An OMIC biomarker detection algorithm TriVote and its application in methylomic biomarker detection

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Transcriptomic and methylomic data are crucial biomarkers for disease diagnosis and prognosis, influenced by genetics and environment.
• High-throughput profiling generates vast datasets, posing significant computational challenges for feature selection.

Purpose of the Study:

• To introduce TriVote, a novel three-step feature selection algorithm.
• To identify a subset of transcriptomic or methylomic residues for highly accurate binary classification.

Main Methods:

• Developed a three-step feature selection algorithm named TriVote.
• Applied TriVote to 17 transcriptomes and two methylomes for performance evaluation.

Main Results:

• TriVote demonstrated superior performance compared to existing filter and wrapper methods.
• Achieved higher classification accuracy with a reduced feature set across multiple datasets.
• Identified biologically relevant methylome biomarkers associated with diseases.

Conclusions:

• TriVote offers an effective solution for feature selection in large-scale transcriptomic and methylomic data.
• The algorithm enhances biomarker discovery for disease diagnosis and prognosis.
• A publicly available Python package facilitates the practical application of TriVote.