Skip to main content

Table 3 Average performance of the algorithms on the four organisms using unfiltered (U) and filtered (F) attributes. The PPV (positive predictive value or precision), TPR (True positive rate or sensitivity), SPC (specificity or true negative rate), and AUC (area under the curve) values were calculated using the trained (voting) algorithm (actual values for all algorithms are given in Additional file 2)

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

GenSET Organism/Attributesa PPV TPR SPC AUC
Phase 1 E. coli U 0.068 1.000 0.980 0.998
F 0.082 1.000 0.984 0.993
P. syringae U 0.300 0.750 0.989 0.988
F 0.520 0.542 0.997 0.970
S. dysenteriae U 0.086 1.000 0.984 0.999
F 0.106 1.000 0.987 0.997
S. Typhimurium U 0.025 1.000 0.962 0.987
F 0.022 1.000 0.956 0.984
Phase 2 E. coli U 0.112 0.950 0.957 0.980
F 0.132 0.800 0.970 0.979
Y. pestis U 0.429 0.273 0.999 0.943
F 0.000 0.000 0.999 0.882
P. syringae U 0.330 0.800 0.982 0.981
F 0.337 0.711 0.984 0.976
S. fredii U 0.364 0.444 0.998 0.967
F 0.000 0.000 0.999 0.955
  1. aThe voting algorithm used unfiltered (U) attributes (utilize all 21 attributes) and filtered (F) attributes (utilizing a subset of attributes selected by the feature selection methods)