Consider support of IUPAC codes in matrixOfRandomBarcodes or convenience function to derive possible barcodes

[j-andrews7]

Rather than truly random synthesis, there are semi-random barcode libraries that adhere to a given structure (e.g. clonTracer). In such cases, the ability to provide constrained barcode templates could be convenient.

Of course, the user could just generate them themselves and feed them to choices in the appropriate function with something like:

library(Biostrings)

iupac_codes <- list(
  A = "A", C = "C", G = "G", T = "T", 
  R = c("A", "G"), Y = c("C", "T"), S = c("G", "C"), 
  W = c("A", "T"), K = c("G", "T"), M = c("A", "C"), 
  B = c("C", "G", "T"), D = c("A", "G", "T"), H = c("A", "C", "T"), 
  V = c("A", "C", "G"), N = c("A", "C", "G", "T")
)

generate_possible_barcodes <- function(dna_string) {
  
  expand_char <- function(char) {
    iupac_codes[[toupper(char)]]
  }

  chars <- strsplit(dna_string, split = "")[[1]]
  possible_sequences_list <- lapply(chars, expand_char)
  sequences <- expand.grid(possible_sequences_list, stringsAsFactors = FALSE)

  sequences_str <- do.call(paste0, sequences)

  return(sequences_str)
}

sequences <- generate_possible_barcodes("ACTGWSWSWSAA")
print(sequences)
 [1] "ACTGAGAGAGAA" "ACTGTGAGAGAA" "ACTGACAGAGAA" "ACTGTCAGAGAA" "ACTGAGTGAGAA" "ACTGTGTGAGAA" "ACTGACTGAGAA" "ACTGTCTGAGAA" "ACTGAGACAGAA"
[10] "ACTGTGACAGAA" "ACTGACACAGAA" "ACTGTCACAGAA" "ACTGAGTCAGAA" "ACTGTGTCAGAA" "ACTGACTCAGAA" "ACTGTCTCAGAA" "ACTGAGAGTGAA" "ACTGTGAGTGAA"
[19] "ACTGACAGTGAA" "ACTGTCAGTGAA" "ACTGAGTGTGAA" "ACTGTGTGTGAA" "ACTGACTGTGAA" "ACTGTCTGTGAA" "ACTGAGACTGAA" "ACTGTGACTGAA" "ACTGACACTGAA"
[28] "ACTGTCACTGAA" "ACTGAGTCTGAA" "ACTGTGTCTGAA" "ACTGACTCTGAA" "ACTGTCTCTGAA" "ACTGAGAGACAA" "ACTGTGAGACAA" "ACTGACAGACAA" "ACTGTCAGACAA"
[37] "ACTGAGTGACAA" "ACTGTGTGACAA" "ACTGACTGACAA" "ACTGTCTGACAA" "ACTGAGACACAA" "ACTGTGACACAA" "ACTGACACACAA" "ACTGTCACACAA" "ACTGAGTCACAA"
[46] "ACTGTGTCACAA" "ACTGACTCACAA" "ACTGTCTCACAA" "ACTGAGAGTCAA" "ACTGTGAGTCAA" "ACTGACAGTCAA" "ACTGTCAGTCAA" "ACTGAGTGTCAA" "ACTGTGTGTCAA"
[55] "ACTGACTGTCAA" "ACTGTCTGTCAA" "ACTGAGACTCAA" "ACTGTGACTCAA" "ACTGACACTCAA" "ACTGTCACTCAA" "ACTGAGTCTCAA" "ACTGTGTCTCAA" "ACTGACTCTCAA"
[64] "ACTGTCTCTCAA"

But what's the fun in that, really? Probably simpler ways to do it, I am not great with Biostrings.

[LTLA]

Hm. Well, for your immediate problem, matrixOfRandomBarcodes will just report what it finds, so you might as well resolve ambiguous codes outside of the function. Helper functions notwithstanding, I'd be doing the same.

In the more general case, the various functions that use a known barcode pool could be modified to accept IUPAC codes in the barcodes. It's not too hard, but it's not a simple change, either, so I don't have a time frame for that.

[LTLA]

Making some progress with crisprVerse/kaori#10.

[LTLA]

All of the functions should now support IUPAC codes (except U and gaps) in the known barcode sequences:

# Creating an example dual barcode sequencing experiment.
known.pool <- c("AAARRRAAA", "AAAYYYAAA")

actual <- c("AAAAAAAAA", "AAACCCAAA", "AAAGGGAAA", "AAATTTAAA")
barcodes <- sprintf("CAGCTACGTACG%sCCAGCTCGATCG", actual)
names(barcodes) <- seq_along(barcodes)

library(Biostrings)
tmp <- tempfile(fileext=".fastq")
writeXStringSet(DNAStringSet(barcodes), filepath=tmp, format="fastq")
                                                                      
# Two of each:
countSingleBarcodes(tmp, choices=known.pool,
    template="CAGCTACGTACGNNNNNNNNNCCAGCTCGATCG")
## DataFrame with 2 rows and 2 columns
##       choices    counts
##   <character> <integer>
## 1   AAARRRAAA         2
## 2   AAAYYYAAA         2

In addition, we have a matchBarcodes() function to match observed sequences to a known pool:

choices <- c("AAARAA", "CCCYCC", "GGGMGG", "TTTSTT")
matchBarcodes(c("AAAAAA", "AAAGAA"), choices)
## DataFrame with 2 rows and 2 columns
##       index mismatches
##   <integer>  <integer>
## 1         1          0
## 2         1          0

This should be available in 1.1.4.