A hierarchical Bayesian model for comparing transcriptomes at the individual transcript isoform level.

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Mammalian transcriptomes exhibit complexity due to alternative splicing, producing multiple transcript isoforms from a single gene.
• Current transcriptome comparison methods primarily focus on gene-level differential analysis, neglecting individual splice variant expression.
• High-throughput sequencing and tiling arrays offer potential for splice variant-level analysis, but read coverage can represent multiple isoforms.

Purpose of the Study:

• To develop a statistical method for inferring differential expression at the individual transcript isoform level.
• To enable accurate transcriptome comparison beyond gene-level analysis by focusing on splice variants.
• To address the challenge of ambiguous read coverage representing multiple splice variants.

Main Methods:

• Proposed a hierarchical Bayesian model named BASIS (Bayesian Analysis of Splicing IsoformS).
• Introduced a latent variable for statistical selection of differentially expressed isoforms.
• Employed a Gibbs sampler to infer model parameters via an ergodic Markov chain.
• Developed a method that leverages information across probes and genes, handling heteroskedasticity in data.

Main Results:

• BASIS successfully infers differential expression levels for individual transcript isoforms.
• The model effectively borrows information across probes and genes, enhancing statistical power.
• Applied BASIS to human tiling-array and mouse RNA-seq data, yielding testable predictions.
• Validated some BASIS predictions using quantitative real-time RT-PCR experiments.

Conclusions:

• BASIS provides a robust framework for analyzing differential isoform expression, advancing transcriptome analysis.
• This method overcomes limitations of existing approaches by enabling splice variant-specific comparisons.
• BASIS has significant implications for understanding gene regulation and functional genomics in complex organisms.