Cell type prediction using ImmGen, Mouse and Human Cell Atlas

To do this you need to download the following databse files from our iCellR data link (more data to come soon).

# download the .rda files from here: https://genome.med.nyu.edu/results/external/iCellR/data/ 
# Load the .rda files as below

load("Immgen.GSE109125.205.rda")
load("Immgen.GSE122108.412.rda")
load("Immgen.GSE122597.83.rda")
load("Immgen.GSE124829.190.rda")
load("Immgen.microarray.GSE15907.653.rda")
load("Immgen.microarray.GSE37448.189.rda")
load("immgen.rna.rda")
load("immgen.uli.rna.rda")
load("mouse.cell.atlas.rda") 

Key	Source	Samples	Description	Cell Types
GSE109125	ImmGen	205	83 populations representing all lineages and several differentiation cascades prepared from unchallenged mice and after LPS, anti-CD3, viral infection cell activation.	B Cells, Stromal Cells, Dendritic Cells, Granulocytes, Innate Lymphocytes, Stem Cells, Macrophages, ab T Cells, gd T Cells
GSE122108	ImmGen	412	130 populations comprising progenitors, residents, and stimulated (C.alb, LPS, injury, APAP+ starved overnight and pIC) mononuclear phagocytes for OpenSource MNP Project.	Macrophages, Kupffer Cell/Macrophages, Dendritic Cells, Microglia, Monocytes.
GSE122597	ImmGen	83	Five highly purified immunocyte populations profiled to unusual depth as multiple replicates (8 to 16). Suitable for exploration of genes expressed at very low levels.	NK Cells, Follicular B, Naive CD4+ abT, gdT cells and peritoneal macrophages.
GSE124829	ImmGen	190	11 diverse immunocyte populations from male and female mice of varying ages stimulated with different dose of IFN to understand the immune system's sexual differences.	B Cells, Dendritic Cells, Neutrophils, Macrophages, Natural Killer T Cells, ab T Cells, gd T Cells, Microglia, Regulatory T Cells.
GSE15907	ImmGen	653	178 populations compromiing of gene-expression microarray datasets ("version1" labeling) from primary cells from multiple immune lineages are isolated ex-vivo, primarily from 6weeks B6 male mice.	gd T Cells, ab T Cells, Dendritic Cells, Macrophages, Stem Cells, B Cells, Stromal Cells, Neutrophils, Fibroblast, NK Cells, NK T Cells, Monocytes, CD4 Naive T Cell.
GSE37448	ImmGen	189	80 populations compromising of gene-expression microarray datasets ("version2" labeling) from primary cells from multiple immune lineages are isolated ex-vivo, primarily from 6weeks B6 male mice. Complements the V1 compendium with additional cells. Unfortunately, the version change in the labeling process, while more efficient, introduced some biases such that the two sections of the data can be compared grossly, but not at fine resolution (we tried...).	gd T Cells, ab T Cells, Dendritic Cells, Macrophages, Stem Cells, B Cells, Stromal Cells, Neutrophils, Fibroblast, NK Cells, NK T Cells, Monocytes, CD4 Naive T Cell.
rna	ImmGen	23	Full depth directional RNA sequencing was performed on the core ImmGen populations to generate reference datasets for the tissues from 5 week-old C57BL/6J (Jackson Laboratory) males and females, double-sorted by flow cytometry, per ImmGen cell preparation SOP.	B, CD4T, CD8T, DC, MQ,NK, NKT, Treg
uli.rna	ImmGen	157
mca	Mouse Cell Atlas	43 tissues	Constructed as a basic scheme for the Mouse Cell Atlas using Microwell-seq.	Uterus, TrophoblastStemCells, Thymus, Testis, Stomach, Spleen, SmallIntestine, Prostate, Placenta, PeripheralBlood, Pancreas, Ovary, NeontalBrain, NeonatalSkin, NeonatalRib, NeonatalMuscle, NeonatalHeart, NeonatalCalvaria, Muscle, Mouse3T3, MesenchymalStemCellsPrimary, MesenchymalStemCells, MammaryGland.Virgin, MammaryGland.Pregnancy, MammaryGland.Lactation, MammaryGland.Involution, Male.fetal.Gonad, Lung, Liver, Kidney, FetalStomach, FetalLung, FetalLiver, FetalKidney, FetalIntestine, FetalBrain, Female.fetal.Gonad, EmbryonicStemCells, EmbryonicMesenchyme, Brain, BoneMarrowcKit, BoneMarrow, Bladder

Choose a cluster and take for example top 10 genes for that cluster and then choose one of the databases that is best for you from the above list and predict your cell type. Note that if you have B cells for example and the database of your choice dose not have B cells, it would predict the closest looking cells to B cells. So it's important to use the right database for the right type of data.

# Choose top 40 genes for cluster 8 for example
MyGenes <- top.markers(marker.genes, topde = 40, min.base.mean = 0.2, cluster = 8)

####### predict
# plot 
cell.type.pred(immgen.data = "rna", gene = MyGenes, plot.type = "point.plot")

cell.type.pred(immgen.data = "uli.rna", gene = MyGenes, plot.type = "point.plot", top.cell.types = 50)
 
cell.type.pred(immgen.data = "rna", gene = MyGenes, plot.type = "heatmap")
 
cell.type.pred(immgen.data = "uli.rna", gene = MyGenes, plot.type = "heatmap")

# As you can see cluster 8 is most likely to be B-cells. 

# more examples
cell.type.pred(immgen.data = "GSE109125", gene = MyGenes, plot.type = "point.plot", top.cell.types = 50)

cell.type.pred(immgen.data = "GSE37448", gene = MyGenes, plot.type = "heatmap", top.cell.types = 50)

# for tissue type prediction use this:
cell.type.pred(immgen.data = "mca", gene = MyGenes, plot.type = "point.plot")

# And finally check the genes in the cells and find the common ones to predict
heatmap.gg.plot(my.obj, gene = MyGenes, interactive = F, cluster.by = "clusters") 

You can automate this for all the clusters as below. Add as many plot as you wish.

Clusters = sort(unique([email protected]$clusters))


for(i in Clusters){
	Cluster = i
	MyGenes <- top.markers(marker.genes, topde = 10, min.base.mean = 0.2, cluster = Cluster)
# first plot
Name <- paste("ImmGen_Cluster_",Cluster,"_pointPlot_RNA.pdf",sep="")
pdf(Name, width = 10, height = 10)
print(cell.type.pred(immgen.data = "rna", gene = MyGenes, plot.type = "point.plot"))
dev.off()
# second plot
Name <- paste("ImmGen_Cluster_",Cluster,"_check.pdf",sep="")
pdf(Name, width = 10, height = 10)
print(heatmap.gg.plot(my.obj, gene = MyGenes, interactive = F, cluster.by = "clusters"))
dev.off()
}

• Pathway analysis

# Pathway  
# pathways.kegg(my.obj, clust.num = 7) 
# this function is being improved and soon will be available