Computational Characterization of Undifferentially Expressed Genes with Altered Transcription Regulation in Lung...

Area of Science:

• Genomics
• Bioinformatics
• Cancer Research

Background:

• Transcriptome analysis typically focuses on individual gene expression.
• Conventional methods overlook potential biomarkers that lack differential expression.
• Inter-feature independence is a common assumption, limiting biomarker discovery.

Purpose of the Study:

• To propose a novel computational protocol, mqTrans, for analyzing transcriptomes.
• To identify novel lung cancer biomarkers by examining high-dimensional inter-feature correlations.
• To uncover 'dark biomarkers'—genes with altered expression patterns not detectable by standard differential analysis.

Main Methods:

• Developed the mqTrans protocol to analyze transcriptomes based on inter-feature correlations.
• Trained a regression model to predict mRNA expression from transcription factor (TF) expression.
• Defined mqTrans features as the difference between predicted and actual mRNA expression.

Main Results:

• Identified thirteen transcriptomic features with differentially expressed mqTrans features in three independent lung cancer datasets.
• These features showed no differential expression in original transcriptomic values, hence termed 'dark biomarkers'.
• GBP5 was discussed as a potential dark biomarker, with long non-coding RNAs suggested as a contributing factor.

Conclusions:

• The mqTrans approach successfully identified previously overlooked potential biomarkers in lung cancer.
• Dark biomarkers, characterized by significant mqTrans values but not differential expression, are strongly associated with lung cancer.
• This method enhances biomarker discovery by considering complex gene expression correlations.