How to find the number of cells in a column that match another data frame's range?

Question

I have a data.frame1 like:

Input_SNP_CHR   Input_SNP_BP     Set_1_CHR   Set_1_BP     Set_2_CHR   Set_2_BP     Set_3_CHR   Set_3_BP
    chr4         184648954        chr18      63760782       chr7      135798891      chr7        91206783  
    chr13        45801432         chr14      52254555       chr1      223293324      chr4        184648954
    chr18        71883393         chr22      50428069       chr7      138698825      chr18       63760782

I have another data.frame2 like:

CHR     BP1             BP2             Score   Value
chr1    29123222        29454711        -5.7648 599
chr13   45799118        45986770        -4.8403 473
chr5    46327104        46490961        -5.3036 536
chr6    50780759        51008404        -4.4165 415
chr18   63634657        63864734        -4.8096 469
chr1    77825305        78062178        -5.4671 559

I would like to find out how many rows in each pair (a pair is Input_SNP_CHR and Input_SNP_BP together) in data.frame1 BOTH match a CHR and fall between BP1 and BP2 of data.frame2. For example, in my first pair (the Input_SNP pair) I have one match. This is the second row, where both the CHR (chr13) and BP (45801432) of Input_SNP match a row by CHR (chr13) and BP range (between 45799118 and 45986770) in data.frame2. For my second pair (the Set_1 pair) I also have 1 match (chr18) and BP 63760782 match with the 5th row of data.frame2 by chr18 and the BP range.

My desired output would be:

Input_SNP     Set_1     Set_2     Set_3
1             1         0         1

How would I go about doing this in R?

Answer 1

Here's another possible solution using data.table . First we melt the data to a long format, add a Set column according to the first df column names, and then running foverlaps combined with table in order to check frequencies

library(data.table) # v 1.9.6+
Ldf <- melt(setDT(df), measure = patterns("CHR", "BP")) # Create a column for BP and CHR
Names <- unique(sub("(.*_.*)_.*", "\\1", names(df))) # Creates a sets names indx
setnames(Ldf[, variable := factor(Names[variable])], c("Set", "CHR", "BP1")) # Rename   
Ldf[, BP2 := BP1] # Creating right boundary for foverlaps
setkeyv(Ldf, names(Ldf)[-1]) # Keying for foverlaps
table(foverlaps(setDT(df2), Ldf, nomatch = 0L)$Set) # Running fovelaps and checking freqs 
# Input_SNP     Set_1     Set_2     Set_3 
#         1         1         0         1 

Answer 2

I think the data in your first data.frame should be formatted like this

#       CHR      type        BP
# 1.1  chr4 Input_SNP 184648954
# 1.2 chr13 Input_SNP  45801432
# 1.3 chr18 Input_SNP  71883393
# 2.1 chr18     Set_1  63760782
# 2.2 chr14     Set_1  52254555
# 2.3 chr22     Set_1  50428069
# 3.1  chr7     Set_2 135798891
# 3.2  chr1     Set_2 223293324
# 3.3  chr7     Set_2 138698825
# 4.1  chr7     Set_3  91206783
# 4.2  chr4     Set_3 184648954
# 4.3 chr18     Set_3  63760782

(Row names are not important though.)

Ideally you would generate the data like that but if you already have it in the format like you provided, you can transform it via (assuming the name of your first data.frame is df )

type_list=lapply(strsplit(colnames(df),"_"),
                 function(x) c(paste0(x[1],"_",x[2])))

df_new=do.call("rbind",
               lapply(split(1:ncol(df),sort(rep(1:(ncol(df)/2),times=2))),
                      function(idxs) {
                        data.frame(CHR=df[,idxs[1]],
                                   type=type_list[[idxs[1]]],
                                   BP=df[,idxs[2]])}))

Then it's just two lines of base R to accomplish your task (assuming the second data.frame is df2 )

df_new_2=within(merge(df_new,df2,by="CHR"),
                cnt<-BP>=BP1&BP<=BP2)

sapply(split(df_new_2,df_new_2$type),function(x) sum(x$cnt))
#Input_SNP     Set_1     Set_2     Set_3 
#        1         1         0         1 

(I only get one hit for Set_3 as only chr18 matches.)