根据数据框 R 的元素字符串创建坐标向量

Question

I am trying to create a coordinate vector to later plot with ggplot.

Suppose I have a data frame that looks like:

keys = c("aa", "aa", "ac", "ag", "gg", "at", "ca", "gc", "cc", "cg", "gt", "gg", "tt", "ta", "ga", "tg")
values = c(9.318796e-05, 1.863759e-04, 5.591278e-04, 1.863759e-04, 2.795639e-04, 9.318796e-05, 9.318796e-05, 1.863759e-04, 1.863759e-04, 2.795639e-04, 2.795639e-04, 1.863759e-04, 2.795639e-04, 9.318796e-05, 9.318796e-05, 5.591278e-04)
df = data.frame(keys, values)

Now I want to create a matrix which will give each letter its own space, specifically:

A(-1,1) [upper left],

T(1,-1)[lower right],

G(1,1)[upper right] and

C(-1,-1)[lower left]

For this I have done:

    array_size = sqrt(4^k) #Where k = 2
    graph_coord = c()

    for(i in range(array_size)){
    graph_coord = append(graph_coord, array_size[1])
    } ##Give the graph_coord its size

    maxx = array_size
    maxy = array_size
    posx = 1
    posy = 1

    for(i in df$keys){
    ##This part is for getting each individual letter of each element of keys.
    for(j in i[[1]]){
      ##If the individual letter is a T then the actual position on x should be maxx/2
      if (i == "T"){
        posx = maxx/2
      }else if(i == "C"){
        posy = maxy/2
      }else if(i == "G"){
        posx =maxx/2
        posy =maxy/2
      }

      ###Up until this point I think that the code is doing well, 
      ###I can grab individual letters of each element of key and 
      ###see which one they are and then decide to move them according 
      ###to the initial coordinate maxx and maxy. The next part escapes me:

      maxx = maxx/2
      maxy /=2 ##This /= is customary to python what would be the R equivalent?
     ##Append the graph coordinates with the df$values. 
      graph_coord = append(graph_coord, posy-1, posx-1, prob) ##This part is especially hard for me to grasp and as such I have left the idea, but the code snippet is absolutely incorrect.
    }

This code is still a work in progress. I am trying to recreate what has been done here: Frequency table extracted from Chaos Game Representation

Answer 1

Here's a tidyverse approach:

library(tidyverse)
pos <- data.frame(key = c("a", "g", "t", "c"),
                  x = c(-1, 1, -1, 1),
                  y = c(1, 1, -1, -1))

df %>%
  separate(keys, c("M1", "M2"), sep = 1, remove = FALSE) %>%
  left_join(pos, by = c("M1" = "key")) %>% #this adds an x and y.
  left_join(pos, by = c("M2" = "key")) %>% #the first x from M1 becomes x.x,
                                           #  2nd from M2 becomes x.y
  mutate(x = x.x + x.y/2, y = y.x + y.y/2) %>%
  ggplot(aes(x, y, fill = values, label = keys)) +
  geom_tile(alpha = 0.2) +
  geom_text()

Your example data has two aa and two gg so this is the output:

---

EDIT: here's a general approach for any k granularity. Here are two example data frames, with k = 3 and 4, respectively.

key <- c("a", "g", "t", "c")

# k = 3
df2 <- expand_grid(key1 = key, 
                   key2 = key,
                   key3 = key) %>%
  unite("keys", key1:key3, sep = "") %>%
  mutate(values = sin(seq(0, 60, length.out = 4^3)))

# k = 4
df2 <- expand_grid(key1 = key, 
                   key2 = key,
                   key3 = key,
                   key4 = key) %>%
  unite("keys", key1:key4, sep = "") %>%
  mutate(values = sin(seq(0, 60, length.out = 4^4)))

Then we could plug those into the following code:

array_size = max(str_count(df2$keys))

# reshape
df2 %>%
  separate(keys, paste0("col",1:array_size), sep = 1:(array_size-1), remove = FALSE) %>%
  mutate(id = row_number()) %>%
  pivot_longer(-c(keys, id, values)) %>%

  # calculate coordinates
  left_join(pos, by = c("value" = "key")) %>%
  mutate(divisor = 2^parse_number(name),
         across(c(x,y), ~./divisor))  %>%
  group_by(keys, id, values) %>%
  summarize(x = sum(x), y = sum(y)) %>% 

  # plotting
  ggplot(aes(x, y, fill = values, label = keys)) +
  geom_tile(alpha = 0.5) +
  geom_text(angle = 20)

Which produces the outputs below: