Fig. 5

A Top 20 important features extracted from all data-all transcripts dataset obtained using (i) SVM and (ii) RF. B Recursive feature elimination (RFE) was used with the top 20 features from (A) to extract a list of the least number of features required to differentiate between healthy and HCC cell models. Three features were selected having an accuracy of 0.97 and kappa of 0.93 (C) Average expression of three features (transcripts), across healthy liver and HCC cell models, chosen in (B)