Fig. 7

Performance of published hypoxic gene signatures. A–D Application of 8 hypoxia gene signatures as described in [56] to training and validation datasets. A PRJNA561635 time series. B Specific RNA fractions other than total mRNA. C ccRCC tumor and healthy adjacent tissue samples. D mouse datasets. E Correlation between the hypoxia scores derived from the 8 molecular signatures tested and the hypoxia probability calculated using an ensemble of our 20 best classifiers calculated for all samples from the validation experiments. Color code goes as follows; blue: samples grown in normoxia, red: samples grown in hypoxia, dark blue: healthy kidney samples, dark red: ccRCC tumoral samples. F–G distribution of F₁-scores for normal and tumoral samples from the validation datasets using the published gene signatures and tree classifiers. F Molecular signatures used in [56]. Samples were classified as hypoxic when the score calculated exceeded 50% of the maximum for that dataset and signature. G 20 individual trees that composed the ensemble. Samples were classified as hypoxic when the probability given by a tree exceeded 0.5