Skip to main content
. 2019 Dec 13;36(8):2466–2473. doi: 10.1093/bioinformatics/btz932

Table 6.

Classification performance of estimated isoform expressions on three experiments

Experiment Task Kallisto Kallisto (bias) Salmon Salmon (bias) eXpress eXpress (bias) RSEM RSEM (bias) BaseEM IntMTQ
SVM
 E1 OV 0.626(0.101) 0.606(0.099) 0.631(0.101) 0.662(0.100) 0.624(0.091) 0.605(0.096) 0.631(0.097) 0.648(0.098) 0.628(0.100) 0.691(0.090)
OV+BRCA 0.607(0.086) 0.598(0.085) 0.572(0.085) 0.573(0.087) 0.634(0.082) 0.604(0.083) 0.595(0.087) 0.596(0.084) 0.551(0.083) 0.562(0.082)
 E2 OV 0.643(0.153) 0.637(0.151) 0.611(0.152) 0.648(0.150) 0.621(0.146) 0.612(0.149) 0.595(0.149) 0.613(0.145) 0.594(0.143) 0.660(0.145)
OV+BRCA 0.852(0.072) 0.851(0.070) 0.856(0.071) 0.893(0.062) 0.852(0.073) 0.863(0.072) 0.862(0.068) 0.869(0.068) 0.855(0.072) 0.881(0.055)
 E3 OV 0.533(0.135) 0.543(0.134) 0.540(0.126) 0.596(0.139) 0.534(0.130) 0.558(0.137) 0.524(0.128) 0.546(0.133) 0.532(0.140) 0.608(0.140)
Random Forest
OV+BRCA 0.643(0.120) 0.721(0.113) 0.644(0.123) 0.695(0.110) 0.655(0.116) 0.672(0.118) 0.671(0.122) 0.672(0.112) 0.767(0.103) 0.779(0.095)
 E1 OV 0.701(0.092) 0.706(0.089) 0.717(0.090) 0.724(0.087) 0.734(0.091) 0.744(0.089) 0.712(0.091) 0.707(0.094) 0.719(0.092) 0.763(0.080)
OV+BRCA 0.628(0.088) 0.648(0.090) 0.632(0.091) 0.664(0.090) 0.660(0.090) 0.663(0.092) 0.622(0.090) 0.636(0.091) 0.650(0.087) 0.660(0.086)
 E2 OV 0.686(0.134) 0.685(0.135) 0.624(0.143) 0.627(0.144) 0.616(0.146) 0.633(0.140) 0.690(0.135) 0.693(0.132) 0.698(0.131) 0.717(0.130)
OV+BRCA 0.804(0.090) 0.798(0.090) 0.721(0.111) 0.742(0.106) 0.728(0.109) 0.729(0.105) 0.803(0.094) 0.807(0.095) 0.799(0.088) 0.851(0.079)
 E3 OV 0.540(0.152) 0.542(0.149) 0.591(0.147) 0.632(0.145) 0.582(0.152) 0.592(0.147) 0.569(0.150) 0.569(0.153) 0.576(0.148) 0.628(0.144)
OV+BRCA 0.639(0.117) 0.638(0.111) 0.664(0.112) 0.671(0.112) 0.686(0.115) 0.689(0.111) 0.641(0.115) 0.648(0.114) 0.680(0.114) 0.694(0.111)

Note: The mean AUC scores (standard deviation in parentheses) of classifying cell lines by estimated transcript expression in 1000 repeats for each experiment are reported. The best AUCs across the ten methods are bold.