Fig. 2

An Overview of GenSET Phase 2 selection of the training and testing sets for T3SS effector prediction. The positive set comprised 30 known effectors (15 each from S. dysenteriae and S. Typhimurium). The negative set was a 10-fold larger group of randomly selected non-effectors from the two organisms. Machine learning was similar to that in GenSET Phase 1 and used unfiltered and filtered attributes. The trained algorithm was applied to the genome of a third organism for T3SS effector prediction. The testing sets comprised all sequences in the test organism’s genome (i.e. E. coli, Y. pestis, P. syringae, and S. fredii)