IGF2BP2-related modification patterns in pancreatic cancer: A machine learning-driven approach towards personalized treatment

Abstract

Pancreatic cancer (PC) is a malignant digestive system tumor with a very poor prognosis. N6-methyladenosine (m6A) is mediated by a variety of readers and participates in important regulatory roles in PC. Based on TCGA_PAAD, ICGC_AU_PAAD, ICGC_CA_PAAD, GSE28735 and GSE62452 datasets, We mapped the multi-omics changes of m6A readers in PC and found that m6A readers, especially IGF2BP family genes, had specific changes and were significantly associated with poor prognosis. An unsupervised consensus clustering algorithm was used to explore the correlation between specific expression patterns of m6A readers in PC and enrichment pathways, tumor immunity and clinical molecular subtypes. Then, the principal component analysis (PCA) algorithm was used to quantify specific expression patterns and screen core genes. Machine learning algorithms such as Bootstrapping and RSF were used to quantify the expression patterns of core genes and construct a prognostic scoring model for PC patients. What's more, pharmacogenomic databases were used to screen sensitive drug targets and small molecule compounds for high-risk PC patients in an all-around and multi-angle way. Our study has not only provided new insights into personalized prognostication approaches, but also thrown light on integrating tailored risk stratification with precision therapy based on IGF2BP2-mediated m6A modification patterns.