Collection and Pretreatment of Publicly Available Expression Data Sets

Clinical features of LUAD and gene expression data were collected retrospectively from TCGA (https://cancergenome.nih.gov/) and NCBI GEO databases (https://www.ncbi.nlm.nih.gov/geo/). Twenty-three m6A regulatory factors were obtained (Supplementary Table S1) (Chen et al., 2019b; Li et al., 2021b; Zhou et al., 2020; Liu et al., 2020b; Arguello et al., 2017; Warda et al., 2017; Barros-Silva et al., 2020; Wu et al., 2021b). TCGA somatic mutation data had been captured through the utilization of TCGAbiolinks R package (Colaprico et al., 2016) and visualized through application of the maftools R package (Mayakonda et al., 2018). The copy number variation (CNV) datasets were obtained from the Xena Public Data Center (https://xenabrowser.net). The copy number variation landscape of these 23 m6A regulators in human chromosomes was identified using R package “Rcircos”. TCGA–LUAD RNA-seq data (FPKM format) were downloaded from the Genomic Data Commons (GDC, https://portal.gdc.cancer.gov/) and converted into transcripts per kilobase million (TPM) format. The Gene Expression Omnibus (GEO) database was used to make a comprehensive query on all qualified LUAD data sets. A total of seven datasets [{"type":"entrez-geo","attrs":{"text":"GSE30219","term_id":"30219"}}GSE30219 (Xie et al., 2011), {"type":"entrez-geo","attrs":{"text":"GSE30219","term_id":"30219"}}GSE30219 (Rousseaux et al., 2013), {"type":"entrez-geo","attrs":{"text":"GSE31210","term_id":"31210"}}GSE31210 (Okayama et al., 2012), {"type":"entrez-geo","attrs":{"text":"GSE37745","term_id":"37745"}}GSE37745 (Botling et al., 2013), {"type":"entrez-geo","attrs":{"text":"GSE50081","term_id":"50081"}}GSE50081 (Der et al., 2014), {"type":"entrez-geo","attrs":{"text":"GSE68465","term_id":"68465"}}GSE68465 (Shedden et al., 2008), and {"type":"entrez-geo","attrs":{"text":"GSE72094","term_id":"72094"}}GSE72094 (Schabath et al., 2016)] were obtained to represent different LUAD-independent studies. All datasets contained clinical data and survival information. The platform files and survival data of these datasets are shown in Supplementary Table S2. In accordance with the corresponding annotation file, the probe was transformed into a gene symbol. For genes with set signals of multiple probes, their values were averaged to produce a single gene expression value. At last, they were further combined to a meta-queue through the “ComBat” algorithm using the “sva” package (Leek et al., 2012) to decrease the batch effect from abiotic deviations.