CODEWITHSUNDEEP
rrowsbioconductorgenome
user2120870
user2120870

Reputation: 869

Estimating distance difference between rows (genetic markers)

I would like to calculate the distance between the markers (Name) in a given chromosome (Chr). The objects dist1.alldown (distance downstream) and dist1.allup (distance upstream) have exactly what I want. However, the below script is computationally inefficient (my real data could contain a milion of markers and this loop is time consuming).

  df <-  'Name  Chr Position
  GGaluGA001820 chr1    34388
  Gga_rs16686671 chr1    67781
  GGaluGA001841 chr1    80477
  Gga_rs15995401 chr1   111556
  Gga_rs15995393 chr1   112481
  GGaluGA001890 chr1   149690
  GGaluGA001902 chr1   176450
  Gga_rs14688751 chr1   185573
  GGaluGA001921 chr1   202425
  GGaluGA001945 chr1   235155'
df <- read.table(text=df, header=T)
probes <- df   
probes.split <- split(probes, probes$Chr)

####### Loop to infer distance upstream #####
{dist1.all <- NULL
 for(k in 1:length(probes.split)){
   probescx <- probes.split[[k]]
   probescx <- probescx[order(probescx$Position, decreasing=F),]
   for(i in 1:nrow(probescx)){
     v <- vector()
     v[k] <- k^2; print(paste(k,i)) 
     rowx <- probescx[i,]
     rowxm1 <- probescx[i-1,]
     if(nrow(rowxm1) > 0){
       lab <- rowx[1,1:2]
       dist1 <- rowx[1,3] - rowxm1[1,3]
       dist1 <- as.data.frame(dist1)
       dist1 <- cbind(lab, dist1)
       dist1.all <- rbind(dist1.all, dist1)
     }
   }
 }
}
### Save a different object
dist1.allup <- dist1.all
##background of up object
dist1.allupback <- dist1.allup

### Loop to infer distance downstream
{dist1.all <- NULL
 for(k in 1:length(probes.split)){
   probescx <- probes.split[[k]]
   probescx <- probescx[order(probescx$Position, decreasing=F),]
   for(i in 1:nrow(probescx)){
     v <- vector()
     v[k] <- k^2; print(paste(k,i)) 
     rowx <- probescx[i,]
     rowxm1 <- probescx[i+1,]
     if(nrow(rowxm1) > 0){
       lab <- rowx[1,1:2]
       dist1 <- rowx[1,3] - rowxm1[1,3]
       dist1 <- as.data.frame(dist1)
       dist1 <- cbind(lab, dist1)
       dist1.all <- rbind(dist1.all, dist1)
     }
   }
 }
}
### Save a different object
dist1.alldown <- dist1.all
##background of down object
dist1.alldownback <- dist1.alldown
## Turn distance in positive integers
dist1.alldown$dist1 <- dist1.alldown$dist1 * -1

Some ideas or known tools to obtain an efficiently approach? Thank you!

Upvotes: 2

Views: 58

Answers (1)

Chris
Chris

Reputation: 6372

Lets simplify your data a bit. You have:

> df
             Name  Chr Position
1   GGaluGA001820 chr1    34388
2  Gga_rs16686671 chr1    67781
3   GGaluGA001841 chr1    80477
4  Gga_rs15995401 chr1   111556
5  Gga_rs15995393 chr1   112481
6   GGaluGA001890 chr1   149690
7   GGaluGA001902 chr1   176450
8  Gga_rs14688751 chr1   185573
9   GGaluGA001921 chr1   202425
10  GGaluGA001945 chr1   235155

Based on 

> dist1.allup
             Name  Chr dist1
2  Gga_rs16686671 chr1 33393
3   GGaluGA001841 chr1 12696
4  Gga_rs15995401 chr1 31079
5  Gga_rs15995393 chr1   925
6   GGaluGA001890 chr1 37209
7   GGaluGA001902 chr1 26760
8  Gga_rs14688751 chr1  9123
9   GGaluGA001921 chr1 16852
10  GGaluGA001945 chr1 32730

You're looking for the rowwise distance between markers (I.e. GGalu -> Gga_rs, Gga_rs -> GGalu).

The most straightforward way to do this (and very computationally quick) would be with data.table.

First, set to a data table

library(data.table)
setDT(df)

Then, order your data so that you have consecutive markers (your data may already be like this, but good to make sure:

df <- df[order(Chr,Position)]

Then, create offsetting data for Chr, Name and Position:

df[, ChrN := Chr[.I + 1]]
df[, NameN := Name[.I + 1]]
df[, PosN := Position[.I + 1]]

We only want to compare on the same chromosome:

df <- df[Chr == ChrN]

And now we can calculate the distances

df[, list(NameFrom = Name, NameTo = NameN, Chr, dist = PosN - Position)]

As this is vectorized, and uses in memory operations, it should be much quicker than the looping approach above.

For all.down, use: 

df <- df[-order(Chr,Position)]

and

df[, list(NameFrom = Name, NameTo = NameN, Chr, dist = PosN - Position)]

becomes

df[, list(NameFrom = Name, NameTo = NameN, Chr, dist = Position - PosN)]

Upvotes: 1

Related Questions