Benchmark set for the HHomp benchmark (Figure 1) in the Paper "HHomp - prediction and classification of outer membrane proteins" (NAR 2009).


The negative set contains 5000 random, non-OMP sequences from the SCOP database (v1.69). 

The positive set is built of sequences from the TransportDB. This database contains 4494 outer membrane channels from fully sequenced organism. To avoid testing HHomp on proteins that are similar to one of the 23 proteins with which is was trained, we exclude all sequences that have a BLAST E-value better than 0.01 to one of the 23 proteins. This yields a test set of 2164 OMBBs.

The file results.tar.gz contains the binned results of the four tested OMP prediction methods, which are used to draw the ROC plot in figure 1 in the paper. These result files are created by counting the true and false positive hits by different probability-/score-threshods. It also contains two score-files with the probabilities/scores of each method for all benchmark proteins.