While RNA-Seq has enabled great progress towards the goal of wide-scale isoform-level mRNA quantification, short reads have limitations when resolving complex or similar sets of isoforms. As a result, estimates of isoform abundance carry far more uncertainty than those made at the gene level. When confronted with this uncertainty, commonly used methods produce estimates that are often high-variance—small perturbations in the data often produce dramatically different results, confounding downstream analysis.
Researchers at the University of Washington introduce a new method, Isolator, which analyzes all samples in an experiment in unison using a simple Bayesian hierarchical model. Combined with aggressive bias correction, it produces estimates that are simultaneously accurate and show high agreement between samples. In a comprehensive comparison of accuracy and variance, the researchers show that this property is unique to Isolator. They further demonstrate that the approach of modeling an entire experiment enables new analyses, which they demonstrate by examining splicing monotonicity across several time points in the development of human cardiomyocyte cells.
A heatmap showing pairwise proportionality correlation between samples sequenced on two flowcells each at five sites, from centrally prepared libraries. Flowcells are numbered arbitrarily 1 or 2 and sequencing sites are abbreviated with three letter codes: Australian Genome Research Facility (AGR), Beijing Genome Institute (BGI), Cornell University (CNL), Mayo Clinic (MAY), and Novartis (NVS). Median proportionality correlation is listed below each heatmap. b The absolute change in correlation induced by enabling bias correction for methods that support it. For clarity this plot excludes points for BitSeq estimates of ”MAY 2”, as bias correction has an extremely detrimental effect on these. Mean improvement in correlation was 0.008 with Salmon, 0.007 with Cufflinks, 0.006 with Isolator, 0.003 with Kallisto, 0.002 with eXpress, and -0.164 with BitSeq.
Availability – Isolator is available under a permissiveopen source license at: https://github.com/dcjones/isolator