Computational Prediction and Experimental Verification of New MAP Kinase Docking Sites and Substrates Including Gli Transcription Factors
Figure 3
D-finder architecture and results of human genome search.
(A) Overview of D-finder. D-finder consists of D-matcher, a pattern matching algorithm employing expert knowledge, and D-learner.T1, a profile HMM trained on the training set shown in Fig. 1B. D-matcher filters out most windows, but found many acceptable windows in most sequences. D-learner assigns a probability score to each window it is passed, and found above-threshold windows in only 403 of the sequences passed to it by D-matcher. When D-learner was run without D-matcher interceding, it found 2,260 above-threshold windows in 1,784 sequences. (B) The top 25 D-sites found by D-finder in the human genome.