Characterization of TGFβ-associated molecular features and drug responses in gastrointestinal adenocarcinoma

Abstract Background Gastrointestinal adenocarcinoma (GIAD) has caused a serious disease burden globally. Targeted therapy for the transforming growth factor beta (TGF-β) signaling pathway is becoming a reality. However, the molecular characterization of TGF-β associated signatures in GIAD requires further exploration. Methods Multi-omics data were collected from TCGA and GEO database. A pivotal unsupervised clustering for TGF-β level was performed by distinguish status of TGF-β associated genes. We analyzed differential mRNAs, miRNAs, proteins gene mutations and copy number variations in both clusters for comparison. Enrichment of pathways and gene sets were identified in each type of GIAD. Then we performed differential mRNA related drug response by collecting data from GDSC. At last, a summarized deep neural network for TGF-β status and GIADs was constracted. Results The TGF-βhigh group had a worse prognosis in overall GIAD patients, and had a worse prognosis trend in gastric cancer and colon cancer specifically. Signatures (including mRNA and proteins) of the TGF-βhigh group is highly correlated with EMT. According to miRNA analysis, miR-215-3p, miR-378a-5p, and miR-194-3p may block the effect of TGF-β. Further genomic analysis showed that TGF-βlow group had more genomic changes in gastric cancer, such as TP53 mutation, EGFR amplification, and SMAD4 deletion. And drug response dataset revealed tumor-sensitive or tumor-resistant drugs corresponding to TGF-β associated mRNAs. Finally, the DNN model showed an excellent predictive effect in predicting TGF-β status in different GIAD datasets. Conclusions We provide molecular signatures associated with different levels of TGF-β to deepen the understanding of the role of TGF-β in GIAD and provide potential drug possibilities for therapeutic targets in different levels of TGF-β in GIAD