Bias correction and Bayesian analysis of aggregate counts in SAGE libraries

Area of Science:

• Transcriptomics
• Bioinformatics
• Statistical Genetics

Background:

• Tag-based RNA sequencing methods, such as SAGE, are widely used for transcriptome analysis.
• Incomplete digestion can lead to multiple tags per mRNA, causing biased sampling of transcript pools.
• Existing methods like SAGE often discard data to mitigate bias, reducing statistical power.

Purpose of the Study:

• To develop novel statistical models that directly address tag formation bias in transcriptome data.
• To improve the accuracy and statistical power of differential gene expression analysis.

Main Methods:

• Proposed three new hierarchical Bayesian models incorporating tag formation probability variation.
• Developed associated Bayesian inference algorithms for both tag and aggregate level data.
• Utilized simulation experiments and real data analysis to evaluate model performance.

Main Results:

• The proposed Bayesian models demonstrated superior performance compared to existing methods like DPB.
• Inference accuracy was influenced by gene expression levels and prior choice.
• The multivariate approach facilitates both univariate and joint differential expression tests.

Conclusions:

• The developed Bayesian algorithms effectively account for tag formation bias, enhancing transcriptome analysis.
• Failure to account for tag formation bias can lead to false positive and negative findings in differential expression testing.
• These methods offer a more powerful and accurate approach to analyzing tag-based sequencing data.