Skip to main content
. 2019 Jul 5;35(14):i31–i40. doi: 10.1093/bioinformatics/btz394

Fig. 5.

Fig. 5.

Results for E3. The training set includes 1/5 healthy out of a total of 243 samples. The testing set has 1/5 (20-versus-80), 1/2 (50-versus-50) or 4/5 (80-versus-20) of samples coming from healthy individuals on the IBD dataset. Methods labeled identically to Figure 3