Stable gene expression for normalisation and single-sample scoring

Abstract

BackgroundTranscriptomic signatures are useful in defining the molecular phenotypes of cells, tissues, and patient samples. Their most successful and widespread clinical application is the stratification of breast cancer patients into molecular (PAM50) subtypes. In most cases, gene expression signatures are developed using transcriptome-wide measurements, thus methods that match signatures to samples typically require a similar degree of measurements. The cost and relatively large amounts of fresh starting material required for whole-transcriptome sequencing has limited clinical applications, and accordingly thousands of existing gene signatures are unexplored in a clinical context.ResultsGenes in a molecular signature can provide information about molecular phenotypes and their underlying transcriptional programs from tissue samples, however determining the transcriptional state of these genes typically requires the measurement of all genes across multiple samples to allow for comparison. An efficient assay and scoring method should quantify the relative abundance of signature genes with a minimal number of additional measurements. We identified genes with stable expression across a range of abundances, and with a preserved relative ordering across large numbers (thousands) of samples, allowing signature scoring, and supporting general data normalisation for transcriptomic data. Based on singscore, we have developed a new method, stingscore, which quantifies and summarises relative expression levels of signature genes from individual samples through the inclusion of these “stably-expressed genes”.ConclusionWe show that our proposed list of stable genes has better stability across cancer and normal tissue data than previously proposed stable or housekeeping genes. Additionally, we show that signature scores computed from whole-transcriptome data are comparable to those calculated using only values for signature genes and our panel of stable genes. This new approach to gene expression signature analysis may facilitate the development of panel-type tests for gene expression signatures, thus supporting clinical translation of the powerful insights gained from cancer transcriptomic studies.

Measuring and Modelling the Epithelial- Mesenchymal Hybrid State in Cancer: Clinical Implications

“…Epithelial and mesenchymal signatures were obtained from Tan et al [2014] for both cell lines and tumour samples separately. Scores were calculated using singscore and ranked using 5 stable genes [Foroutan et al, 2018;Bhuva et al, 2020 middle value of 1 indicates the E/M hybrid state [George et al, 2017]. Maximum variability in terms of EMT is associated with increased E/M hybrid signatures [Chakraborty et al, 2020], consistent with the higher plasticity ascribed with hybrid E/M phenotype(s).…”

“…When applied to differential analysis workflows, EMT gene expression signatures can evaluate the differential enrichment for an EMT programme between 2 groups of samples. Alternatively, single-sample gene set analysis methods such as ssGSEA [Barbie et al, 2009] or singscore [Foroutan et al, 2018;Bhuva et al, 2020] can be used to identify individual samples that are concordant with gene expression signatures and estimate the EMP phenotype of a single sample. In recent work, a refinement of the singscore method has been developed to remove the requirement for wholetranscriptome scale measurement in order to apply EMT signatures [Bhuva et al, 2020].…”

“…Alternatively, single-sample gene set analysis methods such as ssGSEA [Barbie et al, 2009] or singscore [Foroutan et al, 2018;Bhuva et al, 2020] can be used to identify individual samples that are concordant with gene expression signatures and estimate the EMP phenotype of a single sample. In recent work, a refinement of the singscore method has been developed to remove the requirement for wholetranscriptome scale measurement in order to apply EMT signatures [Bhuva et al, 2020]. Bhuva et al [2020] demonstrate that by adding as few as 5 stably expressed anchor genes to the genes included in an EMT signature, resolution of epithelial, mesenchymal and hybrid phenotypes can be achieved in the context of a low-or medium throughput measurement methods such as qRTPCR or Nanostring.…”

“…In recent work, a refinement of the singscore method has been developed to remove the requirement for wholetranscriptome scale measurement in order to apply EMT signatures [Bhuva et al, 2020]. Bhuva et al [2020] demonstrate that by adding as few as 5 stably expressed anchor genes to the genes included in an EMT signature, resolution of epithelial, mesenchymal and hybrid phenotypes can be achieved in the context of a low-or medium throughput measurement methods such as qRTPCR or Nanostring. This refinement enables the translation of EMT signatures commonly used with RNA sequencing in an experimental research context into clinical application or application to archival FFPE tissues where whole-transcriptome RNA sequencing is frequently not possible.…”

Removing unwanted variation from large-scale cancer RNA-sequencing data

“…We used these CMS and MSI to define homogenous biological populations for the purpose of creating the PRPS (Supplementary Figure 32). We used a slightly complicated approach to select a suitable set of negative control genes for the COAD study as follows: a) carry out an ANOVA on the FPKM.UQ normalized gene expression values with CMS subtypes as the factor; b) calculate Spearman correlations between FPKM.UQ normalized gene expression values and tumor purity; c) calculate Spearman correlations between FPKM.UQ normalized gene expression values and the average expression level of a set of housekeeping genes [29]; and then d) selecting genes (262 genes) that have lowest F statistics (F statistics < 20 ) in a), the lowest correlation (ρ < 0.3) in b), and the highest correlations (ρ > 0.9) in c).…”