Fig. 4

Construction and verification of an IRGS-S for LUAD patients. A-B These plots visualize the Lasso regression process of the 24 prognostic-correlated IRGs. C A forest plot reflecting a multivariate Cox regression analysis of the candidate genes in the TCGA. D–H Theses Kaplan–Meier survival curves revealed the OS differences between high- and low-IRGS groups in the TCGA D, GSE42127 E, GSE68465 F, GSE72094 G and merge-GEO H cohorts. GSE29013, GSE30219, GSE31210, GSE37745, and GSE50081 were from a same chip platform and thus integrated to a new cohort (meta cohort) using the ‘ComBat’ algorithm of ‘sva’ R package to reduce the batch effects from non-biological technical biases. I–M The ROC curves of the IRGS-S I, the sigantures from Al-Dherasi et al. J, Zhang et al. K, Shi et al. L, Yi et al. M. N Comparison of the C-index for each model