Impact of Gene Duplicate Handling Strategies on Classification Performance and Feature Selection in Gene Expression Data
Year: 2025
Type: conference-paper
Abstract: A typical gene expression dataset includes a large number of features (i.e., genes) and a limited number of samples. Gene duplicates are likely to exist in the data as one gene may produce multiple mRNAs, and in fact a gene corresponds to the expression level measured for a specific mRNA in a gene expression dataset. In this study, we seek to analyze the effects of duplicate genes under different duplicate handling settings. We considered averaging the duplicates (mean setting), selecting duplicates with the highest variance (variance setting), and leaving duplicates unmodified (unmodified setting). The experimental results show similar classification performance and feature numbers in the three settings. However, we also observe that variance and unmodified settings are safer choices in determining significant genes in a gene selection task. The findings of this study can guide the handling of gene duplicates in gene expression datasets.
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W4415524813