Multi-omics Integrative Analysis And In Silico Drug Screening In Subtypes Of KRAS-mutant Lung Adenocarcinoma

In lung adenocarcinoma,KRAS mutations account for 30%of the total.Due to the tertiary structure and intracellular location of the KRAS mutant protein,it has been become a protein that is difficult to target.Unlike EGFR mutation that occurring in lung adenocarcinomas,KRAS-mutant lung adenocarcinomas(LUADs)are heterogeneous and frequently occur in smokers.Whether targeting downstream proteins or immunotherapy,only a subset of KRAS-mutant lung adenocarcinoma patients responding to the therapy.The heterogeneity of KRAS-mutant LUAD has been an obstacle for the drug discovery.We integrated multiplatform datatypes which including mRNA expression,DNA methylation,somatic mutation,miRNA expression and somatic copy number alteration and identified two subtypes in 128 KRAS-mutant LUAD patients.At the same time,we screened out three data types that contributed the most in data integration,including DNA methylation,gene expression,and somatic mutation.We integrated these three data types of KRAS-mutated lung adenocarcinoma cell lines and identified two different molecular subtypes that consistent with the subtypes of KRAS-mutant LUAD patients.We further characterized the features of these two subtypes in multiplatform.In DNA methylation level,Patient-Subtype 2(PS2)was hypermethylated compared with Patient-Subtype 1(PS1).Meanwhile,PS2 has higher smoking-related methylation signature acitivity than PS1.In gene expression level,genes significantly highly expressed in PS1 enriched cell cycle,poor survival and metabolism pathways,while genes significantly highly expressed in PS2 enriched KRAS signal,immune response pathways.Otherwise,TIL-associating score and KRAS dependency score of PS2 are both higher than PS1.In somatic mutation level,we found that smoking-related mutational signature that characterized by C>A base substitution was higher in PS1 than PS2.STK11 and KEAP1 mutations were significantly enriched within PS1.However,KRAS mutation types had no significant difference between two subtypes.The cell line subtypes faithfully recapitulated all the patients’ features.Finally,we performed drug screening to identify subtype-specific drugs and used the defining features of the KRAS subtypes for drug sensitivity prediction.Drug screening of the two cell line subtypes yielded 13 potential candidates.12 drugs including Cytarabine and Enzastaurin showed better efficacy in CS1 than CS2.The only drug that is sensitive to CS2 is BTK inhibitor,QL-XII-61.Based on the subtype-specific drugs.We build linear model to predict drug sensitivity using above biological features,such as smoking-related methylation signature and mutational signature.We found that smoking-related methylation signature might be used as strong biomarkers for drug sensitivity prediction.