IntAPT

IntAPT reconstructs phenotype-specific transcripts from high-throughput RNA-seq data using a two-layer Bayesian model to identify and quantify alternatively spliced isoforms across multiple samples.

Key Features:

• Two-layer Bayesian model: Models isoform existence at the group level and isoform abundance at the sample level across multiple RNA-seq profiles.
• Spike-and-slab prior: Enforces sparsity in expressed isoforms and explicitly models dependencies between isoform existence and expression levels.
• Gibbs sampling: Uses iterative Gibbs sampling to estimate model parameters and infer the joint posterior distribution.
• Isoform identification and quantification: Identifies alternatively spliced isoforms and quantifies their abundance across samples.
• Robustness to sequencing errors: Maintains performance in the presence of sequencing noise and sample variability.
• Detection of low-abundance isoforms: Demonstrates improved identification of low-abundance expressed isoforms compared to existing methods.
• Implementation: Developed in C++.

Scientific Applications:

• Phenotype-specific transcriptome reconstruction: Reconstructs transcripts that are specific to a given phenotype from combined RNA-seq datasets.
• Alternative splicing analysis: Detects and characterizes alternatively spliced isoforms associated with phenotypes.
• Comparative isoform expression analysis: Quantifies isoform abundance to compare expression patterns across samples and groups.
• Gene regulation and molecular mechanism studies: Supports investigation of gene regulation and expression patterns linked to specific phenotypes, including discovery of novel isoforms.

Methodology:

Integrates multiple high-throughput RNA-seq profiles using a two-layer Bayesian framework (group-level existence and sample-level abundance), applies a spike-and-slab prior to enforce sparsity and model dependencies, and uses Gibbs sampling for iterative parameter estimation to infer the joint posterior distribution; implemented in C++.