CODEWITHSUNDEEP
r
Eduardo Herreros
Eduardo Herreros

Reputation: 27

Select rows where at least the condition is present in 2 columns in R

I have a large dataset with one column of gene names and 4 columns of detection methods (that in this case I called them X1, X2, X3 and X4). I would like to select the rows where the genes are selected by at least 2 detection method. Example of the table is:

Table:
Row   Gene   X1   X2   X3   X4  
 1      A     1    0    0    0
 2      A     0    0    1    0
 3      A     0    1    0    0
 4      B     0    0    1    0
 5      B     0    0    1    0
 6      C     0    0    0    1
 7      D     0    0    1    0
 8      D     0    1    0    0
 9      D     0    1    0    0
 10     E     0    0    1    0
 11     E     0    0    1    0

In summary, I want to select the rows 1,2,3 (Methods X1, X2 and X3 detected gene A) and rows 7,8,9 where methods X2 and X3 detected gene D.

Thanks for your help.

Upvotes: 2

Views: 596

Answers (2)

Jason
Jason

Reputation: 2617

To show which genes were detected by two or more methods, this will work.

Short version:

if zz is your data.frame, then:

yy <- by(zz, zz$Gene, function(dat) {sum(apply(dat[,-c(1,2)], 2, any)) >= 2} )
zz[zz$Gene %in% which(yy),]

Long version

# load the data:
zz <- read.table(header = TRUE, text = "
Row Gene X1 X2 X3 X4 
1 A 1 0 0 0
2 A 0 0 1 0
3 A 0 1 0 0
4 B 0 0 1 0
5 B 0 0 1 0
6 C 0 0 0 1
7 D 0 0 1 0
8 D 0 1 0 0
9 D 0 1 0 0
10 E 0 0 1 0
11 E 0 0 1 0")

# now check, gene by gene, whether at least two columns have at least one 1.
# note that the repeated any() statements can be replaced by a loop or  
# apply(), but for only four columns this works, is easy enough to type, 
# and much easier to understand
yy <- by(zz, zz$Gene, function(dat) {(any(dat$X1) + 
                                      any(dat$X2) + 
                                      any(dat$X3) + 
                                      any(dat$X4) ) >= 2} )

# or, the apply way, in case there are a lot of columns.
# "-c(1,2)" as a column index means "every column except the first two",
# so if the data has 3, 4, or 30 methods, this code stays the same.
yy <- by(zz, zz$Gene, function(dat) {sum(apply(dat[,-c(1,2)], 2, any)) >= 2} )

yy
zz$Gene: A
[1] TRUE
--------------------------------------------------------------------------- 
zz$Gene: B
[1] FALSE
--------------------------------------------------------------------------- 
zz$Gene: C
[1] FALSE
--------------------------------------------------------------------------- 
zz$Gene: D
[1] TRUE
--------------------------------------------------------------------------- 
zz$Gene: E
[1] FALSE

Now to find the matching rows to the Genes that got TRUE results.

Find the names of zz (A, B, C, ... ) that correspond to yy values of TRUE, and index the data.frame based on that...

which(yy)  # equivalent to which(yy == TRUE)

gives

A D 
1 4

and

names(which(yy))

gives

[1] "A" "D"

so...

zz[zz$Gene %in% names(which(yy)),]

gives

  Row Gene X1 X2 X3 X4
1   1    A  1  0  0  0
2   2    A  0  0  1  0
3   3    A  0  1  0  0
7   7    D  0  0  1  0
8   8    D  0  1  0  0
9   9    D  0  1  0  0

Upvotes: 1

GKi
GKi

Reputation: 39737

You can use rowsum and rowSums to find those with more than 1 method and %in% to find the matched rows.

x <- rowSums(rowsum(zz[3:6], zz[,2]) > 0) > 1
zz$Row[zz$Gene %in% names(x[x])]
#[1] 1 2 3 7 8 9

Upvotes: 2

Related Questions