Assessment of dispersion metrics for estimating single-cell transcriptional variability is drawing significant interest across the industry.
Author summary Single-cell RNA sequencing (scRNA-seq) data allows for the study of transcriptional variability. However, the contribution of transcriptional variability to gene expression has not been fully appreciated in part due to a lack of consensus on how to estimate and apply noise metrics for downstream analytical modeling. The study of transcriptional variability provides a new lens through which we can study how transcriptional dynamics impact complex biological phenomena. Here, we simulated single-cell data to test six dispersion metrics for their relative sensitivity to variability in single-cell counts. From our simulations, we found that the variance-to-mean ratio (VMR or Fano factor) appears to be the most suitable metric among those tested for quantifying transcriptional variability as it is scale-invariant and is easily interpretable with respect to changes in data dispersion. We then applied the VMR to analyze changes in transcriptional variability in scRNA-seq datasets from platforms with different capture rates. We find that the Fano factor can identify genes distinct from differentially expressed genes and that variable genes relate to specific functional categories that likely reflect the underlying biology. For most distributions, VMR/Fano factor is a reasonable, robust choice for modeling transcriptional noise. However, for certain niche distributions, other metrics may be better suited. Together, we demonstrate that model choice for measuring transcriptional variability can provide new biological insights into how cells respond and adapt in complex systems.
Experts suggest this could influence future trends and innovation in the sector.
More updates are expected as the story develops.
Source: Original →