein contact prediction from residue co-evolution

Abstract

Background: 20 years of improved technology and growing sequences now renders residue-residue contact constraints in large protein families through correlated mutations accurate enough to drive de novo predictions of protein three-dimensional structure. The method EVfold broke new ground using mean-field Direct Coupling Analysis (EVfold-mfDCA); the method PSICOV applied a related concept by estimating a sparse inverse covariance matrix. Both methods (EVfold-mfDCA and PSICOV) are publicly available, but both require too much CPU time for interactive applications. On top, EVfold-mfDCA depends on proprietary software.

Results: Here, we present FreeContact, a fast, open source implementation of EVfold-mfDCA and PSICOV. On a test set of 140 proteins, FreeContact was almost eight times faster than PSICOV without decreasing prediction performance. The EVfold-mfDCA implementation of FreeContact was over 220 times faster than PSICOV with negligible performance decrease. EVfold-mfDCA was unavailable for testing due to its dependency on proprietary software. FreeContact is implemented as the free C++ library "libfreecontact", complete with command line tool "freecontact", as well as Perl and Python modules. All components are available as Debian packages. FreeContact supports the BioXSD format for interoperability.

Conclusions: FreeContact provides the opportunity to compute reliable contact predictions in any environment (desktop or cloud).

Figure 1

Runtimes for FreeContact. We measured the runtime (logarithmic y-axis) for different program components (x-axis) on a single thread. The program components were: “seqw” – sequence weighting; “pairfreq” – pairwise residue frequencies; “shrink” – shrinking of covariance matrix; “inv” – sparse inverse covariance estimation/covariance matrix inversion. The different colors distinguish: the original PSICOV implementation (blue), our acceleration of PSICOV (FC.psicov, yellow), our acceleration of the faster PSICOV version “sensible default” (FC.psicov-fast, green), and our implementation of EVfold-mfDCA (FC.evfold, red). The whiskers on the box plots show the most extreme data point that is less than 1.5-times the interquartile range from the box. Outliers are not shown. Total runtime of all methods tested is dominated by the sparse inverse covariance estimation/covariance matrix inversion component.

Figure 2

Speedup using multiple threads. A: Sequence weighting. Speed is calculated as: proteins in alignment² length of target protein/runtime. B: Pairwise residue frequency calculation. Speed is calculated as: proteins in alignment length of target protein²/runtime. Dashed lines indicate linear correlation, extrapolated from one thread. The whiskers extend to the most extreme data point that is less than 1.5-times the interquartile range from the box. The surprisingly clear correlation between the number of threads and speed demonstrates how well our implementation scales for multi-threading.