DAFS: a data-adaptive flag method for RNA-sequencing data to differentiate genes with low and high expression

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Next-generation sequencing (NGS) enables high-throughput analysis of genetic and genomic variation.
• RNA-sequencing (RNA-seq) detects transcripts across a wide dynamic range, including low expression levels.
• Defining biologically meaningful expression thresholds in RNA-seq remains a challenge due to high variability and experimental factors.

Purpose of the Study:

• To develop and validate a data-adaptive method for distinguishing low and high expression in RNA-seq data.
• To address the lack of consensus on defining low expression levels.
• To improve the interpretation of transcript abundance classes in RNA-seq analysis.

Main Methods:

• A data-adaptive approach was developed using the Kolmogorov-Smirnov statistic and multivariate adaptive regression splines.
• The method determines an optimal cutoff for separating high and low expressed transcripts.
• Performance was validated against a two-component mixture distribution model using diverse RNA-seq datasets.

Main Results:

• The proposed data-adaptive method effectively estimates the lower bound of high expression.
• Robustness was demonstrated across datasets with varying sequencing depth, species, and data distributions.
• The method outperforms arbitrary cutoffs and highlights limitations of two-component mixture models for poorly separated expression classes.

Conclusions:

• Data-adaptive methodologies are superior to arbitrary cutoffs for classifying low and high RNA-seq expression.
• Accurate classification of stably expressed RNA-seq data is essential for robust data analysis.
• The developed algorithm offers a promising tool for RNA-seq data analysis and interpretation.