Computational Approaches for Disease Gene Identification

TL;DR

This paper reviews and proposes computational models using semi-supervised and ensemble learning techniques to improve disease gene identification from human genome data, addressing challenges like data imbalance and biological complexity.

Contribution

It introduces three novel computational models that leverage ensemble and semi-supervised learning to enhance disease gene prediction accuracy.

Findings

Ensemble models outperform single-model approaches.

Semi-supervised methods effectively utilize unlabeled data.

Improved prediction accuracy demonstrated on biological datasets.

Abstract

Identifying disease genes from human genome is an important and fundamental problem in biomedical research. Despite many publications of machine learning methods applied to discover new disease genes, it still remains a challenge because of the pleiotropy of genes, the limited number of confirmed disease genes among whole genome and the genetic heterogeneity of diseases. Recent approaches have applied the concept of 'guilty by association' to investigate the association between a disease phenotype and its causative genes, which means that candidate genes with similar characteristics as known disease genes are more likely to be associated with diseases. However, due to the imbalance issues (few genes are experimentally confirmed as disease related genes within human genome) in disease gene identification, semi-supervised approaches, like label propagation approaches and positive-unlabeled learning, are used to identify candidate disease genes via making use of unknown genes for training - typically in the scenario of a small amount of confirmed disease genes (labeled data) with a large amount of unknown genome (unlabeled data). The performance of Disease gene prediction models are limited by potential bias of single learning models and incompleteness and noise of single biological data sources, therefore ensemble learning models are applied via combining multiple diverse biological sources and learning models to obtain better predictive performance. In this thesis, we propose three computational models for identifying candidate disease genes.