# Prepare expression sets from Tusi dataset:

# Steps: 
# 1. Import data 
# 2. Re-format (traspose and remove meta)
# 2.1 Convert alias to symbols
# 2.2 Convert mouse to human symbols
# 3. Create expression set with PhenoData added with info

library(data.table)
library(nichenetr)
library(dplyr)

# for inputs see Tusi et al manuscript. 

infiles <- list(P1 = "GSM2985844_P1.raw_umifm_counts.csv", 
                P1CD71 = "GSM2985845_P1-CD71hi.raw_umifm_counts.csv",
                P2 = "GSM2985846_P2.raw_umifm_counts.csv",
                P5 = "GSM2985849_P5.raw_umifm_counts.csv") 

convmts <- list()

for(i in c("P1","P1CD71","P2","P5")){
  
  obj <- read.csv(file = infiles[[i]], header = T, stringsAsFactors = F, quote = "", as.is = T)
  obj$CellID <- paste(obj$Sample, obj$Library, gsub(obj$Barcode_Seq, pattern = "-", replacement = ""), sep = "_")
  row.names(obj) <- obj$CellID
  obj$CellID <- NULL
  obj_t <- t(obj)
  mymat <- obj_t[10:nrow(obj_t),]
  ids <- rownames(mymat)
  human_symbols = ids %>% nichenetr::convert_mouse_to_human_symbols()
  names(human_symbols) <- NULL
  mymat <- mymat[!is.na(human_symbols),]
  rownames(mymat) <- human_symbols[!is.na(human_symbols)]
  #mymat <- noquote(mymat)
  convmts[[i]] <- as.data.frame(mymat)
}

saveRDS(object = convmts, file = "convmts.rds")

full <- merge.data.frame(convmts$P1, convmts$P1CD71, by = 0, sort = F)
rownames(full) <- full$Row.names
full$Row.names <- NULL
full <- merge.data.frame(full, convmts$P2, by = 0, sort = F)
rownames(full) <- full$Row.names
full$Row.names <- NULL
full <- merge.data.frame(full, convmts$P5, by = 0, sort = F)
rownames(full) <- full$Row.names
full$Row.names <- NULL

full_num <- full
full_num <- apply(full, 2, as.numeric)
rownames(full_num) <- rownames(full)

# create Expression set

library(Biobase)

minimalSet <- ExpressionSet(assayData=full_num)
saveRDS(minimalSet,file = "minimalset.rds")