Fig. 4

LNM classifier construction and the efficiency of the 12-ASE-based classifier. a Z-score of the top 30 important ASEs and 20 randomly picked ASEs using Boruta algorithm. b The mean cross-validation error of the five-round fivefold cross-validation about different numbers of ASEs. c The heat map showing PSI levels of ASEs in the LNM classifier. The data were normalized using R function scale. d ROC curves for the fivefold cross-validation of the classifier to identify LNM statuses of LUAD patients. LNM lymph node metastasis. Important ASEs, the ASEs confirmed as important features for the identification of LNM for LUAD patients by Boruta algorithm. Top 30 ASEs (rejected), the ASEs had the top 30 Z-score but rejected as unimportant features by Boruta algorithm for the identification of LNM for LUAD patients