Quick Start Tutorial

Viewmaster for automated cell annotation

1) load greenleaf MPAL dataset into monocle3

mpal<-readRDS(file ="healthy_bone_marrow_greenleaf.rds")
plot_cells(mpal, color_cells_by = "BioClassification", label_cell_groups = F)+scale_color_manual(values = sfc(n = 26))

2) load hca bone marrow seurat object, convert to monocle using viewmaster's helpful function

hca<-readRDS(file = "palantir_bm_r1.rds")
hca<-seurat_to_monocle3(hca, seu_rd = "tsne")
plot_cells(hca , color_cells_by = "cell_annotation", label_cell_groups = F)+scale_color_manual(values = sfc(10, scramble = F))

3) Detect common variant genes between two cell data sets

vg<-common_variant_genes(mpal, hca, top_n = 5000)

4) Use softmax regression for viewmaster. We will use the Greenleaf lab's MPAL healthy bone marrow cell annotations and transfer them to Human Cell Atlas of bone marrow

view<-viewmaster(ref_cds  = mpal, query_cds = hca, ref_celldata_col = "BioClassification", selected_genes = vg, verbose=T, FUNC = "softmax_regression", tf_idf = F)

plot_cells(view, color_cells_by = "smr_celltype", label_cell_groups = F)+scale_color_manual(values = sfc(11))

scRNA Logistic regression matrix

code wrapped from Sam Behjati

library(viewmaster)
library(doMC)
library(glmnet)

hca_seu<-monocle3_to_seurat(hca)
mpal_seu<-monocle3_to_seurat(mpal, mon_rd = "UMAP")

log_mat<-log_reg_matrix(trainDat = mpal_seu, testDat = hca_seu, trainClass = "BioClassification", testClass = "cell_annotation", downsample = 500)

testDat<- loadTrainingData(hca_seu)

clusterPreds = apply(log_mat,2,function(e) sapply(split(e,testDat$mDat[["cell_annotation"]]),mean))

#Convet to percentages
clusterPreds = (1+exp(-clusterPreds))**-1

pheatmap(clusterPreds, cluster_cols=T,cluster_rows = T ,col = ArchR::paletteContinuous("coolwarm", n=10, reverse = F), border_color = "black")