基于两列值的计算过滤数据表的优雅方法是什么？ [R]

Question

I have a data table, let's call it lung :

> lung
                     variant_id             transcript_id is_NL counts nrows
     1: chr10_129450960_T_C_b38 chr10_129467297_129536240     0  33029   458
     2: chr10_129450960_T_C_b38 chr10_129467297_129536240     1   3477    54
     3: chr10_129450960_T_C_b38 chr10_129467297_129536240     2    130     3
     4: chr10_129450960_T_C_b38 chr10_129536378_129563778     0     51   458
     5: chr10_129450960_T_C_b38 chr10_129536378_129563778     1      8    54
    ---
500148:   chr9_34699703_G_C_b38    chr9_34649082_34649409     1   4214    57
500149:   chr9_34699703_G_C_b38    chr9_34649082_34649409     2    171     2
500150:   chr9_34699703_G_C_b38    chr9_34649565_34650368     0  48713   456
500151:   chr9_34699703_G_C_b38    chr9_34649565_34650368     1   4932    57
500152:   chr9_34699703_G_C_b38    chr9_34649565_34650368     2    208     2

I would like to filter it such that when is_NL == 0 , the only rows preserved are those which counts/nrows < 50 ( 50 being an arbitrary number), and when is_NL is 1 or 2 , the only rows that are preserved are those which counts/nrows > 50 .

So far, I've only been able to come up with this:

> lung[which(lung[is_NL == 0][,counts]/lung[is_NL == 0][,nrows] < 50),]
                     variant_id             transcript_id is_NL counts nrows
     1: chr10_129450960_T_C_b38 chr10_129467297_129536240     1   3477    54
     2: chr10_129450960_T_C_b38 chr10_129536378_129563778     0     51   458
     3: chr10_129450960_T_C_b38 chr10_129536378_129563778     1      8    54
     4: chr10_129450960_T_C_b38 chr10_129536378_129707894     0  37918   458
     5: chr10_129450960_T_C_b38 chr10_129701913_129707894     0    188   458
    ---
147877:  chr17_45825156_G_A_b38   chr17_46148240_46152903     2     17    20
147878:  chr17_45825156_G_A_b38   chr17_46152967_46156773     0      3   336
147879:  chr17_45825156_G_A_b38   chr17_46152967_46169530     0      5   336
147880:  chr17_45825156_G_A_b38   chr17_46152967_46169530     1    137   159
147881:  chr17_45825156_G_A_b38   chr17_46156896_46170854     0     18   336
> lung[which(lung[is_NL > 0]$counts/lung[is_NL > 0]$nrows > 50),]
                    variant_id             transcript_id is_NL counts nrows
    1: chr10_129450960_T_C_b38 chr10_129467297_129536240     0  33029   458
    2: chr10_129450960_T_C_b38 chr10_129536378_129563778     1      8    54
    3: chr10_129450960_T_C_b38 chr10_129701913_129707894     1     24    54
    4: chr10_129450960_T_C_b38 chr10_129701913_129707894     2      2     3
    5: chr10_129450960_T_C_b38 chr10_129708044_129715519     2      0     3
   ---
50195:  chr17_46025930_T_C_b38   chr17_46039885_46050532     0  14129   337
50196:  chr17_46025930_T_C_b38   chr17_46050705_46066536     0  14106   337
50197:  chr17_46025930_T_C_b38   chr17_46050705_46066536     1   6658   158
50198:  chr17_46025930_T_C_b38   chr17_46050705_46066536     2    809    20
50199:  chr17_46025930_T_C_b38   chr17_46066733_46067548     0  12842   337

which, as you can tell by looking at the is_NL column, does not work. I could probably subset into two different tables first, apply the comparison filter ( < or > 50 ), and then figure out how to merge them, but I feel like there should be a simpler way to do this that I don't know about.

Answer 1

In base R, you could do something like:

lung[with(lung, (is_NL == 0 & counts/nrows < 50) | 
                (is_NL %in% c(1,2) & counts/nrows > 50)),]
# output
               variant_id             transcript_id is_NL counts nrows
2 chr10_129450960_T_C_b38 chr10_129467297_129536240     1   3477    54
4 chr10_129450960_T_C_b38 chr10_129536378_129563778     0     51   458

where I created lung as the first 5 lines in your example:

lung <- structure(list(variant_id = structure(c(1L, 1L, 1L, 1L, 1L), .Label = "chr10_129450960_T_C_b38", class = "factor"), 
    transcript_id = structure(c(1L, 1L, 1L, 2L, 2L), .Label = c("chr10_129467297_129536240", 
    "chr10_129536378_129563778"), class = "factor"), is_NL = c(0L, 
    1L, 2L, 0L, 1L), counts = c(33029L, 3477L, 130L, 51L, 8L), 
    nrows = c(458L, 54L, 3L, 458L, 54L)), class = "data.frame", row.names = c(NA, 
-5L))

Answer 2

Using data.table

library(data.table)
setDT(lung)[!is_NL & counts/.N < 50|(is_NL %in% c(1, 2) & counts/.N > 50)]

Answer 3

I would create a flag using tidyverse :

lung %>% 
mutate(FLG = if_else(is_NL == 0 & counts/nrows < 50, 1 
             if_else(is_NL in (1,2) &counts/nrows >50, 1,0))) %>% 
filter(FLG == 1)

If you want to use base R then

lung[which((is_NL == 0 & counts/nrows < 50)|(is_NL in (1,2) &counts/nrows >50)),]