Net-RSTQ

Net-RSTQ integrates protein domain-domain interaction networks with short-read RNA-Seq alignments to improve isoform-level transcript quantification accuracy by leveraging correlated abundances among interacting isoforms.

Key Features:

• Network-based integration: Incorporates known protein domain-domain interaction networks as prior knowledge and leverages observed positive correlations among neighboring isoform abundances.
• Short-read alignment likelihoods: Uses likelihoods from short read alignments derived from high-throughput mRNA sequencing (RNA-Seq) against individual transcripts.
• Dirichlet priors for expression modeling: Models expression levels of neighboring transcripts using Dirichlet priors applied to the likelihood of observed read alignments.
• Joint estimation with alternating optimization: Employs an alternating optimization strategy that solves multiple Expectation-Maximization (EM) problems to jointly estimate transcript abundances across all genes.

Scientific Applications:

• Improved isoform quantification: Simulation studies show enhanced isoform transcript quantifications when isoform co-expression correlates with their interactions.
• Validation with qRT-PCR: Experimental validation by qRT-PCR on 25 multi-isoform genes across stem cells, ovarian cancer, and breast cancer cell lines produced more consistent isoform proportion estimates than traditional RNA-Seq analyses.
• Clinical relevance in TCGA datasets: Transcript abundances estimated by Net-RSTQ were more informative for patient sample classification in ovarian cancer, breast cancer, and lung cancer in TCGA RNA-Seq datasets.

Methodology:

Net-RSTQ integrates protein domain-domain interaction networks as priors, applies Dirichlet priors to the likelihoods of observed short-read alignments against transcripts, and uses alternating optimization solving multiple Expectation-Maximization (EM) problems for joint estimation of transcript abundances across genes.