A model based criterion for gene expression calls using RNA-seq data.

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Deep sequencing technologies like RNA-seq offer high sensitivity but create ambiguity in distinguishing low gene expression from transcriptional noise.
• Traditional RNA quantification methods often struggle with low sensitivity, making it difficult to separate true signals from background noise.

Purpose of the Study:

• To develop a statistical model for RNA sequencing (RNA-seq) data to accurately distinguish between actively transcribed and inactive genes.
• To establish a reliable criterion for classifying genes as expressed or non-expressed based on RNA-seq read counts.

Main Methods:

• A statistical model was proposed, treating RNA-seq read counts as a mixture of two distributions: exponential for inactive genes and negative binomial for active genes.
• The model was applied to multiple RNA-seq datasets to assess its fit and the consistency of derived criteria.

Main Results:

• The proposed statistical model demonstrated a strong fit across various RNA-seq datasets.
• A consistent criterion emerged: genes with more than two transcripts per million (TPM) are highly likely to be actively transcribed.
• This criterion aligns with previously proposed thresholds (e.g., 1 RPKM), validating the model's findings.

Conclusions:

• The statistical model effectively differentiates between expressed and non-expressed genes in RNA-seq data.
• The established criterion ( > 2 TPM) provides an operational method for interpreting RNA-seq results and classifying gene expression levels.
• This approach facilitates more robust analysis and interpretation of gene expression profiles from RNA-seq studies.