如何计算r中哪些行满足条件？

Question

I have a dataset of groups of genes that I filter to select the best gene per group with some conditions:

Data:

 Group Gene      Score     direct_count   secondary_count 
    1   AQP11    0.5566507       4               5
    1   CLNS1A   0.2811747       0               2
    1   RSF1     0.5469924       3               6
    2   CFDP1    0.4186066       1               2
    2   CHST6    0.4295135       1               3
    3   ACE      0.634           1               1
    3   NOS2     0.6345          1               1
    4   Gene1    0.7             0               1
    4   Gene2    0.61            1               0
    4   Gene3    0.62            0               1   

Filtering:

dt %>% 
  group_by(Group) %>% 
  filter((max(Score) - Score)<0.05) %>% 
  slice_max(direct_count, n = 1) %>% 
  slice_max(secondary_count, n = 1) %>% 
  ungroup()

I am looking to be able to count how may genes are being filtered at which step in the above code.

So for example my conditions I am applying with this code are:

1. Select the gene with the highest score if the score difference between the top scored gene and all others in the group is >0.05

2. If the score difference between the top gene and any other genes in a group is <0.05 then select the gene with a higher direct_count only selecting between those genes with a <0.05 distance to the top scored gene per group

3. If the direct_count is the same select the gene with the highest secondary_count

4. If all counts are the same select all genes <0.05 distance to each other.

I've been able to count the genes meeting my first condition (>0.05 score) doing:

new_df <- dt %>% 
  group_by(Group) %>% 
  filter((max(Score) - Score)<0.05)

count1 <- new_df[!(duplicated(new_df$Group) | duplicated(new_df$Group, fromLast = TRUE)), ] 

I've been trying to apply similar rules to get counts of how many genes are meeting conditions for higher direct_count or higher secondary_count or matching direct_count and secondary_count , but different code I try gives different numbers so I'm not sure what is the best way.

Input data:

#Input data before filtering with code above:

structure(list(Group = c(1L, 1L, 1L, 2L, 2L, 3L, 3L, 4L, 4L, 4L), Gene = c("AQP11", 
"CLNS1A", "RSF1", "CFDP1", "CHST6", "ACE", "NOS2", "Gene1","Gene2","Gene3"), Score = c(0.5566507, 
0.2811747, 0.5269924, 0.4186066, 0.4295135, 0.634, 0.6345, 0.7, 0.62, 0.61), direct_count = c(4L, 
0L, 3L, 1L, 1L, 1L, 1L, 0L, 1L, 0L), secondary_count = c(5L, 2L, 6L, 2L, 
3L, 1L, 1L, 0L, 0L, 1L)), row.names = c(NA, -10L), class = c("data.table", 
"data.frame"))

#Input data after filtering with code applied above:

structure(list(Group = c(1L, 1L, 1L, 2L, 2L, 3L, 3L, 4L, 4L, 
4L), Gene = c("AQP11", "CLNS1A", "RSF1", "CFDP1", "CHST6", "ACE", 
"NOS2", "Gene1", "Gene2", "Gene3"), Score = c(0.5566507, 0.2811747, 
0.5269924, 0.4186066, 0.4295135, 0.634, 0.6345, 0.7, 0.62, 0.61
), direct_count = c(4L, 0L, 3L, 1L, 1L, 1L, 1L, 0L, 1L, 0L), 
    secondary_count = c(5L, 2L, 6L, 2L, 3L, 1L, 1L, 0L, 0L, 1L
    )), row.names = c(NA, -10L), class = c("data.table", "data.frame"
))

This example data I've made should have 1 gene group selected by >0.05 score, 1 gene group filtered by larger direct_count and 2 gene groups filtered by secondary_count . Ideally I am aiming to count the number of groups and be able to pull them out of the dataset.

Output from example would just be a count like:

Genes filtered by >0.05 score: 1
Genes filtered by direct_count: 1
Genes filtered by secondary_count: 2

Answer 1

Count

Basically, before and after each filter, you calculate the number of rows available and you save it in a new column.

library(dplyr)
dt %>% 
  group_by(Group) %>% 
  mutate(filter0 = n()) %>% 
  filter((max(Score) - Score)<0.05) %>% 
  mutate(filter1 = n()) %>% 
  slice_max(direct_count, n = 1) %>% 
  mutate(filter2 = n()) %>% 
  slice_max(secondary_count, n = 1) %>% 
  mutate(filter3 = n()) %>% 
  ungroup()

#> # A tibble: 5 x 9
#>   Group Gene  Score direct_count secondary_count filter0 filter1 filter2 filter3
#>   <int> <chr> <dbl>        <int>           <int>   <int>   <int>   <int>   <int>
#> 1     1 AQP11 0.557            4               5       3       2       1       1
#> 2     2 CHST6 0.430            1               3       2       2       2       1
#> 3     3 ACE   0.634            1               1       2       2       2       2
#> 4     3 NOS2  0.634            1               1       2       2       2       2
#> 5     4 Gene1 0.7              0               0       3       1       1       1

---

Show filters explicitly

Or you can keep track of the filters in this way. Each column shows if the row was selected or not at each filtering.

library(dplyr)
dt %>% 
  group_by(Group) %>% 
  mutate(filter1 = (max(Score) - Score)<0.05) %>% 
  mutate(filter2 = rank(-replace(direct_count, !filter1, -Inf), ties.method = "min") == 1) %>% 
  mutate(filter3 = rank(-replace(secondary_count, !filter2, -Inf), ties.method = "min") == 1) %>% 
  ungroup()

#> # A tibble: 10 x 8
#>    Group Gene   Score direct_count secondary_count filter1 filter2 filter3
#>    <int> <chr>  <dbl>        <int>           <int> <lgl>   <lgl>   <lgl>  
#>  1     1 AQP11  0.557            4               5 TRUE    TRUE    TRUE   
#>  2     1 CLNS1A 0.281            0               2 FALSE   FALSE   FALSE  
#>  3     1 RSF1   0.527            3               6 TRUE    FALSE   FALSE  
#>  4     2 CFDP1  0.419            1               2 TRUE    TRUE    FALSE  
#>  5     2 CHST6  0.430            1               3 TRUE    TRUE    TRUE   
#>  6     3 ACE    0.634            1               1 TRUE    TRUE    TRUE   
#>  7     3 NOS2   0.634            1               1 TRUE    TRUE    TRUE   
#>  8     4 Gene1  0.7              0               0 TRUE    TRUE    TRUE   
#>  9     4 Gene2  0.62             1               0 FALSE   FALSE   FALSE  
#> 10     4 Gene3  0.61             0               1 FALSE   FALSE   FALSE  

If you filter by the last column ( filter3 ) you actually get the same output of the dplyr pipe you shared in your question.

library(dplyr)
dt %>% 
  group_by(Group) %>% 
  mutate(filter1 = (max(Score) - Score)<0.05) %>% 
  mutate(filter2 = rank(-replace(direct_count, !filter1, -Inf), ties.method = "min") == 1) %>% 
  mutate(filter3 = rank(-replace(secondary_count, !filter2, -Inf), ties.method = "min") == 1) %>% 
  ungroup() %>%
  filter(filter3)
#> # A tibble: 5 x 8
#>   Group Gene  Score direct_count secondary_count filter1 filter2 filter3
#>   <int> <chr> <dbl>        <int>           <int> <lgl>   <lgl>   <lgl>  
#> 1     1 AQP11 0.557            4               5 TRUE    TRUE    TRUE   
#> 2     2 CHST6 0.430            1               3 TRUE    TRUE    TRUE   
#> 3     3 ACE   0.634            1               1 TRUE    TRUE    TRUE   
#> 4     3 NOS2  0.634            1               1 TRUE    TRUE    TRUE   
#> 5     4 Gene1 0.7              0               0 TRUE    TRUE    TRUE  

---

Visual Aid

If it is easier for you to visually see how filters evolve, remember you can split your data with group_split , like this:

library(dplyr)
dt %>% 
  group_by(Group) %>% 
  mutate(filter1 = (max(Score) - Score)<0.05) %>% 
  mutate(filter2 = rank(-replace(direct_count, !filter1, -Inf), ties.method = "min") == 1) %>% 
  mutate(filter3 = rank(-replace(secondary_count, !filter2, -Inf), ties.method = "min") == 1) %>% 
  group_split()

OUTPUT:

<list_of<
  tbl_df<
    Group          : integer
    Gene           : character
    Score          : double
    direct_count   : integer
    secondary_count: integer
    filter1        : logical
    filter2        : logical
    filter3        : logical
  >
>[4]>
[[1]]
# A tibble: 3 x 8
  Group Gene   Score direct_count secondary_count filter1 filter2 filter3
  <int> <chr>  <dbl>        <int>           <int> <lgl>   <lgl>   <lgl>  
1     1 AQP11  0.557            4               5 TRUE    TRUE    TRUE   
2     1 CLNS1A 0.281            0               2 FALSE   FALSE   FALSE  
3     1 RSF1   0.527            3               6 TRUE    FALSE   FALSE  

[[2]]
# A tibble: 2 x 8
  Group Gene  Score direct_count secondary_count filter1 filter2 filter3
  <int> <chr> <dbl>        <int>           <int> <lgl>   <lgl>   <lgl>  
1     2 CFDP1 0.419            1               2 TRUE    TRUE    FALSE  
2     2 CHST6 0.430            1               3 TRUE    TRUE    TRUE   

[[3]]
# A tibble: 2 x 8
  Group Gene  Score direct_count secondary_count filter1 filter2 filter3
  <int> <chr> <dbl>        <int>           <int> <lgl>   <lgl>   <lgl>  
1     3 ACE   0.634            1               1 TRUE    TRUE    TRUE   
2     3 NOS2  0.634            1               1 TRUE    TRUE    TRUE   

[[4]]
# A tibble: 3 x 8
  Group Gene  Score direct_count secondary_count filter1 filter2 filter3
  <int> <chr> <dbl>        <int>           <int> <lgl>   <lgl>   <lgl>  
1     4 Gene1  0.7             0               0 TRUE    TRUE    TRUE   
2     4 Gene2  0.62            1               0 FALSE   FALSE   FALSE  
3     4 Gene3  0.61            0               1 FALSE   FALSE   FALSE  

But if you are more of a "visual" guy, you can plot the evolution of the filters for each group.

Use geom_tile to create a heatmap of the selected rows.

The plot has to be read from left to right. The red tiles are the ones discarded by the filter.

library(ggplot2)
library(tidyr)
library(dplyr)

dt %>% 
  group_by(Group) %>% 
  mutate(filter1 = (max(Score) - Score)<0.05) %>% 
  mutate(filter2 = rank(-replace(direct_count, !filter1, -Inf), ties.method = "min") == 1) %>% 
  mutate(filter3 = rank(-replace(secondary_count, !filter2, -Inf), ties.method = "min") == 1) %>% 
  
  select(Group, Gene, starts_with("filter")) %>% 
  pivot_longer(starts_with("filter")) %>% 
  
  ggplot() +
  geom_tile(aes(x = name, y = Gene, fill = value), colour = "black") +
  facet_wrap("Group", scales = "free") +
  labs(title = "Gene selected from left to right",
       x = "Filters",
       y = "Genes",
       fill = "Selected")

---

Definitive filter

Following, I'll leave the code to see how many Genes were selected at each step.

Also, as last columns, you can see at which filter you got down to the minimum number of genes selected at the end, in this way you can see how many times each filter was the definitive one.

library(dplyr)
dt1 <- dt %>% 
  group_by(Group) %>% 
  mutate(filter0 = n()) %>% 
  mutate(filter1 = (max(Score) - Score)<0.05) %>% 
  mutate(filter2 = rank(-replace(direct_count, !filter1, -Inf), ties.method = "min") == 1) %>% 
  mutate(filter3 = rank(-replace(secondary_count, !filter2, -Inf), ties.method = "min") == 1) %>% 

  # sum the number of genes selected for each filter
  group_by(Group) %>% 
  summarise(across(starts_with("filter"), sum)) %>% 
  
  # show the number of the decisive filter!
  rowwise() %>% 
  mutate(definitive = which.min(c_across(starts_with("filter")))-1) %>%
  ungroup()

dt1
#> # A tibble: 4 x 6
#>   Group filter0 filter1 filter2 filter3 definitive
#>   <int>   <int>   <int>   <int>   <int>      <dbl>
#> 1     1       9       2       1       1          2
#> 2     2       4       2       2       1          3
#> 3     3       4       2       2       2          1
#> 4     4       9       1       1       1          1

count(dt1, definitive)
#> # A tibble: 3 x 2
#>   definitive     n
#>        <dbl> <int>
#> 1          1     2
#> 2          2     1
#> 3          3     1

ggplot(dt1) + geom_bar(aes(definitive))