Table 3

Classification results of 34 allele datasets
MHC # of peptides AUC
ARB SMM SMMPMBEC NetMHC EpicCapo EpicCapo+
HLA-A*01:01 1157 0.964 0.980 0.977
    0.982
0.972 ± 0.004 0.977 ± 0.003
HLA-A*02:01 3089 0.934 0.952 0.946
    0.957
0.950 ± 0.004 0.951 ± 0.004
HLA-A*02:02 1447 0.875 0.899 0.899
    0.900
0.901 ± 0.004 0.909 ± 0.004
HLA-A*02:03 1443 0.884 0.916 0.916
    0.921
0.920 ± 0.003 0.923 ± 0.003
HLA-A*02:06 1437 0.872 0.914 0.916
    0.927
0.925 ± 0.004 0.927 ± 0.004
HLA-A*03:01 2094 0.908
    0.940
0.928 0.937 0.934 ± 0.004 0.938 ± 0.003
HLA-A*11:01 1985 0.918 0.948 0.939
    0.951
0.945 ± 0.004 0.951 ± 0.002
HLA-A*24:02 197 0.718 0.780 0.801
    0.825
0.853 ± 0.012 0.865 ± 0.011
HLA-A*26:01 672 0.907 0.931 0.924
    0.956
0.941 ± 0.005 0.957 ± 0.007
HLA-A*29:02 160 0.755 0.911 0.916
    0.935
0.944 ± 0.008 0.945 ± 0.010
HLA-A*31:01 1869 0.909
    0.930
0.925 0.928 0.930 ± 0.002 0.935 ± 0.003
HLA-A*33:01 1140 0.892
    0.925
    0.925
0.915 0.926 ± 0.004 0.934 ± 0.004
HLA-A*68:01 1141 0.840 0.885
    0.885
0.883 0.891 ± 0.003 0.899 ± 0.003
HLA-A*68:02 1434 0.865 0.898 0.889
    0.899
0.901 ± 0.005 0.907 ± 0.003
HLA-B*07:02 1262 0.952 0.964 0.960
    0.965
0.960 ± 0.004 0.964 ± 0.002
HLA-B*08:01 708 0.936 0.943
    0.956
0.955 0.942 ± 0.005 0.951 ± 0.004
HLA-B*15:01 978 0.900
    0.952
0.940 0.941 0.940 ± 0.006 0.950 ± 0.005
HLA-B*18:01 118 0.573 0.853
    0.880
0.838 0.886 ± 0.013 0.911 ± 0.009
HLA-B*27:05 969 0.915 0.940
    0.941
0.938 0.949 ± 0.005 0.958 ± 0.003
HLA-B*35:01 736 0.851 0.889
    0.889
0.875 0.900 ± 0.004 0.907 ± 0.007
HLA-B*40:02 118 0.541 0.842
    0.843
0.754 0.811 ± 0.007 0.912 ± 0.011
HLA-B*44:02 119 0.533 0.740 0.739
    0.778
0.798 ± 0.009 0.861 ± 0.013
HLA-B*44:03 119 0.461
    0.770
0.753 0.763 0.813 ± 0.010 0.871 ± 0.008
HLA-B*51:01 244 0.822 0.868
    0.895
0.886 0.930 ± 0.012 0.948 ± 0.015
HLA-B*53:01 254 0.871 0.882 0.885
    0.899
0.916 ± 0.008 0.940 ± 0.008
HLA-B*54:01 255 0.847 0.921
    0.935
0.903 0.927 ± 0.008 0.938 ± 0.006
HLA-B*57:01 59 0.428
    0.871
0.843 0.826 0.792 ± 0.009 0.854 ± 0.010
HLA-B*58:01 988 0.889
    0.964
0.945 0.961 0.959 ± 0.005 0.964 ± 0.004
H-2 Db 303 0.865 0.912 0.901
    0.933
0.940 ± 0.014 0.968 ± 0.006
H-2 Dd 85 0.696 0.853 0.837
    0.925
0.956 ± 0.016 0.985 ± 0.017
H-2 Kb 223 0.792 0.810 0.833
    0.850
0.844 ± 0.021 0.880 ± 0.017
H-2 Kd 176 0.798 0.936 0.931
    0.939
0.950 ± 0.015 0.966 ± 0.009
H-2 Kk 164 0.758 0.770
    0.793
0.790 0.883 ± 0.009 0.926 ± 0.008
H-2 Ld 102 0.551 0.924 0.942
    0.977
0.984 ± 0.012 0.992 ± 0.013
Average 0.801 0.895 0.895 0.900 0.912 0.931
t-test|ARB NA 4.37E-5 3.69E-5 1.25E-5 5.21E-6 2.64E-6
t-test|SMM NA 8.61E-1 2.30E-1 8.28E-3 2.87E-5
t-test|SMMPMBEC NA 2.61E-1 3.50E-3 8.49E-6
t-test|NetMHC NA 8.57E-3 7.74E-5
t-test|EpicCapo NA 1.95E-5

For each dataset, AUCs were evaluated based on 5-fold cross validation. In the lower part, p-values of average AUCs were calculated using paired t-tests (two-tailed).

Means and standard deviations were calculated by 20 iterations of 5-fold cross validation for EpicCapo and EpicCapo+.

Underlined values represent the highest performance among ARB, SMM, SMMPMBEC, and NetMHC. Values in bold represent significant improvements of EpicCapo or EpicCapo+ AUCs from 20 iterations of 5-fold cross validation over the underlined values according to t-tests (one-tailed, significance level = 0.01).

Saethang et al.

Saethang et al. BMC Bioinformatics 2012 13:313   doi:10.1186/1471-2105-13-313

Open Data