TROM: A Testing-Based Method for Finding Transcriptomic Similarity of Biological Samples

Abstract

Comparative transcriptomics has gained increasing popularity in genomic research thanks to the development of high-throughput technologies including microarray and next-generation RNA sequencing that have generated numerous transcriptomic data. An important question is to understand the conservation and divergence of biological processes in different species. We propose a testing-based method TROM (Transcriptome Overlap Measure) for comparing transcriptomes within or between different species, and provide a different perspective, in contrast to traditional correlation analyses, about capturing transcriptomic similarity. Specifically, the TROM method focuses on identifying associated genes that capture molecular characteristics of biological samples, and subsequently comparing the biological samples by testing the overlap of their associated genes. We use simulation and real data studies to demonstrate that TROM is more powerful in identifying similar transcriptomes and more robust to stochastic gene expression noise than Pearson and Spearman correlations. We apply TROM to compare the developmental stages of six Drosophila species, C. elegans, S. purpuratus, D. rerio and mouse liver, and find interesting correspondence patterns that imply conserved gene expression programs in the development of these species. The TROM method is available as an R package on CRAN (https://cran.r-project.org/package=TROM) with manuals and source codes available at http://www.stat.ucla.edu/~jingyi.li/software-and-data/trom.html.

Large-scale mapping of mammalian transcriptomes identifies conserved genes associated with different cell states

“…However, in contrast to TROM, these correlation analyses failed to find clear or informative correspondence patterns among the mammalian cell states ( Supplementary Data ). This was expected and consistent with our previous findings ( 28 , 44 ), because correlation values depend heavily on the accuracy of gene expression estimates. On the other hand, TROM finds correspondence between cell states based on their associated genes and is more robust to noise and biases in gene expression estimates.…”

“…To capture the transcriptome characteristics of different cell states, we define the ‘associated genes’ of a cell state as the protein-coding and long non-coding genes whose expression is relatively high in that cell state and relatively low in some other cell states. To identify the associated genes of different cell states and subsequently use them to compare the cell states from different species, we adapted a statistical method TROM, which were recently developed for comparing developmental stages of D. melanogaster and C. elegans ( 28 , 44 ). We first identified the associated genes of cell states by the following criterion: in a given cell state, its associated genes must have (i) FPKM ( 35 ) above a positive constant c ( c = 1 for protein-coding genes and c = 0 for lncRNAs, because lncRNAs are generally more lowly expressed than protein-coding genes); (ii) normalized Z -score in the top 5% among its normalized Z -scores in all cell states (see Material and Methods for details, Supplementary Data ).…”

“…We then used the TROM algorithm to map two mammalian cell states by their transcriptome similarity ( 28 , 44 ). In the within-species cell state mapping, for every two cell states, we tested the significance of the number of their common associated genes using an ‘overlap test’, whose null hypothesis is that the two cell states’ associated genes are independent samples from the gene population of the species.…”

“…We collected and processed 307 publicly available polyadenylated RNA-seq data sets for more than 40 tissues and cell types (including ESCs, iPSCs, in vivo tissues and cultured cell lines) from four mammalian species (human, chimpanzee, bonobo and mouse). In order to quantitatively characterize the lineage relationships of those tissues and cell types, we used a statistical method TROM (Transcriptome Overlap Measure) ( 28 , 44 ) to find their correspondence via a comprehensive transcriptome mapping. The results also provide a catalog of protein-coding and long non-coding associated genes, which well capture transcriptome characteristics of various cell identities in different species.…”

Learning a genome-wide score of human-mouse conservation at the functional genomics level

“…The challenge in taking advantage of such data is that many experiments do not have an obvious corresponding experiment, and even when one is assumed there could in practice be confounding differences. Previous work partly addressed some of these issues 24,[37][38][39][40][41][42][43] , but still limited their work to either one data type or dimension of the data at a time and thus only utilized a small fraction of the available data to find evidence of conservation.…”