vborjesson / gene_expression

Analysis of RNA seq data to explore gene expression in different types of cancer

Gene Expression

---

Abstract

An overview of RNA sequencing and microarray gene expression analysis. This analysis includes exploring, visualizing and clustering the gene expression data by hierarchically-clustered heatmap, principal component analysis (PCA) and t-distributed Stochastic Neighbor Embedding (t-SNE). The analysis also includes code for cancer classification by using supervised machine learning on gene expression data.

Data

The RNA-sequencing gene expression level dataset is a fraction of the Illumina HiSeq dataset from the cancer genome atlas (TCGA) pan-cancer analysis project (Weinstein, J.N. et al. 2013), and was retrieved from UCI machine learning repository. The data contains 801 observations, expression levels of 20532 genes, and labels of five cancers; breast (BRCA), kidney (KIRC), lung adeno (LUAD), colon (COAD), and prostate (PRAD).

Analysis

The data was first explored by performing hierarchically-clustering, and visualizing the result using a heatmap.

Dimensionality reduction

This dataset contains gene expressions of 20532 genes, corresponding to 20532 dimensions. In order to get an intuition of the data and cluster the data without label information, a dimensionality reduction is performed using PCA and t-SNE.

PCA was performed using 2 principal components.

t-SNE

Supervised machine learning for cancer type classification

The dataset was split 70/30, train/test dataset. The methods that was used was K-Nearest neighbors classifier, Support Vector Machines (SVM) and Decision Tree (see figure below)

Weinstein, J. N. et al. (2013), 'The Cancer Genome Atlas Pan-Cancer analysis project', Nat Genet 45 (10), 1113-1120.