We introduce Signac, a versatile R package to facilitate the analysis workflow for single-cell data. It helps to find marker genes faster and more accurate, search for cells with similar expression profiles, integrate multiple datasets in the BioTuring Browser database (know more about BioTuring Browser), etc. For users with a limited computational resource, we provide the helper functions to exercise all analyses for the large-scale datasets from disk. Because of its speed and flexibility, it can be adapted to any existing R analysis pipeline to help explore single-cell data more efficient.
This package can also be used as the reference for the computational methods used in BioTuring Browser software.
Please visit this blog to read about Venice, the fast and accurate method for finding marker genes, which is incorporated into Signac. Venice function can be accessed via Signac::VeniceMarker
Below is the benchmark when finding 4 types of DE genes in a simulated dataset (using scDD R package). Total 15 methods included Venice are tested. The dataset has 2000 DE genes that are even divided into 4 groups (DE, DP, DB, DM), and 18000 non-DE genes (EP and EE).
devtools::install_github("bioturing/signac")> class(pbmc.mat)
[1] "dgCMatrix"
attr(,"package")
[1] "Matrix"
> head(clusters)
[1] "Memory CD4 T" "B" "Memory CD4 T" "CD14+ Mono" "NK" "Memory CD4 T"
### Find markers for "NK" cluster
> markers <- Signac::VeniceMarker(pbmc.mat, cluster = (clusters != "NK") + 1)
> head(markers) # Type 1: Up-regulated; -1: Down-regulated
Gene.Name Similarity Log.P.value P.adjusted.value Type
1 NKG7 0.1366077 -552.6917 1.277099e-236 0.9870968
2 GNLY 0.1575343 -520.4139 6.656220e-223 1.0000000
3 GZMB 0.1730481 -502.0630 4.138385e-215 1.0000000
4 PRF1 0.2285456 -418.0301 9.703964e-179 1.0000000
5 CST7 0.2894125 -346.8274 6.500973e-148 1.0000000
6 CTSW 0.2838700 -345.5459 1.951351e-147 0.9869281