Workshop 2: Unsupervised approaches to uncover and interpret transcriptional patterns in high-dimensional transcriptomic data
$0.00
Description
Workshop Synopsis:
Gene expression profiles derived from tumor biopsies are often confounded by cellular heterogeneity and technical artifacts. Since biopsies contain both tumor and non-tumor cells, the measured expression levels represent an average across diverse cell populations. Consequently, subtle but biologically relevant signals—such as those driven by copy number alterations or weak transcriptional programs—can be overshadowed by dominant signals from abundant or highly variable components. This necessitates the use of unsupervised approaches that can isolate statistically independent sources of variation to reveal meaningful transcriptional patterns.
In this workshop, we will present methods to: (i) detect batch effects without prior annotations; (ii) mitigate these effects in an unsupervised manner; (iii) extract statistically independent transcriptional patterns; (iv) interpret these patterns using gene set enrichment and cofunctionality frameworks; and (v) link pattern activity to known phenotypes. We will also demonstrate how to identify and characterize phenotype-specific cohorts and evaluate cross platform generalizability
Intended audience:
This workshop is designed for bachelor’s/master’s students, PhD students, postdoctoral researchers, and data scientists working in the fields of bioinformatics, computational biology,and transcriptomics. A very basic understanding of transcriptomic technologies (e.g., RNA-seq or microarray data), dimensionality reduction, and statistical analysis is helpful. No prior experience with unsupervised learning is required, although familiarity with R will be helpful for hands-on sessions.
