MotifFinder takes in a file containing the weka specific output to a RandomForrest or Decision tree classifier and outputs a list of probable motifs.
Java code parses throught he output of the weka classifier trees and recreates the each of the trees. It then traverses through each node of the tree storing the position and residue at each branching node. When it gets to a leaf node it adds the motif to a map of either toxic, neutral, or antitoxic motifs depending on the classification at that leaf node. The map is a multimap sorted by the classification score of the motif that was added. When all trees have been traversed the program will output the k most liekly motifs created from the decision forest.
Classification score is calculated from the total number of training data points that were classified in that node divided by the number of training points in that node that were missclassified. This ensurres that the most likely motifs are both highly represented, yet still reliably accurate.
Combiner iterates over the list of motifs for each of the toxicity classes and checks to see if each motif matches any of the other motifs. If these two motifs can be combined, they are put together into a new consesus motif that is added to a list of combined motifs.
- A text file with results from a weka Decision Tree, j48, or Random Forrest classifier run.
- A number k corresponding to how many motifs in each class you want to find. Default = 10. [Use -a or 0 to find all possible motifs]
- -RF if searching in a RandomForrest. By default algorithm searches in a single decision tree. Must include the -RF option to search through mutliple trees in the output.
- -c: output an additional list of combined motifs. Motifs found from finder could be partial motifs. -c calls a combiner class to combine these partial motifs together to create a consesus motif. Outputs these potential consesus motifs to std.err so they can be seperated from the normal motifs
Motif Finder prints the k most probable motifs for each class to standard output along with the total number of classification instances plus the number of missclasified instances, and the total score for classification. The motifs are written in regular expression format. Perfect for using in a grep command to find the peptides in the original data set that match the specified motif.
Following shows the example output of running the following command:
java -jar MotifFinder.jar randomForest.txt 10 -c...S.... toxic (16/0) 16.000000
......D. toxic (15/0) 15.000000
LH...... toxic (63/4) 12.600000
L...L... toxic (90/7) 11.250000
G.....H. toxic (62/5) 10.333333
G......G toxic (40/3) 10.000000
.......V toxic (10/0) 10.000000
L....N.. toxic (18/1) 9.000000
.G...ND. toxic (27/2) 9.000000
..N...CD toxic (9/0) 9.000000
.I...... anti-toxic (36/0) 36.000000
V..V.... anti-toxic (30/2) 10.000000
.I....V. anti-toxic (84/8) 9.333333
F......F anti-toxic (101/10) 9.181818
...F..Y. anti-toxic (35/3) 8.750000
.I...G.. anti-toxic (168/19) 8.400000
..C.V... anti-toxic (56/6) 8.000000
..C...C. anti-toxic (93/11) 7.750000
.D..F... anti-toxic (59/7) 7.375000
.....F.H anti-toxic (50/6) 7.142857
....L.R. neutral (72/6) 10.285714
YSV..... neutral (20/1) 10.000000
.....NR. neutral (215/22) 9.347826
V.....R. neutral (46/4) 9.200000
.Y..C... neutral (16/1) 8.000000
.L...S.. neutral (53/6) 7.571429
.....SV. neutral (106/13) 7.571429
......RF neutral (22/2) 7.333333
.....DY. neutral (35/4) 7.000000
V.....D. neutral (130/18) 6.842105
COMBINED MOTIFS:
===================
...S..D. toxic
LH.S.... toxic
L..SL... toxic
G..S..H. toxic
G..S...G toxic
...S...V toxic
L..S.N.. toxic
.G.S.ND. toxic
..NS..CD toxic
LH....D. toxic
L...L.D. toxic
G.....DG toxic
......DV toxic
L....ND. toxic
.G...ND. toxic
LH..L... toxic
LH.....V toxic
LH...N.. toxic
LHN...CD toxic
L...L..V toxic
L.N..NCD toxic
.G...ND. toxic
..N...CD toxic
VI.V.... antitoxic
.I....V. antitoxic
FI.....F antitoxic
.I.F..Y. antitoxic
.I...G.. antitoxic
.IC.V... antitoxic
.IC...C. antitoxic
.I...F.H antitoxic
VI.V..V. antitoxic
VI.V.G.. antitoxic
V.CVV... antitoxic
V.CV..C. antitoxic
VD.VF... antitoxic
V..V.F.H antitoxic
FI....VF antitoxic
.I...GV. antitoxic
.IC.V.V. antitoxic
.I...FVH antitoxic
F..F..YF antitoxic
FI...G.F antitoxic
F.C.V..F antitoxic
.I...FVH antitoxic
F..F..YF antitoxic
FI...G.F antitoxic
.IC..GC. antitoxic
..C.V.C. antitoxic
..C.VF.H antitoxic
.DC.F.C. antitoxic
..C..FCH antitoxic
.D..FF.H antitoxic
.....F.H antitoxic
YSV.L.R. neutral
....LNR. neutral
V...L.R. neutral
.L..LSR. neutral
....L.RF neutral
YSV..NR. neutral
YSV..SV. neutral
YSV...RF neutral
YSV..DY. neutral
V....NR. neutral
.Y..CNR. neutral
.....NRF neutral
VY..C.R. neutral
VL...SR. neutral
V.....RF neutral
.Y..CSV. neutral
.Y..C.RF neutral
.Y..CDY. neutral
VY..C.D. neutral
.L...SV. neutral
.L...SRF neutral
VL...SD. neutral
Installed with the git repository is a jar file containing all the necesary code. Either run Motif Finder with the java -jar command or export the java file to the CLASSPATH environment variable and run java MotifFinder.
java -jar path/to/MotifFinder.jar decisionTree.txt -aor:
export CLASSPATH=/path/to/the/MotifFinder.jar
java MotifFinder decisionTree.txt 30 -cAdding the following line to the .bashrc or .bash_profile will create a new alias to use MotifFinder simply by tpying the command MotifFinder:
alias MotifFinder="java -jar path/to/MotiFFinder.jar"Then From the Command line:
MotifFinder randomForestTrees.txt 50 -c