EMDomics: a robust and powerful method for the identification of genes differentially expressed between heterogeneous classes

Abstract

Motivation: A major goal of biomedical research is to identify molecular features associated with a biological or clinical class of interest. Differential expression analysis has long been used for this purpose; however, conventional methods perform poorly when applied to data with high within class heterogeneity.Results: To address this challenge, we developed EMDomics, a new method that uses the Earth mover’s distance to measure the overall difference between the distributions of a gene’s expression in two classes of samples and uses permutations to obtain q-values for each gene. We applied EMDomics to the challenging problem of identifying genes associated with drug resistance in ovarian cancer. We also used simulated data to evaluate the performance of EMDomics, in terms of sensitivity and specificity for identifying differentially expressed gene in classes with high within class heterogeneity. In both the simulated and real biological data, EMDomics outperformed competing approaches for the identification of differentially expressed genes, and EMDomics was significantly more powerful than conventional methods for the identification of drug resistance-associated gene sets. EMDomics represents a new approach for the identification of genes differentially expressed between heterogeneous classes and has utility in a wide range of complex biomedical conditions in which sample classes show within class heterogeneity.Availability and implementation: The R package is available at http://www.bioconductor.org/packages/release/bioc/html/EMDomics.htmlContact: abeck2@bidmc.harvard.eduSupplementary information: supplementary data are available at Bioinformatics online.

Choice of library size normalization and statistical methods for differential gene expression analysis in balanced two-group comparisons for RNA-seq studies

“…Due to technical limitations, scRNAseq data generally have low library sizes resulting in a large fraction of 'dropout' events as well as huge heterogeneity, which introduces a major challenge in identification of DEGs. Given the special characteristics, many new methods have been developed especially for DE analysis of scRNA-seq data [57][58][59][60]. A few studies have compared DEG analysis tools for scRNA-seq data and found that existing methods for analysis of bulk RNA-seq data perform as well as, or not worse than, those specifically developed for scRNA-seq data in terms of the power and FDR [52,61].…”

Statistical Methods for RNA Sequencing Data Analysis

“…Single-cell RNA sequencing is becoming popular in recent years to better understand the stochastic process and gene regulations in a granular resolution (13,15,16,39). The commonly used gene differential expression analysis in single-cell RNA sequencing can be classified into two categories, with one category modeling excess zeros (SCDE, MAST, scDD, DEsingle, and SigEMD) (40)(41)(42)(43)(44) and the other category without modeling the excess zeros in the single-cell RNA sequencing data (DESeq2, SINCERA, D 3 E, EMDomics, Monocle2, Linnorm, and Discriminative Learning) (12,27,(45)(46)(47)(48)(49)(50). DESeq2 is a popular method used for bulk RNA sequencing data analysis, which is also often used for analyzing single-cell RNA sequencing data for testing of differential expression between groups.…”

“…EMDomics detects significantly differentially expressed genes between groups for single-cell RNA sequencing data by comparing the two distribution functions of gene expressions between groups (48). EMDomics compares the differences between groups based on EMD, a commonly used approach to compare two histograms in imaging analysis.…”

Data-based RNA-seq Simulations by Binomial Thinning

“…One of the more common applications of RNA-seq data is estimating and testing for differences in gene expression between two groups. Many packages and techniques exist to perform this task [Robinson and Smyth, 2007b, Hardcastle and Kelly, 2010, Van De Wiel et al, 2012, Kharchenko et al, 2014, Law et al, 2014, Love et al, 2014, Finak et al, 2015, Guo et al, 2015, Nabavi et al, 2015, Delmans and Hemberg, 2016, Korthauer et al, 2016, Costa-Silva et al, 2017, Qiu et al, 2017, Miao et al, 2018, Van den Berge et al, 2018, Wang and Nabavi, 2018, Wang et al, 2019, and so developing approaches and software to compare these different software packages would be of great utility to the scientific community. Generating data from the two-group model is a special case of (1) and (2), where…”