Additional file 10 of A gene-based score for the risk stratification of stage IA lung adenocarcinoma

Additional file 10: Figure S3. Feature selection of genes representing different patterns. Consensus cumulative distribution function (CDF) (A) and Delta area (B) of consistent cluster analysis of genes through unsupervised clustering; consensus clustering of genes into four clusters (C); GO analysis of four clusters (D); prediction effect of 114 genes into the logistic regression model (classifying high-risk, H and low-risk, L) in training set (E) and external validation set (GSE30219) (F); prediction effect of 41 genes with significant coefficients into the logistic regression model (classifying high-risk, H and low-risk, L) in training set (G) and external validation set (GSE30219) (H); Heatmaps of 114 genes in the training set (original: I, smoking related residual values removed: K) and the validation set (J) respectively; GO analysis of 64 genes of IA score (L); Expression distribution of risk genes or oncogenes ((OR > 1) (M) and protective genes or tumor suppressor genes (OR < 1) (N) in early-stage LUAD (AIS, MIA and IAC). CDF, Consensus cumulative distribution function, OR, Odds Ratio; LASSO, least absolute shrinkage and selection operator; LUAD, lung adenocarcinoma; carcinoma in situ (AIS), microinvasive adenocarcinoma (MIA), invasive adenocarcinoma (IAC)