Filter data frame to remove some repeated factors with different sign-CodePudding

I have one big data frame with different columns like name, position, expression level, q value and so on, and i have many repeats for most of the objects with same name but different expression levels, so I want to filter them if expression levels are in opposite of each other for example up( ) and down (-) regulated values, omit and remove those, but if it finds repeats with different expressions but all up ( ) or all down (-) regulated, keep them. here is an example of my file:

df1<-data.frame(gene.name=c( "DEC1","DEC1","DEC1","ATP","ANXA2","ANXA1","ANXA1","ANXA1"),
                expression.level=c(2.01,0.5,-1.56,3.1,0.67,0.1,1.2,3),
                q.value=c(0.001,0.002,0.0001,0.9,0.00001,0.9,0.0002,0.002))

and output like this:

output<-data.frame(gene.name=c( "ATP","ANXA2","ANXA1","ANXA1","ANXA1"),
                   expression.level=c(3.1,0.67,0.1,1.2,3),
                   q.value=c(0.9,0.00001,0.9,0.0002,0.002))

Thanks in advance for your help.

CodePudding user response：

We can use sign() to check whether they are positive or negative or zero. Then use filter to include those that have the same sign.

library(dplyr)

df1 %>% 
  group_by(gene.name) %>% 
  filter(length(unique(sign(expression.level))) == 1) %>% 
  ungroup()

  gene.name expression.level q.value
1       ATP             3.10   9e-01
2     ANXA2             0.67   1e-05
3     ANXA1             0.10   9e-01
4     ANXA1             1.20   2e-04
5     ANXA1             3.00   2e-03

CodePudding user response：

Using ave you can do this with a one-liner.

df1[with(df1, ave(expression.level, gene.name, FUN=\(x) length(unique(sign(x))))) == 1, ]
#   gene.name expression.level q.value
# 4       ATP             3.10   9e-01
# 5     ANXA2             0.67   1e-05
# 6     ANXA1             0.10   9e-01
# 7     ANXA1             1.20   2e-04
# 8     ANXA1             3.00   2e-03

CodePudding user response：

Using data.table

library(data.table)
setDT(df1)[df1[, .I[uniqueN(sign(expression.level)) == 1], gene.name]$V1]

-output

  gene.name expression.level q.value
      <char>            <num>   <num>
1:       ATP             3.10   9e-01
2:     ANXA2             0.67   1e-05
3:     ANXA1             0.10   9e-01
4:     ANXA1             1.20   2e-04
5:     ANXA1             3.00   2e-03