Effect of compression on mean best CS (Correspondence Score) at various threshold values. (A) Mean best CS shown here is calculated for the InterPro keywords. Mean best CS decreases from right to left, as more clusters are eliminated due to compression. Filled circles represent the mean best CS for compression according to LT thresholds. Open circles represent the mean best CS of a random compression to the same extent as explained in the text. Standard deviation of the random mean best CS is too small to be seen on the graph.