Interpretable Machine Learning for Biomarker Identification in RNA Seq Cancer Data

Aug 3, 2026

Existing research on RNA-Seq gene-expression biomarkers uses statistical methods to shrink thousands of genes into a compact cancer-marker set, while machine learning often adds predictive selection. This thesis trains inherently interpretable models—Logistic Regression, Random Forest, and Linear SVM—on TCGA data across 16 cancer types, then refines candidates using model-agnostic iML such as SHAP and Permutation Importance. iML-selected biomarkers reach comparable or better classification performance than external or purely statistical biomarkers, with MI proving especially effective and pathway/annotation checks supporting validity.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages143

Reading Time6 h 0 min

AccessFree to download

Chapter 1: Introduction

Motivation

Background

Experiment Goals

Chapter 2: Related Work

Early Work in machine learning and cancer

Current State of Research

Biomarkers from Previous Studies

Chapter 3: Methodology

Overview

Data Source

Exploratory Data Analysis (EDA)

Statistics

Mutual Information

Step 1: Classification Using All Genes

Step 2: Classification using 250 Genes, and Additional Feature Importance

Step 3: Gene-set Classification Performance Evaluation

Step 4 Biomarker Analysis

Chapter 4: Results

Step 1 Results

Step 2 results

Step 3 results

Step 4: Potential Biomarker Evaluation

Discussion

Chapter 5: Conclusion

Bibliography

Appendix A: Appendix A

©Copyright 2025 Jeremy Newton Interpretable Machine Learning for Biomarker Identification in RNA Seq Cancer Data Jeremy Newton A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science University of Washington 2025 Reading Committee: Wooyoung Kim Douglas Wacker Dharma Dailey Program Authorized to Offer Degree: Computer Science & Software Engineering University of Washington Abstract Interpretable Machine Learning for Biomarker Identification in RNA Seq Cancer Data Jeremy Newton Chair of the Supervisory Committee: Wooyoung Kim Department of Computing and Software Systems Existing research on RNA Seq gene expression biomarkers has provided various methods to select a small list of genes as cancer biomarkers from a large number of gene expression data. Previous methods for identifying potential gene expression cancer biomarkers have focused on statistical analysis, but other methods have incorporated machine learning, often including Interpretable Machine Learning (iML) techniques. On 16 cancer types from TCGA data, we used inherently interpretable machine learning models: Logistic Regression, Random Forest, and Linear Support Vector Machine to narrow down subsets of potential genes as biomarkers using the trained models’ feature importance rankings. We subsequently applied model-agnostic iML techniques, such as Shapley Additive Explanations (SHAP) and Permutation Importance, to narrow down the subsets even further. We compared classification performance between machine learning models trained on iML selected features with features selected by statistical methods, and biomarkers from external research. We found that iML biomarker selection methods lead to comparable or better classification performance on these datasets than the biomarkers from outside research, or from statistical analysis alone. Mutual Information estimation (MI) was a surprisingly useful technique for initial feature selection, and iML techniques improved the MI selected features for classification. We cross-checked potential biomarkers with biomedical annotations and gene pathway analysis, finding some support for the validity of the biomarkers. TABLE OF CONTENTS Page Abstract ........................................... i Table of Contents ...................................... ii List of Figures ....................................... iii List of Tables ........................................ xiii Glossary ........................................... xiv Acknowledgments ...................................... xx Chapter 1: Introduction ................................ 1 1.1 Motivation ..................................... 1 1.2 Background .................................... 2 1.3 Experiment Goals ................................. 4 Chapter 2: Related Work ................................ 6 2.1 Early Work in machine learning and cancer ................... 6 2.2 Current State of Research ............................ 6 2.3 Biomarkers from Previous Studies ....................... 7 Chapter 3: Methodology ................................ 12 3.1 Overview ...................................... 12 3.2 Data Source .................................... 14 3.3 Exploratory Data Analysis (EDA) ....................... 14 3.4 Statistics ..................................... 17 3.5 Mutual Information ................................ 20 3.6 Step 1: Classification Using All Genes ...................... 22 3.7 Step 2: Classification using 250 Genes, and Additional Feature Importance . 30 3.8 Step 3: Gene-set Classification Performance Evaluation ............ 31 3.9 Step 4 Biomarker Analysis ............................ 35 Chapter 4: Results ................................... 38 4.1 Step 1 Results ................................... 38 4.2 Step 2 results ................................... 47 4.3 Step 3 results .......................

FAQs

Which interpretable machine learning models are used for biomarker selection?

Logistic Regression, Random Forest, and Linear Support Vector Machine are trained to rank feature importance and narrow gene subsets as potential biomarkers.

How does the thesis further refine biomarker candidates after training the models?

It applies model-agnostic iML techniques, including SHAP (Shapley Additive Explanations) and Permutation Importance, to reduce candidate subsets further based on interpretability-driven feature contributions.

How does iML-based biomarker selection compare with statistical methods and external biomarkers?

On TCGA datasets from 16 cancer types, iML biomarker selection achieves comparable or better classification performance than external research biomarkers and than statistical analysis alone.

Aug 3, 2026

Upload by Valentina