Identifiability of isoform deconvolution from junction arrays and RNA-Seq

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Quantifying splice variants is crucial for understanding gene expression.
• Current methods like splice junction microarrays and RNA-sequencing (RNA-seq) face challenges in determining isoform expressions from exon and splice junction data.
• The extent of this identifiability problem across different platforms and potential solutions remain underexplored.

Purpose of the Study:

• To propose criteria ensuring the identifiability of isoform deconvolution models.
• To evaluate the performance of these criteria on various gene quantification platforms, including RNA-seq, exon arrays, and splice junction arrays.
• To assess the identifiability of gene models for alternatively spliced genes.

Main Methods:

• Development of criteria for guaranteed identifiability in isoform deconvolution models.
• Application and evaluation of these criteria on RNA-seq data.
• Comparative analysis of model identifiability across different microarray platforms (Human Exon array, splice junction array) and RNA-seq.
• Utilizing a dataset of 2256 alternatively spliced human genes from the RefSeq database.

Main Results:

• RNA-sequencing (RNA-seq) demonstrated high identifiability, with up to 97% of tested genes yielding identifiable models.
• Human Exon arrays showed significantly lower identifiability, with only 26% of genes having identifiable models.
• Even comprehensive splice junction arrays achieved only 69% identifiability, highlighting platform-specific limitations.

Conclusions:

• RNA-sequencing offers superior performance for isoform deconvolution compared to current microarray technologies.
• The proposed criteria are effective in guaranteeing model identifiability, particularly in RNA-seq data.
• Further research may be needed to improve identifiability on microarray platforms or explore alternative analytical approaches.