Inferring Cellular Localization from Single Cell RNA-seq

Thanks to RNA-seq, we have ways to estimate expression abundances decently (at least at gene level :-)).  One of the other missing dimensions in realm of  expression is where is that transcript expressed from? what is the physical location of the transcript?. Till recently, our ability to answer the positional information about the expressing transcript relied mainly on not-so-high-throughput Flourescent In Situ Hybridization (FISH).

Just a year ago George Church’s team came up with FISSEQ, a technique that combined benefits of sequencing and FISH to offer a high-throughput approach to probe the location of thousands of genes and their expression.

Now, get ready to get a zoomed in view of gene expression inside cell.  Nature Biotechnology has two interesting papers on spatial reconstruction using single cell RNA-seq. These two papers offer new computational methods to infer spatial origin of expression using single-cell RNA-seq and in situ hybridization data.

The paper from Aviv Regev’s group at Broad Institute describes a new statistical framework that uses expression profiles from single cell RNA-seq with in situ hybridization data for a small number of genes to anchor spatial inference. The statistical approach is implemented in R & available as package “Seurat“. ( Thanks to Rahul Satija’s comment, the first author of the paper, corrected the incorrect description of the method.) Seurat is freely available here with tutorials

In case you are wondering about the name, the paper says that Seurat is

named for the artist Georges Seurat to invoke the analogy between the intricate spatial patterning of single cells and a pointillist painting.

The team applied Seurat to a data from 851 dissociated single cells from zebrafish embryos all from a single developmental stage and

confirmed the method’s accuracy with several experimental assays, used it to predict and validate patterns where in situ data were not available, identified and correctly localized rare cell populations—either spatially restricted or intermixed throughout the embryo—and defined additional markers of these populations.

The paper from John Marioni’s group at Wellcome Trust Sanger Institute addresses the same problem and have come with the new computational approach and propose

an integrated approach that combines previously generated in situ hybridization (ISH)-based gene expression atlases with unbiased single-cell transcriptomics .

Our approach is based on comparing complete, specificity-weighted mRNA profiles of a cell with positional gene expression profiles derived from a gene expression atlas. We show that this method allocates cells to precise locations in the brain of the marine annelid Platynereis dumerilii with a success rate of 81%. Our method is applicable to any system that has a reference gene expression database of sufficiently high resolution.

In case you are wondering about P. dumerilli..

Platynereis dumerilii, a marine polychaetous annelid or ragworm.is considered aliving fossil. Platynereis dumerilii lives in the same environment as its ancestors millions of years ago and still has many ancestral features. On of the most interestingancient features this marine worm has retained is an eye like structure called aneyespot.

R Scripts used for analyzing the data is accessible from

Speak Your Mind

*