MSIQ: Joint modeling of multiple RNA-seq samples for accurate isoform quantification

Abstract

Next-generation RNA sequencing (RNA-seq) technology has been widely used to assess full-length RNA isoform abundance in a high-throughput manner. RNA-seq data offer insight into gene expression levels and transcriptome structures, enabling us to better understand the regulation of gene expression and fundamental biological processes. Accurate isoform quantification from RNA-seq data is challenging due to the information loss in sequencing experiments. A recent accumulation of multiple RNA-seq data sets from the same tissue or cell type provides new opportunities to improve the accuracy of isoform quantification. However, existing statistical or computational methods for multiple RNA-seq samples either pool the samples into one sample or assign equal weights to the samples when estimating isoform abundance. These methods ignore the possible heterogeneity in the quality of different samples and could result in biased and unrobust estimates. In this article, we develop a method, which we call “joint modeling of multiple RNA-seq samples for accurate isoform quantification” (MSIQ), for more accurate and robust isoform quantification by integrating multiple RNA-seq samples under a Bayesian framework. Our method aims to (1) identify a consistent group of samples with homogeneous quality and (2) improve isoform quantification accuracy by jointly modeling multiple RNA-seq samples by allowing for higher weights on the consistent group. We show that MSIQ provides a consistent estimator of isoform abundance, and we demonstrate the accuracy and effectiveness of MSIQ compared with alternative methods through simulation studies on D. melanogaster genes. We justify MSIQ’s advantages over existing approaches via application studies on real RNA-seq data from human embryonic stem cells, brain tissues, and the HepG2 immortalized cell line. We also perform a comprehensive analysis of how the isoform quantification accuracy would be affected by RNA-seq sample heterogeneity and different experimental protocols.

scINSIGHT for interpreting single-cell gene expression from biologically heterogeneous data

“…Rapid advances of scRNA-seq technologies have resulted in the generation of large-scale single-cell gene expression datasets from different platforms in different laboratories [ 6 , 7 ], using samples that span a broad range of species, tissue types, and experimental conditions [ 8 – 10 ]. The increasing number of scRNA-seq datasets emphasizes the need for integrative biological analysis to help assess and interpret similarities and differences between single-cell samples and to obtain in-depth insights into the underlying biological systems [ 11 – 13 ]. For example, integrative analysis of human and mouse transcriptomes has identified conserved cell types and transcription factors in pancreatic cells [ 14 ]; integrative analysis of scRNA-seq data from multiple melanoma tumors has identified a resistance program in malignant cells that is associated with T cell exclusion and immune evasion [ 15 ].…”

Modeling and analysis of RNA‐seq data: a review from a statistical perspective

“…Even though the point estimates of expression levels have led to new scientific discoveries in many biological studies, it is important to consider estimation uncertainty, especially when the differential expression analysis is of interest, or when some candidate isoforms are highly similar in structures (related to the collinearity issue in linear model estimation). One way to evaluate the uncertainty in Bayesian methods is to construct posterior or credible intervals of the estimated abundance levels [63,75]. In regression‐based methods, it is possible to calculate the standard errors of the abundance estimates (the coefficients in regression models).…”

“…There have been multiple efforts to quantify transcripts for better accuracy based on multiple RNA‐seq samples (especially biological replicates), thanks to reduced sequencing costs and the rapid accumulation of publicly available RNA‐seq samples. Model‐based methods include CLIIQ [76], MITIE [77], FlipFlop [78], and MSIQ [63]. These methods generalize the models designed for isoform quantification based on a single sample, and their results show that aggregating the information from multiple samples can achieve better accuracy in isoform abundance estimation.…”

“…These methods generalize the models designed for isoform quantification based on a single sample, and their results show that aggregating the information from multiple samples can achieve better accuracy in isoform abundance estimation. It has been noted in MSIQ [63] that it is important to consider the possible heterogeneity in the quality of different samples to obtain robust and accurate estimation results.…”

“…Some other methods, including WemIQ [60], Salmon [61], iReckon [62], and MSIQ [63], introduce hidden variables to denote the isoform origins of reads and use these variables to simplify the form of the likelihood function. Suppose that the isoform origins of reads are denoted as , where if read comes from isoform j .…”