Explainable Multi-Omics AI Framework for Breast Cancer Stage Classification: Integrating Gene Expression, Proteomics, and Mutation Data

Samala, Sriram (2025) Explainable Multi-Omics AI Framework for Breast Cancer Stage Classification: Integrating Gene Expression, Proteomics, and Mutation Data. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (637kB) | Preview
Preview	PDF (Configuration Manual) Download (1MB) | Preview

Abstract

Detecting cancer at the earliest possible stage is a key factor in enhancing patient outcomes through timely intervention. This study devised an explainable AI framework integrating multi-omics of gene expression, somatic mutations, and proteomics from large clinically annotated cohorts of breast cancer patients. Data was processed through strict data cleaning, imputation, log-transformation, and standardization. Dimensionality reduction techniques employed both statistical feature selection and deep autoencoder representations. The XGBoost classifier trained using class balancing distinguished between early (Stage I/II) and late (Stage III/IV) stage cancer with nearly perfect performance (ROC-AUC ~0.99). Model interpretability was obtained through SHAP and LIME, indicating all molecular contributors obtained from each layer of omics data.

Model evaluation utilized stratified cross-validation and lays the conceptual groundwork for future external validation that can employ independent proteogenomic datasets. Furthermore, an autoencoder-based anomaly detection method was demonstrated to identify tumors in an unsupervised manner. Limitations include the lack of sequence-based transformer models that were outside of the scope for this study, given the state of the available data, which will be considered for future work. The explainable multi-omics AI framework represents a significant step forward in being able to identify the precise breast cancer stage.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Vamadevan, Arundev UNSPECIFIED
Uncontrolled Keywords:	Multi-omics; Liquid Biopsy; Explainable AI; XGBoost; SHAP; LIME; Gene Expression; Somatic Mutation; Proteomics; Autoencoder; Breast Cancer; Early-Stage Detection
Subjects:	Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence Q Science > Life sciences > Medical sciences > Pathology > Tumors > Cancer R Medicine > Healthcare Industry
Divisions:	School of Computing > Master of Science in Artificial Intelligence
Depositing User:	Ciara O'Brien
Date Deposited:	12 Aug 2026 10:04
Last Modified:	12 Aug 2026 10:04
URI:	https://norma.ncirl.ie/id/eprint/9515

Actions (login required)

View Item