UniBlood Reference Creation

Overview

This readme documents the creation process of the UniBlood sorted cell DNAme references used in the CellsPickMe R package. The package deconvolute cellular composition of cord, pediatric, and adult blood samples based on their DNA methylation (DNAme) profile, as measured on Illumina methylation arrays (compatible with HM450k, EPICv1, and EPICv2)

library(ExperimentHub)
library(minfi)
library(dplyr)
library(ggplot2)

UniBlood references

The UniBlood references, including UniBlood7, UniBlood13, and UniBlood19, were generated based on combining preexisting cord blood and adult blood-based references. By combining sorted cells’ DNAme from these two distinct developmental time points, we aimed to capture cell-type-specific signals that account for changes in the hematopoietic landscape. The composition of each UniBlood references is shown in Table 1.

Table 1. Cellular composition and dataset included for the three reference datasets we compiled in this study, UniBlood7, UniBlood13, and UniBlood19.

	UniBlood7	UniBlood13	UniBlood19
FlowSorted.Blood.450k (Reinius)	🗸
FlowSorted.Blood.EPIC (IDOL)	🗸	only monocytes, neutrophils, and NK cells	only monocytes, neutrophils, and NK cells
FlowSorted.Blood.Extended.EPIC (Extended)		🗸	🗸
FlowSorted.CordBlood.450k (Cord)	🗸	6 randomly selected nRBC samples	6 randomly selected samples for every cell type
Adult and cord cell types grouped?	Yes	No	No
Cell types included	Seven cell types: B cells, CD4+ T cells, CD8+ T cells, NK cells, monocytes, granulocytes, and nRBCs	Thirteen cell types: naïve and memory B cells, naïve and memory CD4+ T cells, naïve and memory CD8+ T cells, Tregs, NK cells, monocytes, neutrophils, basophils, eosinophils, and nRBCs	Nineteen cell types: neonatal, naïve, and memory B cells, neonatal, naïve, and memory CD4+ T cells, neonatal, naïve, and memory CD8+ T cells, adult Tregs, neonatal and adult NK cells, neonatal and adult monocytes, neonatal granulocytes, adult neutrophils, basophils, and eosinophils, and nRBCs

Load existing reference datasets

hub <- ExperimentHub()

### Extended reference
query(hub, "FlowSorted.BloodExtended.EPIC")  
FlowSorted.BloodExtended.EPIC <- hub[["EH5425"]]
n
FlowSorted.BloodExtended.EPIC <- FlowSorted.BloodExtended.EPIC[, FlowSorted.BloodExtended.EPIC$Sample_Group != "FlowSorted.Blood.EPIC"]
meta_1 <- pData(FlowSorted.BloodExtended.EPIC)
meta_1 <- meta_1[, c("Sex", "Age", "CellType", "Ethnicity_wide", "purity")]
meta_1$Ref <- "Extended"

### IDOL reference
FlowSorted.Blood.EPIC <- FlowSorted.Blood.EPIC::libraryDataGet('FlowSorted.Blood.EPIC')
meta_2 <- pData(FlowSorted.Blood.EPIC)
meta_2 <- meta_2[, c("Sex", "Age", "CellType", "Ethnicity_wide", "purity")]
meta_2$Ref <- "IDOL"

### Cord reference
FlowSorted.CordBlood.450k <- FlowSorted.CordBloodCombined.450k::libraryDataGet('FlowSorted.CordBloodCombined.450k')
meta_3 <- pData(FlowSorted.CordBlood.450k)
meta_3 <- meta_3[, c("Sex", "Age", "CellType")] %>% cbind(., Ethnicity_wide = NA, purity = NA)
meta_3$Ref <- "Cord"

### Reinius
library(FlowSorted.Blood.450k)
data("FlowSorted.Blood.450k")
meta_4 <- pData(FlowSorted.Blood.450k)
meta_4 <- meta_4[, c("Sex", "CellType")] %>% cbind(., Age = NA, Ethnicity_wide = NA, purity = NA)
meta_4$Ref <- "Reinius"

UniBlood7

UniBlood7 includes the Reinius, Cord, and IDOL references. With only 7 cell types, the reference treat neonatal and adult cells of the same identity (e.g. CD4T cells) as one cell type.

Ref_meta <- do.call(rbind, list(meta_2, meta_3, meta_4))

# Combine reference datasets
UniBlood7 <- combineArrays(FlowSorted.Blood.450k, 
                           FlowSorted.Blood.EPIC, 
                           outType = "IlluminaHumanMethylation450k")
UniBlood7 <- combineArrays(UniBlood7, 
                           FlowSorted.CordBlood.450k, 
                           outType = "IlluminaHumanMethylation450k")

UniBlood7_meta <- Ref_meta[sampleNames(UniBlood7), ]
pData(UniBlood7) <- UniBlood7_meta

save(UniBlood7, compression_level = 9, file = "data/UniBlood7.rda")

UniBlood13

UniBlood13 is mostly built on the Extended reference, with additional samples from the IDOL reference and nRBCs from the Cord reference. With 13 cell types, the reference assumes that neonatal and adult cells share similar DNAme profiles except for the nRBCs, which only exist at the neonatal time point.

Ref_meta <- do.call(rbind, list(meta_1, meta_2, meta_3))

# Select samples to exclude from cord blood samples based on odd clustering from PCA
samp_IDOL <- rownames(meta_2)[meta_2$CellType %in% c("Mono", "Neu", "NK")]

samp_Extended <- rownames(meta_1)[meta_1$CellType != "MIX"]

FlowSorted.CordBlood.nRBC <- FlowSorted.CordBlood.450k[, meta_3$CellType == "nRBC"]

# Combine reference datasets
UniBlood13 <- combineArrays(FlowSorted.BloodExtended.EPIC[, samp_Extended], 
                            FlowSorted.Blood.EPIC[, samp_IDOL], 
                            outType = "IlluminaHumanMethylationEPIC")
UniBlood13 <- combineArrays(UniBlood13, 
                            FlowSorted.CordBlood.nRBC, 
                            outType = "IlluminaHumanMethylation450k")

UniBlood13_meta <- Ref_meta[sampleNames(UniBlood13), ]
pData(UniBlood13) <- UniBlood13_meta

save(UniBlood13, compression_level = 9, file = "data/UniBlood13.rda")

UniBlood19

Finally, UniBlood19 is the reference with the most cell types. It combines reference datasets from Extended, Cord, and IDOL. The reference treated neonatal and adult cells as distinct identities, with the cord blood cells labelled with the “_cord” suffix.

Ref_meta <- do.call(rbind, list(meta_1, meta_2, meta_3))

# Select samples to exclude from cord blood samples based on odd clustering from PCA
out <- c("X3999492113_R01C01", 
         "X3999492113_R06C02", 
         "X3999997142_R02C02", 
         "X3999984059_R01C02", 
         "X3999492043_R06C01", 
         "X3999984058_R06C01", 
         "X201868590206_R06C01", 
         "3999492043_R06C01", 
         "3999984058_R06C01", 
         "200705360101_R04C01", 
         "Gran_4", 
         "Gran_5", 
         "200705360089_R06C01", 
         "200705360058_R07C01", 
         "200693480131_R02C01", 
         "3999984059_R04C02", 
         "200705360013_R06C01", 
         "200705360068_R02C01", 
         "3999492113_R01C01", 
         "CD8+_6", 
         "200705360060_R06C01", 
         "3999997142_R03C01",
         "200693480078_R02C01")

Ref_meta <- Ref_meta[!rownames(Ref_meta) %in% out, ]

set.seed(1234)
samp_Cord <- sapply(c("CD4T", "NK", "nRBC", "Bcell", "Mono", "Gran", "CD8T", "WBC"), function(x){
    sample(rownames(Ref_meta)[Ref_meta$CellType == x & Ref_meta$Ref == "Cord"], size = 10)
}) %>% as.vector()

samp_IDOL <- rownames(meta_2)[meta_2$CellType %in% c("Mono", "Neu", "NK")]

samp_Extended <- rownames(meta_1)#[meta_1$CellType != "MIX"]

# Rename cord blood cells
Ref_meta$CellType[Ref_meta$Ref == "Cord"] <- paste0(Ref_meta$CellType[Ref_meta$Ref == "Cord"], "_cord") %>%
    gsub("nRBC_cord", "nRBC", .)

# Combine reference datasets
UniBlood19 <- combineArrays(FlowSorted.BloodExtended.EPIC[, samp_Extended], 
                            FlowSorted.Blood.EPIC[, samp_IDOL], 
                            outType = "IlluminaHumanMethylationEPIC")
UniBlood19 <- combineArrays(UniBlood19, 
                            FlowSorted.CordBlood.450k[, samp_Cord], 
                            outType = "IlluminaHumanMethylation450k")

UniBlood19_meta <- Ref_meta[sampleNames(UniBlood19), ]
pData(UniBlood19) <- UniBlood19_meta

save(UniBlood19, compression_level = 9, file = "data/UniBlood19.rda")