pop_generator.Rmd

---
title: "POP generator"
output: html_document
params:
  pop_number: "05_05_a"
  cells: 5000
  seed: 23
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

# Parameters
```{r}
params
```

```{r INFO}
# MFI values for NEG and POS cells
neg_mean <- 50
neg_sd <- 1000
pos_mean <- 5000
pos_sd <- 1000

# phenotypes OF INTEREST
B <-      c("CD3" = "low",  "CD4" = "low",  "CD8" = "low",  "CD56" = "low",  "CD19" = "high")
NK <-     c("CD3" = "low",  "CD4" = "low",  "CD8" = "low",  "CD56" = "high", "CD19" = "low")
T4 <-     c("CD3" = "high", "CD4" = "high", "CD8" = "low",  "CD56" = "low",  "CD19" = "low")
T8 <-     c("CD3" = "high", "CD4" = "low",  "CD8" = "high", "CD56" = "low",  "CD19" = "low")
NKT_NN <- c("CD3" = "high", "CD4" = "low",  "CD8" = "low",  "CD56" = "high", "CD19" = "low")
NKT_4 <-  c("CD3" = "high", "CD4" = "high", "CD8" = "low",  "CD56" = "high", "CD19" = "low")
NKT_8 <-  c("CD3" = "high", "CD4" = "low",  "CD8" = "high", "CD56" = "high", "CD19" = "low")
U1 <- c("CD3" = "low",  "CD4" = "low",  "CD8" = "low", "CD56" = "low", "CD19" = "low")
U2 <- c("CD3" = "low",  "CD4" = "high", "CD8" = "low", "CD56" = "low", "CD19" = "low")
U3 <- c("CD3" = "high", "CD4" = "low",  "CD8" = "low", "CD56" = "low", "CD19" = "low")
U4 <- c("CD3" = "high", "CD4" = "low",  "CD8" = "low", "CD56" = "low", "CD19" = "high")

pheno <- ( as.data.frame( rbind(B, NK, T4, T8, NKT_NN, NKT_4, NKT_8, U1, U2, U3, U4) ) )


# PERCENTAGE OF CELLS PER CELL TYPE
# Let it be the same for all cell types:
p <- 100/nrow(pheno)

# or specify it here:
pc <- c("B" = 15, "NK" = 6, "T4" = 44, "T8" = 17.5, "NKT_NN" = 2, "NKT_4" = 0.5, "NKT_8" = 1.5, 
        "U1" = 4.25, "U2" = 0.75, "U3" = 6.5, "U4" = 2)
# save the proportions to a file
saveRDS(pc, paste("props", params$pop_number, Sys.Date(), sep = "_"))

# total number of cells
n <- params$cells

# number of cells per cell type
ns <- c()
for(i in 1:length(pc)){
  ns[i] <- round(n*pc[[i]]/100)
}

# final number of cells
n <- sum(ns)

# cell types
cells <- rownames(pheno)

# markers
markers <- colnames(pheno)

pheno
```

```{r}
# creating the dataframe
set.seed(params$seed)
for(i in (1:length(cells))){      # i takes values [1, number of CELL TYPES]
  # For cell type i, repeat its name as many times as the number of cells 
  c <- rep(cells[i], ns[i]) 
  # One column will be created for every marker with intensity values
  for(j in (1:length(markers))){  
    if(pheno[i,j] == "low"){
      p <- rnorm(ns[[i]], mean = neg_mean, sd = neg_sd)
    }else{
      p <- rnorm(ns[[i]], mean = pos_mean, sd = pos_sd)
    }
     # all the columns (one per marker) are joined:
    # convert the "c" list as data frame to preserve numeric "p" values after cbind-ing
     c <- cbind(as.data.frame(c), p)  
  }
  if(i==1){
    # Values for the first cell type start to fill the dataframe "pop"
    pop <- c
  }else{
    # Values for the other cell types are joined to "pop"
    pop <- rbind(pop, c)
  }
}

# Name the variables:
names(pop) <- c("cells", markers)


#### REARRANGING ###
# Randomly sampling
set.seed(42)
x <- sample(1:n, n, replace = F)

pop <- pop[x,] # Rearranging the dataframe

rownames(pop) <- 1:n # Renaming the rows (in order)


# Save the column with the cell labels to a file
saveRDS(pop$cells,  paste("labels", params$pop_number, Sys.Date(), sep = "_"))

head(pop)
```

```{r}
#par(mfrow = c(ceiling(length(markers)/2), 2))
for(i in 1:length(markers)){
  boxplot(pop[,i+1] ~ pop$cells, xlab = markers[i])
}
```

```{r, eval = F}
plot(pop[,2:(length(markers)+1)])
```

# Conversion to `FlowFrame` class objects
```{r}
library(flowCore)

ff <- new("flowFrame", exprs = as.matrix(pop[,-1])) 
head(ff@exprs)
```

```{r, eval = F}
library(ggcyto)
autoplot(ff, markers[1], markers[5])
```

```{r, eval = F}
autoplot(ff, markers[1])
```

# Export to .fcs
```{r, eval = T}
write.FCS(ff, paste("ff_", params$pop_number, "_", Sys.Date(), ".fcs", sep = "")) 
```