Guessing V/J annotation from CDR3 sequence for cases when it is missing

Author: mikessh

Diff for: .gitignore

@@ -2,6 +2,7 @@ tmp/
 database/
 releases/
 out/
+misc/export
 
 ## Groovy / Java
 

Diff for: misc/misc.Rmd

@@ -0,0 +1,21 @@
+---
+title: "misc"
+output: html_document
+---
+
+```{r setup, include=FALSE}
+knitr::opts_chunk$set(echo = TRUE)
+```
+
+```{r}
+source("utils.R")
+```
+
+Generate a dump of annotated V/J segment sequences. This dump is used by CDRFixer to annotate cases when V/J is not supplied
+
+```{r}
+fread("../database/vdjdb.slim.txt") %>%
+  as.data.frame %>%
+  get_segment_parts %>%
+  fwrite("../res/segments.aaparts.txt", sep = "\t")
+```

Diff for: summary/pub_diff.sh renamed to misc/pub_diff.sh

@@ -0,0 +1,69 @@
+library(dplyr)
+library(stringr)
+library(data.table)
+
+### Segment processing
+get_segment_parts <- function(.df) {
+  .df.v <- .df %>%
+    select(species, gene, v.segm, v.end, cdr3) %>%
+    filter(v.end > 0) %>%
+    mutate(v.segm = str_split_fixed(v.segm, "[*,]", 2)[,1],
+           cdr3 = substr(cdr3, 1, v.end)) %>%
+    group_by(species, gene, v.segm, cdr3) %>%
+    summarise(count = n()) %>%
+    group_by(species, gene, cdr3, type = "V") %>%
+    summarise(segm = v.segm[which(count == max(count))][1])
+  
+  .df.j <- .df %>%
+    select(species, gene, j.segm, j.start, cdr3) %>%
+    filter(j.start > 0) %>%
+    mutate(j.segm = str_split_fixed(j.segm, "[*,]", 2)[,1],
+           cdr3 = substr(cdr3, j.start, nchar(cdr3))) %>%
+    group_by(species, gene, j.segm, cdr3) %>%
+    summarise(count = n()) %>%
+    group_by(species, gene, cdr3, type = "J") %>%
+    summarise(segm = j.segm[which(count == max(count))][1])
+  
+  rbind(.df.v, .df.j)
+}
+  
+
+
+### VDJtools export
+
+mock_codons <- c('GCT', 'TGT', 'GAT', 'GAA', 'TTT',
+                 'GGT', 'ATT', 'CAT', 'AAA', 'TTA',
+                 'ATG', 'AAT', 'CCT', 'CAA', 'CGT',
+                 'TCT', 'ACT', 'GTT', 'TGG', 'TAT')
+
+names(mock_codons) <- c('A', 'C', 'D', 'E', 'F',
+                        'G', 'I', 'H', 'K', 'L',
+                        'M', 'N', 'P', 'Q', 'R',
+                        'S', 'T', 'V', 'W', 'Y')
+
+mock_back_translate <- function(x) {
+  paste0(mock_codons[x], collapse = "")
+}
+
+# "CASS" %>% strsplit('') %>% lapply(mock_back_translate)
+
+as.vdjtools.df <- function(.df, .chain = c("beta", "alpha")) {
+  if (.chain == "beta") {
+    .df$cdr3aa <- .df$cdr3.beta
+    .df$v <- .df$v.beta
+    .df$j <- .df$j.beta
+  } else {
+    .df$cdr3aa <- .df$cdr3.alpha
+    .df$v <- .df$v.alpha
+    .df$j <- .df$j.alpha
+  }
+  
+  .df$cdr3nt <- cdr3.beta %>% 
+    strsplit('') %>% 
+    lapply(mock_back_translate)
+  
+  .df %>%
+    mutate(count = 1, freq = 1 / n(), d = "",
+           vend = -1, dstart = -1, dend = -1, jstart = -1) %>%
+    select(count, freq, cdr3nt, cdr3aa, v, d, j, vend, dstart, dend, jstart)
+}