High heterogeneity undermines generalization of differential expression results in RNA-Seq analysis

Abstract

Background RNA sequencing (RNA-Seq) has been widely applied in oncology for monitoring transcriptome changes. However, the emerging problem that high variation of gene expression levels caused by tumor heterogeneity may affect the reproducibility of differential expression (DE) results has rarely been studied. Here, we investigated the reproducibility of DE results for any given number of biological replicates between 3 and 24 and explored why a great many differentially expressed genes (DEGs) were not reproducible. Results Our findings demonstrate that poor reproducibility of DE results exists not only for small sample sizes, but also for relatively large sample sizes. Quite a few of the DEGs detected are specific to the samples in use, rather than genuinely differentially expressed under different conditions. Poor reproducibility of DE results is mainly caused by high variation of gene expression levels for the same gene in different samples. Even though biological variation may account for much of the high variation of gene expression levels, the effect of outlier count data also needs to be treated seriously, as outlier data severely interfere with DE analysis. Conclusions High heterogeneity exists not only in tumor tissue samples of each cancer type studied, but also in normal samples. High heterogeneity leads to poor reproducibility of DEGs, undermining generalization of differential expression results. Therefore, it is necessary to use large sample sizes (at least 10 if possible) in RNA-Seq experimental designs to reduce the impact of biological variability and DE results should be interpreted cautiously unless soundly validated.

Effect of high variation in transcript expression on identifying differentially expressed genes in RNA‐seq analysis

“…However, as shown in Figure 1, when (FDR(0.05), LFC(1)) is applied, the number of DEGs detected positively correlates with n if n is below 10. Similar phenomena have been reported in previous studies using RNA‐seq read count data from mouse strains (Soneson & Delorenzi, 2013), yeast (Schurch et al., 2016), tomato plants (Lamarre et al., 2018), and human tissues (Cui et al., 2021), and cell lines (Liu et al., 2014). The results of the parallel analysis using HeLa cells (Supporting information Figures S7‐S9) are also consistent with those obtained from human tumor and normal tissue samples.…”

“…Poor reproducibility of DEGs has been shown in several studies using different datasets from tomato plants (Lamarre et al., 2018), yeast (Schurch et al., 2016), mouse strains (Soneson & Delorenzi, 2013), and human tissues(Cui et al., 2021), and cell lines (Liu et al., 2014); however, the findings of all these studies were based on DE analyses using the common threshold value for FC (│log2FC│≥ 1) and FDR (FDR < 0.05). The question was whether the reproducibility of DE results could be improved by increasing significance stringency.…”

Deep Radiotranscriptomics of Non-Small Cell Lung Carcinoma for Assessing Molecular and Histology Subtypes with a Data-Driven Analysis

“…Some challenges for single source RNA-seq data that significantly affect the generalization of the analysis are the varying data acquisition protocols [ 71 ], intratumor heterogeneity [ 72 ] and local mutation burden [ 73 ]. These are prominent aspects of non-small cell lung carcinomas.…”

Meta-Analysis of Microdissected Breast Tumors Reveals Genes Regulated in the Stroma but Hidden in Bulk Analysis

“…High-throughput analyses of gene expression hold great promise for the identification of biomarkers of clinical status, with the potential of predicting outcome, response to therapy, or informing researchers about molecular mechanisms underpinning disease onset and progression and identifying therapeutic targets [1]. Nevertheless, lists of candidate genes obtained through transcriptome-based studies have proven difficult to reproduce [2][3][4][5][6], raising a note of caution regarding conclusions driven by single sets of experiments. Sample collection and processing methods, protocols, and platforms may impact on the resulting gene signatures, making them non-overlapping between studies [7].…”