About

Single-cell RNA-seq (scRNA-seq) is widely used to investigate the composition of complex tissues. However, it is often challenging to directly compare the cells identified in two different experiments. scmap allows you to project cells from an scRNA-seq experiment (the Projection) on to the cell-types identified in a different experiment (the Reference).

Links

A copy of the scmap manuscript is available on bioRxiv.
An R package containing scmap source code is available on GitHub.
More information about the existing Reference can be found here.
Please send your feedback/comments/suggestions to Vladimir Kiselev.

Notes

scmap is based on scater format. Please make yourself familiar with it before running scmap.

featureData slots of both the Reference and Projection dataset must have the feature_symbol column which contains Feature (gene/transcript) names from the same organism.

Reference

What would like to do?

Projection

Select an .rds file containing data in scater format

(example-segerstolpe - a human pancreatic dataset with 3514 cells)


What organism is your data from?

Your data is not in scater format! Please upload a data in the correct format.

featureData slot of your scater object does not contain a feature_symbol column. Please add it to your object and re-upload it.

Feature Selection

scmap selects 500 most informative features (genes/transcripts) of the Reference dataset by fitting a linear model to the log(expression) vs log(dropout) distribution of features. The most informative features are shown in red on the plot below. Conceptually, these features have a higher dropout rate for a given mean expression. More information about the selected features is provided in the table below the plot. scmap scores are defined as the residuals of the linear model.

Selected Features





Selecting features...

Cell Projection

scmap projects all cells of the Projection dataset to the reference calculated from the Reference dataset. The Reference is computed by calculating the median expression in each of 500 selected features across all cells in each cell type.

Cell Projection

scmap projects all cells of the Projection dataset to the precomputed References. The references were computed by selecting 500 most informative features of each Reference dataset and calculating the median expression in each of 500 features across all cells in each cell type.





Projecting cells...