Note: this feature is deprecated in VISION 3.0+.
To compute Signature consistency scores, VISION needs a cell-cell similarity metric in order to build the neighborhood graph. By default, VISION runs PCA and uses euclidean distances in the reduced space for this purpose. However, if the cells’ variation is better modeled by a trajectory (either linear or tree-like), then distances along the trajectory structure can be used instead.
Many tools exist to perform trajectory inference in single cells. The Dynverse project has developed wrappers for 59 recently-published methods and VISION is able to accept the wrapped output of these methods as an input trajectory model.
Within the Dynverse trajectory object, VISION depends on two components:
You should have a working installation of VISION, dyno, and tidyverse.
devtools::install_github("YosefLab/VISION")
devtools::install_github("dynverse/dyno")
install.packages('tidyverse')In this vignette, we’ll be analyzing a set of ~5,000 cells during haematopoiesis (Tusi et al, Nature 2018).
We begin by loading in the data and requisite libraries:
counts = as.matrix(read.table("data/hemato_counts.csv.gz", sep=',', header=T, row.names=1))We’ll now follow the dyno vignette in the github README. You’ll need to make sure you have Docker installed. Although we use Slingshot here, there is a wide variety of methods that can be used, as listed on the Dynmethods page.
Before running the model, we’ll need to perform some gene filtering. We first only select the genes used in the original Tusi et al study, then for further efficiency, we further filter using VISION’s Fano filtering.
k.genes = read.table("data/bBM.filtered_gene_list.paper.txt.gz", sep=',')[,1]
filt.counts = counts[k.genes, ]
f.genes = VISION:::filterGenesFano(filt.counts)
filt.counts = filt.counts[f.genes, ]Now, resuming the workflow recommended on the dyno vignette,
umis <- colSums(counts)
scale.factor <- median(umis)
expr <- t(t(counts) / umis) * scale.factor
filt.expr <- log(1 + expr[f.genes,])
dataset <- wrap_expression(
counts = t(filt.counts),
expression = t(filt.expr)
)
model <- infer_trajectory(dataset, "projected_slingshot", verbose=T)
model <- model %>% add_dimred(dyndimred::dimred_mds, expression_source = dataset$expression)
plot_dimred(
model,
expression_source = dataset$expression
)Now, with this model saved we can pass it to VISION via the
latentTrajectory parameter. The model will automatically be
analyzed with the same signatures passed to the VISION object, and with
this model, the Trajectory tab will be activated in the
VISION output report.
meta <- read.table("data/hemato_covariates.txt.gz", sep='\t', header=T, row.names=1)
meta <- meta[colnames(expr), -1]
vis <- Vision(expr, signatures = c("data/h.all.v5.2.symbols.gmt"),
meta = meta, latentTrajectory = model)
vis <- addProjection(vis, "MDS", model$dimred[,c(1,2)])
vis <- analyze(vis)## R version 3.6.2 (2019-12-12)
## Platform: x86_64-apple-darwin15.6.0 (64-bit)
## Running under: macOS Catalina 10.15.7
##
## Matrix products: default
## BLAS: /Library/Frameworks/R.framework/Versions/3.6/Resources/lib/libRblas.0.dylib
## LAPACK: /Library/Frameworks/R.framework/Versions/3.6/Resources/lib/libRlapack.dylib
##
## locale:
## [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
##
## attached base packages:
## [1] stats graphics grDevices utils datasets methods base
##
## other attached packages:
## [1] BiocStyle_2.20.2
##
## loaded via a namespace (and not attached):
## [1] knitr_1.33 magrittr_1.5 R6_2.4.1
## [4] ragg_1.2.1 rlang_1.0.1 fastmap_1.1.0
## [7] stringr_1.4.0 tools_3.6.2 xfun_0.25
## [10] cli_3.1.1 jquerylib_0.1.4 systemfonts_1.0.4
## [13] htmltools_0.5.2 yaml_2.2.1 digest_0.6.23
## [16] rprojroot_2.0.2 pkgdown_2.0.2 crayon_1.3.4
## [19] bookdown_0.23 textshaping_0.3.6 BiocManager_1.30.16
## [22] purrr_0.3.3 sass_0.4.0 fs_1.5.0
## [25] memoise_2.0.0 cachem_1.0.6 evaluate_0.14
## [28] rmarkdown_2.10 stringi_1.4.5 compiler_3.6.2
## [31] bslib_0.3.1 desc_1.3.0 jsonlite_1.6