TCGA Biomarkers Identification Using Machine Learning - Complete Walkthrough

This workshop demonstrates how to download and preprocess TCGA prostate cancer gene-expression data in R and RStudio, then train a simple Keras/TensorFlow neural network to identify candidate biomarkers associated with Gleason scores. It also covers weight inspection, gene extraction, and gene set enrichment analysis, while discussing limitations of the approach.

Syllabus

Introduction and background
Chapter 1 - Installing packages and importing libraries
Using TCGA Biolinks
Structuring Input data and filtering
PlotMDS from limma and edgeR
Normalization of data
PCA Analysis
Making Train Label and One -hot Encoding
Chapter 2 - Neural network construction
Neural networking Training model fitting
Saving Model as hdf5 files
Extraction weights and bias
Extraction of GOI using weights and bias
Chapter 3 - Gene set enrichment analysis
Results!!!!!!
Some major issues with this approach