Modelling and measuring single cell RNA expression levels find considerable transcriptional differences among phenotypically identical cells

Abstract

Background: Phenotypically identical cells demonstrate predictable, robust behaviours. However, there is uncertainty as to whether phenotypically identical cells are equally similar at the underlying transcriptional level or if cellular systems are inherently noisy. To answer this question, it is essential to distinguish between technical noise and true variation in transcript levels. A critical issue is the contribution of sampling effects, introduced by the requirement to globally amplify the single cell mRNA population, to observed measurements of relative transcript abundance.

microRNA target prediction programs predict many false positives

“…Gene expression typically fluctuates by twofold when comparing two individuals in the human population (Cheung et al 2003), two pools of individuals from a genetically homogenized Drosophila population (Laurie-Ahlberg et al 1982), or two phenotypically identical mouse neural stem cells (Subkhankulova et al 2008). Natural polymorphism in cis-regulatory sequences in the human population causes a large variability in gene expression, which is often greater than twofold (Rockman and Wray 2002).…”

High-Throughput Single-Cell Analysis for Wound Healing Applications

“…31,32 However, the former is a time-intensive technique that is not scalable to large numbers of genes, and the latter, although capable of simultaneously evaluating the expression of thousands of genes, requires considerable global mRNA amplification which introduces significant noise in the resulting measurements. 33 Furthermore, considering the stochastic nature of transcription and translation, 34,35 which can themselves introduce variability in excess of twofold expression change, it is often necessary to study individual cells in larger numbers in order to account for the underlying temporal variation. 36 Higher-throughput methods such as immunohistochemistry and fluorescent activated cell sorting (FACS) are viable alternatives that can simultaneously evaluate thousands and millions of cells, respectively.…”

Current and Future Applications of Transcriptomics for Discovery in CNS Disease and Injury

“…LCM was subsequently used to isolate fluorojade-labelled neurons in a rat TBI model, where the expression of the neuroprotective genes, glutathione peroxidase 1, heme oxygenase 1, and brain-derived neurotrophic factor, were found to be significantly decreased in injured CA3 neurons compared with adjacent uninjured neurons [22] , demonstrating that transcription signatures of adjacent neurons vary greatly depending upon survival. This variation does not appear to be due to technical noise, suggesting that cellular systems are inherently noisy at the single-cell level [23] . Using murine neural stem cells (NSCs), Subkhankulova et al [23] determined that the majority of transcripts (44%) are present at less than 25 copies/cell and proposed that this low abundance along with translational regulation may account for the observed heterogeneity.…”

“…This variation does not appear to be due to technical noise, suggesting that cellular systems are inherently noisy at the single-cell level [23] . Using murine neural stem cells (NSCs), Subkhankulova et al [23] determined that the majority of transcripts (44%) are present at less than 25 copies/cell and proposed that this low abundance along with translational regulation may account for the observed heterogeneity.…”