engleski

Computational subtyping of mouse breast tumors

Studies of gene expression patterns of breast tumors derived from cDNA microarrays reported their distinctive molecular portraits according to which tumors can be classified into five intrinsic subtypes with distinct clinical outcomes: luminal A, luminal B, HER2 enriched, basal and normal- like. PAM50 is a molecular classifier developed by minimizing expanded intrinsic gene set to the top 50 genes that contribute to distinguishing intrinsic subtypes. The R/Bioconductor package genefu implements bioinformatics algorithms and gene signatures for molecular subtyping of breast cancer, including PAM50 molecular classifier. In this project, an algorithm from package genefu was modified to subtype breast tumors using RNA-Seq instead of microarray data as an input and then used to subtype RNA sequenced mouse breast tumors in relation to human tumors. Furthermore, a recent study showed that human and mouse data can be integrated using canonical correlation analysis from package Seurat. The motivation for integrating diverse datasets lies in potential to use information from one dataset for the interpretation of another. In this project canonical correlation analysis was also used to integrate human and mouse bulk RNA-Seq data based on the set of PAM50 genes and used to determine intrinsic breast tumor subtypes. Finally, results of genefu and Seurat subtyping were compared and the performance of the two independent approaches was assessed.

breast cancer, tumor heterogeneity, PAM50, intrinsic subtypes, clustering

nije evidentirano