Question

I am not sure the title is clear enough.
I have a dataframe (see below) which contains values across 5 columns. What I would like to do is to "split" this dataframe into three classes where the rows can be assigned into a "High", "Medium", "Low" state.

What I mean is :

High: the values are "high" in at least 3 columns

Medium: the values are "medium" in a least 3 columns

Low: the values are "Low"(or NA) in a least 3 columns

I guess it involve two things, defining the value cutoff for the 3 groups, then assinging rows into High, Medium and Low category… but thats a guess

The data file is available here

tmp = read.table("tmp2.txt", header=TRUE)
head(tmp)
           Geneid     Hsap      Mmul      Mmus      Rnor     Cfam
1 ENSG00000197711 365823.5 243429.20 44337.267 156874.50 128015.0
2 ENSG00000198712 198613.0        NA 47767.767 200176.50 210559.8
3 ENSG00000198899 189421.5        NA        NA 283425.50 367112.8
4 ENSG00000198804 182559.5        NA 87301.900 277861.00 324438.0
5 ENSG00000198840 142424.5        NA  8400.457  45844.80 115027.9
6 ENSG00000171564 119147.9  93564.66  6675.290  45938.85  45140.2

Any advices strongly appreciated, as I don’t have the slightest idea on how to tackle this !

Thanks,

---

This is the answer below :

I have now replaced the file by a more realistic one (more rows)

tbl <- read.csv("http://db.tt/L2ehGh8", header=FALSE)
colnames(tbl) <- c("Geneid","Hsap","Mmul","Mmus","Rnor","Cfam")

Using cut() :
I have lots of 0s, and the values are quiet stretched, so by using log, or here asinh, you get rid of this.

tbl.data <- apply(asinh(tbl.data),2,
                  function(x) as.numeric(as.factor(cut(x,4)))  )
head(tbl.data)
     Hsap Mmul Mmus Rnor Cfam
[1,]    2    2    1    1    2
[2,]    2    2    2    2    2
[3,]    1    1    1    1    1
[4,]    1    1    1    1    1
[5,]    2    3    2    2    3
[6,]    2    2    2    2    2

Another way is to use Quantiles, which as been shown to me.

quantile(tbl.data[,1],0.25)
quantile(tbl.data[,1],0.5)
quantile(tbl.data[,1],0.75)

tbl.data2 <- apply(tbl.data,2,
                   function(x) as.numeric(as.factor(cut(x,c(-1,
                       quantile(x, 0.25)+0.0001,
                       quantile(x,0.5),
                       quantile(x,0.75), max(x))))))
head(tbl.data2)
     Hsap Mmul Mmus Rnor Cfam
[1,]    3    3    3    2    3
[2,]    2    3    4    3    3
[3,]    2    1    1    1    2
[4,]    1    2    1    1    1
[5,]    4    4    4    4    4
[6,]    3    4    4    3    4

Answer 1

Assuming you want NAs to be handled by not counting them rather than tossing the whole row:

tbl <- read.table("http://db.tt/Eb6qM4h",header=TRUE)
tbl.data <- subset(tbl,select=-Geneid)
tbl.data <- apply(tbl.data,2,function(x) as.numeric(as.factor(cut(x,3)))  )


countLevels <- function(tbl.data,lvl) {
  apply(tbl.data,1,function(x) sum( x[!is.na(x)] == lvl ) )
}

tbl.final <- tbl.new <- subset(tbl,select=Geneid)
for(lvl in seq(3) ) {
  tbl.new[,paste('Level',lvl)] <- (countLevels(tbl.data,lvl) > 3) * lvl
}

tbl.final$Levels <- rowSums(subset(tbl.new,select=-Geneid))

Which returns the data.frame as follows:

> head(tbl.final,20)
            Geneid Levels
1  ENSG00000197711      0
2  ENSG00000198712      0
3  ENSG00000198899      0
4  ENSG00000198804      0
5  ENSG00000198840      0
6  ENSG00000171564      1
7  ENSG00000171557      1
8  ENSG00000198727      1
9  ENSG00000163631      0
10 ENSG00000198888      1
11 ENSG00000198695      1
12 ENSG00000198763      1
13 ENSG00000198786      1
14 ENSG00000158874      0
15 ENSG00000138207      1
16 ENSG00000109072      1
17 ENSG00000130203      3
18 ENSG00000106927      1
19 ENSG00000110169      1
20 ENSG00000104760      1