Computational approach to predict species-specific type III secretion system (T3SS) effectors using single and multiple genomes

Table 6 GenSET 1 T3SS effector prediction on four organisms were compared to three other available machine learning programs. GenSET 1 performed better than other programs in three out of the four organisms tested except for S. dysenteriae (see Additional files 3 and 4 for the actual data)

	Top 40 positive prediction out of total confirmed effectors
Program^a	S. dysenteriae	E. coli	P. syringae	S. Typhimurium
EffectiveT3	7/24 (29.2%)	11/21 (52.4%)	9/51 (17.7%)	2/24 (8.3%)^d 1/8 (12.5%)^e
T3MM	15/24 (62.5%) ^c	5/21 (23.8%)	21/51 (41.2%)	4/24 (16.7%)^d 5/8 (62.5%)^e
BPBAac	13/24 (54.2%)	12/21 (57.1%)	20/51 (39.2%)	7/24 (29.2%)^d 6/8 (75.0%)^e
GenSET 1^b	5/9 (55.6%)	5/6 (83.3%)	16/30 (53.3%)	8/9 (88.9%) ^d 4/8 (50%)^e

^aOther programs namely SIEVE, T3SEpre, and Meta-analytic were not available or accessible at the time of investigation
^bFor the GenSET 1 method, 15 or 21 effectors were taken out from the total effectors as the positive data sets. Thus, the totals were less in numbers when compared to others
^cBold number denotes the highest value in a given column
^dTop 40 positive prediction for SPI-2 effectors
^eTop 40 positive prediction for SPI-1 effectors

ISSN: 1471-2164