RNA-Seq Cancer Subtype Identification

Breast cancers differ in their gene expression and in how they behave. This project finds subtypes in TCGA-BRCA RNA-seq without labels by consensus clustering (k-means, hierarchical and NMF base clusterers, PAC criterion), assigns new tumours to them with a linear SVM, checks that they replicate in held-out TCGA tumours and in an independent Swedish cohort (GEO GSE81538), compares PAM50 classifiers across platforms and describes the survival of the subtypes. The most stable structure is two subtypes - ER+ luminal and ER- basal - which replicate (ARI 0.91 and 0.95); for PAM50 the simple nearest-centroid rule beats every SVM, and cross-platform scaling matters more than the classifier.

Category: Bioinformatics - Transcriptomics and Gene-Expression Analysis

Transcriptomics & Gene-Expression Analysis Project B17 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.