Define ROSMAP Variables

This script first performs preprocessing of ROSMAP FPKM and covariate data so that they have same order, then defines variables in the covariate data.

Dependencies

Load requisite packages.

library(data.table)

Read FPKM Data

Read FPKM data, remove the version from Ensembl ID, and remove duplicated samples.

# read FPKM data
fpkm_file = "../Data/ROSMAP_RNAseq_FPKM_gene.tsv"
rosmap_fpkm = fread(fpkm_file, stringsAsFactors = FALSE, header = T, sep = "\t")
all(rosmap_fpkm$tracking_id == rosmap_fpkm$gene_id)
gene_id = rosmap_fpkm$gene_id
rosmap_fpkm = rosmap_fpkm[, c(-1,-2), with = F]
rosmap_fpkm = as.data.frame(rosmap_fpkm)

# remove version from Ensembl ID
rownames(rosmap_fpkm) = gsub("\\.\\d*$", "", gene_id)

# remove duplicated samples
duplicate_sample_ids = c("492_120515_0","492_120515_6")
rosmap_fpkm = rosmap_fpkm[, -which(colnames(rosmap_fpkm) %in% duplicate_sample_ids)]

Read Covariate Data

Read ROSMAP ID mapping file.

# read ROSMAP ID map
key_map_file = "../Data/ROSMAP_IDkey.csv"
key_map = fread(key_map_file)
key_map = unique(key_map[, .(projid, mrna_id)])

Read and pre-process covariate data.

# read and process covariate data
cov_file = "../Data/ROSMAP_clinical.csv"
cov = fread(cov_file)

# rename covariate data and merge with ROSMAP IDs
setnames(cov, c("projid", "cts_mmse30_lv", "braaksc", "ceradsc", "cogdx", "apoe_genotype"), 
         c("projid", "MMSE", "Braak", "CERAD", "ClinicalDiagnosis", "RawAPOE"))
cov[key_map, on = .(projid), mrna_id := mrna_id]

# no version number
cov[, mrna_id_nov := gsub("_\\d$", "", mrna_id)]

Reorder Data

# extract clinical data in the same order as in the data
colnames(rosmap_fpkm) = gsub("_\\d$", "", colnames(rosmap_fpkm)) 

# remove samples without clinical data
colnames(rosmap_fpkm)[!colnames(rosmap_fpkm) %in% c(cov$mrna_id_nov)]
rosmap_fpkm = rosmap_fpkm[, colnames(rosmap_fpkm) %in% cov$mrna_id_nov]
all(colnames(rosmap_fpkm) %in% cov$mrna_id_nov)
cov = cov[data.table(mrna_id_nov = colnames(rosmap_fpkm)), on = .(mrna_id_nov)]

# check that cov and rosmap_fpkm have same order
all(cov$mrna_id_nov == colnames(rosmap_fpkm))
colnames(rosmap_fpkm) = cov$projid

Define Clinical Variables

Braak

Group levels into three categories:

• 1 = Normal/I/II
• 2 = III/IV
• 3 = V/VI

# refactor
cov[, B := as.factor(cov$Braak)]
levels(cov$B) = c(1,1,1,2,2,3,3)
cov[, B := as.character(B)]
cov[, B := factor(B)]

# double-check
table(cov$B, cov$Braak)

CERAD

Old ROSMAP levels:

• 1 = Definite
• 2 = Probable
• 3 = Possible
• 4 = Normal

Define new levels:

• 1 = Normal
• 2 = Possible
• 3 = Probable
• 4 = Definite

# refactor
cov[, C := as.factor(CERAD)]
levels(cov$C) = c(3,2,1,0)
cov[, C := as.character(C)]
cov[, C := factor(C)]

# double-check
table(cov$C, cov$CERAD)

APOE

Group levels into three categories:

• E2 = E2/E2 and E2/E3
• E3 = E3/E3
• E4 = E2/E4, E3/E4, and E4/E4

# refactor
cov[, APOE := as.factor(RawAPOE)]
levels(cov$APOE) = c("E2", "E2", "E4", "E3", "E4", "E4")
cov[, E := as.character(APOE)]
cov[, E := factor(E, levels = c("E3", "E2", "E4"))]

# double-check
table(cov$E, cov$RawAPOE)

E4 Count

Group levels into three categories:

• 0 = E2/E2, E3/E3, and E2/E3
• 1 = E2/E4 and E3/E4
• 2 = E4/E4

# refactor
cov[, E4num := as.factor(RawAPOE)]
levels(cov$E4num) = c(0, 0, 1, 0, 1, 2)
cov[, E4num := as.numeric(as.character(E4num))]

# double-check
table(cov$E4num, cov$RawAPOE)

E2 Count

Group levels into three categories:

• 0 = E4/E4, E3/E3, and E3/E4
• 1 = E2/E4 and E2/E3
• 2 = E2/E2

# refactor
cov[, E2num := as.factor(RawAPOE)]
levels(cov$E2num) = c(2, 1, 1, 0, 0, 0)
cov[, E2num := as.numeric(as.character(E2num))]

# double-check
table(cov$E2num, cov$RawAPOE)

Age at Death

Clean age at death variable.

# clean age at death
cov[, age_at_death := gsub("90\\+", "91", age_death)]
cov[, age_at_death := as.numeric(age_at_death)]

Save Results

# save results
rosmap = list(fpkm = rosmap_fpkm,
              cov = cov)
save(rosmap, file = "../Data/ROSMAP-24.Rdata")