CellBarocde: Cellular DNA Barcode Analysis toolkit

Documented in bc_splitVDJ

#########################################################################################
#########################################################################################
#
# File: splitVDJ.R
#
# Script to split barcodes from the genetic 'barcode mouse' construct as generated in
# the lab of Ton Schumacher (NKI, NL) in its remaining constant V, D and J elements and
# the modified elements (additions/deletions) in between those constant parts.
#
# Expected input: a data frame with a column named var, which contains the barcode names
# 	(a specific part of the sequence)
#
# Output: two data frames are created (add.del.ok and add.del.err), which contain
# 	columns with the remaining constant parts and inserted/deleted parts
#
#
# Author: Joost Beltman (LACDR, Leiden University, NL)
# Date: May 15, 2019
#
# Modified: Wenjie Sun
# Data: Mar 2021
#
#########################################################################################
#########################################################################################

#load required packages
# library(seqinr)
# library(Biostrings)

#' Parse VDJ recombination (experimental)
#'
#' Script to split barcodes from the genetic 'barcode mouse' construct as generated in
#' the lab of Ton Schumacher (NKI, NL) in its remaining constant V, D and J elements and
#' the modified elements (additions/deletions) in between those constant parts.
#'
#' @param seqs a character vector contains the barcode sequences.
#' @param v_part a string given the V part sequence. 
#' @param d_fwd a string given the D region forwrad sequence. 
#' @param d_inv a string given the D region inverted sequence. 
#' @param j_part a string given the J region sequence.
#' @return A list contains two data.frame named \code{add.del.ok} and \code{add.del.err}, which contain
#' columns with the remaining constant parts and inserted/deleted parts
#' 
#' @examples
#' ## prepare input sequence
#' seq_v <- c(
#'     "TCCAGTAGCTACTATCGTTACGAGTAGCTACTACCG",
#'     "TCCAGTAGCTACTATCGTTACGACGTAGCTACTACCG",
#'     "TCCATACTATCGTTACGACGTAGCTACTACG",
#'     "TCCAGTAGTCGTAACGATAGTAGAGTAGCTACTACCG"
#'     )
#'
#' ## split the sequences
#' bc_splitVDJ(seq_v)
#'
#' @export
bc_splitVDJ <- function(
    seqs, 
    v_part = "TCCAGTAG", 
    d_fwd = "TCTACTATCGTTACGAC", 
    d_inv = "GTCGTAACGATAGTAGA", 
    j_part = "GTAGCTACTACCG"
    ) {

    #define V, D and J input element
    Vpart <- "TCCAGTAG"
    Vpart.s <- seqinr::s2c(Vpart)
    Dfwd <- "TCTACTATCGTTACGAC"
    Dfwd.s <- seqinr::s2c(Dfwd)
    Dinv <- "GTCGTAACGATAGTAGA"
    Dinv.s <- seqinr::s2c(Dinv)
    Jpart <- "GTAGCTACTACCG"
    Jpart.s <- seqinr::s2c(Jpart)

    ref_seq <- list(
        Vpart = Vpart, Dfwd = Dfwd, Dinv = Dinv, Jpart = Jpart,
        Vpart.s = Vpart.s, Dfwd.s = Dfwd.s, Dinv.s = Dinv.s, Jpart.s = Jpart.s
    )

    #make data frame for additions and deletions when compared to input sequence
    #   add.del = as.data.frame( matrix(nrow = length(seqs), ncol = 13) )
    #   names(add.del) = c("V", "insVD", "D", "insDJ", "J", "delV", "delDl", "delDr", "delJ", "D.fwdorinv", "errV", "errD", "errJ")
    #   rownames(add.del) = seqs
    # NOTE:  Avoid sapply(); use vapply()
    add.del <- sapply(seqs, splitVDJ1seq, ref_seq = ref_seq) %>% t %>% data.frame
    rownames(add.del) <- seqs

    add.del.err <- add.del[add.del$errV==1 | add.del$errD==1 | add.del$errJ==1 ,]
    add.del.ok <- add.del[add.del$errV==0 & add.del$errD==0 & add.del$errJ==0 ,]

    list(add.del.ok = add.del.ok, add.del.err = add.del.err)
}


splitVDJ1seq <- function(curseq, ref_seq) {
    output <- as.list(vector("character", 13))
    names(output) <- c("V", "insVD", "D", "insDJ", "J", "delV", "delDl", "delDr", "delJ", "D.fwdorinv", "errV", "errD", "errJ")

    #get and split current sequence in list
    curseq.s <- seqinr::s2c(curseq)


    if (length(curseq.s) > 0) { #test if empty sequence
        #find part of V that overlaps	
        overlapV <- 0

        for (i in 1:min(length(curseq.s),length(ref_seq$Vpart.s)) ) {

            if (curseq.s[i] == ref_seq$Vpart.s[i]) {
                overlapV <- overlapV+1
            } else {
                break
            }
        }

        curseq.V <- substr(curseq, 1, overlapV)
    }	else { #empty sequence

        curseq.V <- ""

    }

    #remove detected V part from sequence	
    curseq.DJ <- substr(curseq, overlapV+1, nchar(curseq))
    curseq.DJ.s <- seqinr::s2c(curseq.DJ)

    if (length(curseq.DJ.s) > 0) {

        #find part of J that overlaps (note that in the loop we now need to start at end)
        overlapJ <- 0
        testlen <- min(length(ref_seq$Jpart.s), length(curseq.DJ.s))
        for (i in length(ref_seq$Jpart.s):(length(ref_seq$Jpart.s) - testlen + 1)) {

            #find index in sequence to compare to index in J part
            j <- length(curseq.DJ.s) - (length(ref_seq$Jpart.s)-i)
            if (curseq.DJ.s[j] == ref_seq$Jpart.s[i]) {
                overlapJ <- overlapJ+1
            } else {
                break
            }
        }

        curseq.J <- substr(curseq, nchar(curseq)-overlapJ+1, nchar(curseq))

    } else {

        curseq.J <- ""

    }

    #remove detected J part from sequence	
    curseq.D <- substr(curseq.DJ, 1, nchar(curseq.DJ)-overlapJ)

    if (nchar(curseq.D) > 0) {

        #match first to forward D sequence
        #start with longest possible sequences to find match, and decrease each time by 1
        err.fwd <- -1 # to see whether a match was found
        for (l.fwd in min(nchar(curseq.D), nchar(ref_seq$Dfwd)):1) {

            perms.fwd <- {}
            for (j in 1:(nchar(ref_seq$Dfwd)-l.fwd+1)) {
                perms.fwd <- c(perms.fwd, substr(ref_seq$Dfwd, j, j+l.fwd-1))
            }

            trymatch.fwd <- perms.fwd

            #first search exact match	for current length		
            for (i in seq_along(trymatch.fwd)) {

                #find position at which match occurs (-1 for no match)
                matchpos.fwd <- regexpr(trymatch.fwd[i], curseq.D, fixed=TRUE)

                #matchpos.fwd <- Biostrings::matchPattern(trymatch.fwd[i], curseq.D, max.mismatch=0, with.indels=FALSE)
                #if match found, step out of loop
                if (matchpos.fwd>0) {
                    #startpos <- start(matchpos.fwd)
                    #endpos <- end(matchpos.fwd)					
                    startpos <- matchpos.fwd
                    endpos <- matchpos.fwd+l.fwd-1

                    err.fwd <- 0 #remember wether match contains an error
                    break
                }
            }

            #if no exact match, allow for one mismatch in middle (last two at both sides must match)
            if (err.fwd<0 & l.fwd >= 5) {

                for (i in seq_along(trymatch.fwd)) {

                    matchpos.fwd <- Biostrings::matchPattern(trymatch.fwd[i], curseq.D, min.mismatch=1, max.mismatch=1, with.indels=FALSE)

                    #if match found, step out of loop
                    if (length(matchpos.fwd)>0) {

                        startpos <- start(matchpos.fwd)
                        endpos <- end(matchpos.fwd)

                        if (substr(curseq.D, startpos, startpos+1) == substr(trymatch.fwd[i], 1, 2) &
                            substr(curseq.D, endpos-1, endpos) == substr(trymatch.fwd[i], l.fwd-1, l.fwd)) {

                            err.fwd <- 1 #remember wether match contains an error
                            break

                        }
                    }

                }
            }

            #if match found, step out of loop
            if (err.fwd>=0) {

                #remember D part in curseq that was matched
                curseq.D.fwdmatch <- substr(curseq.D, startpos, endpos)
                curseq.D.fwdmatchVD <- substr(curseq.D, 1, startpos-1)
                curseq.D.fwdmatchDJ <- substr(curseq.D, endpos+1, nchar(curseq.D))

                #remember parts of D that are left for this match
                Dfwd.leftVD <- substr(ref_seq$Dfwd, 1, i-1)
                Dfwd.leftVD.s <- seqinr::s2c(Dfwd.leftVD)
                Dfwd.leftDJ <- substr(ref_seq$Dfwd, i+l.fwd, nchar(ref_seq$Dfwd))
                Dfwd.leftDJ.s <- seqinr::s2c(Dfwd.leftDJ)

                break
            }

        }

        #then match to inverted D sequence
        #start with longest possible sequences to find match, and decrease each time by 1
        err.inv=-1 # to see whether a match was found
        for (l.inv in min(nchar(curseq.D), nchar(ref_seq$Dinv)):1) {

            perms.inv <- {}
            for (j in 1:(nchar(ref_seq$Dinv)-l.inv+1)) {
                perms.inv <- c(perms.inv, substr(ref_seq$Dinv, j, j+l.inv-1))
            }

            trymatch.inv <- perms.inv

            #first search exact match	for current length		
            for (i in seq_along(trymatch.inv)) {

                #find position at which match occurs (-1 for no match)
                matchpos.inv <- regexpr(trymatch.inv[i], curseq.D, fixed=TRUE)

                #matchpos.inv <- Biostrings::matchPattern(trymatch.inv[i], curseq.D, max.mismatch=0, with.indels=FALSE)
                #if match found, step out of loop
                if (matchpos.inv>0) {
                    #startpos <- start(matchpos.inv)
                    #endpos <- end(matchpos.inv)					
                    startpos <- matchpos.inv
                    endpos <- matchpos.inv+l.inv-1

                    err.inv <- 0 #remember wether match contains an error
                    break
                }
            }

            #if no exact match, allow for one mismatch in middle (last two at both sides must match)
            if (err.inv<0 & l.inv >= 5) {

                for (i in seq_along(trymatch.inv)) {

                    matchpos.inv <- Biostrings::matchPattern(trymatch.inv[i], curseq.D, min.mismatch=1, max.mismatch=1, with.indels=FALSE)

                    #if match found, step out of loop
                    if (length(matchpos.inv)>0) {

                        startpos <- start(matchpos.inv)
                        endpos <- end(matchpos.inv)

                        if (substr(curseq.D, startpos, startpos+1) == substr(trymatch.inv[i], 1, 2) &
                            substr(curseq.D, endpos-1, endpos) == substr(trymatch.inv[i], l.inv-1, l.inv)) {

                            err.inv <- 1 #remember wether match contains an error
                            break

                        }
                    }

                }
            }

            #if match found, step out of loop
            if (err.inv>=0) {

                #remember D part in curseq that was matched
                curseq.D.invmatch <- substr(curseq.D, startpos, endpos)
                curseq.D.invmatchVD <- substr(curseq.D, 1, startpos-1)
                curseq.D.invmatchDJ <- substr(curseq.D, endpos+1, nchar(curseq.D))

                #remember parts of D that are left for this match
                Dinv.leftVD <- substr(ref_seq$Dinv, 1, i-1)
                Dinv.leftVD.s <- seqinr::s2c(Dinv.leftVD)
                Dinv.leftDJ <- substr(ref_seq$Dinv, i+l.inv, nchar(ref_seq$Dinv))
                Dinv.leftDJ.s <- seqinr::s2c(Dinv.leftDJ)

                break
            }

        }


        #find out whether forward or inverted D fits best when allowing a mismatch at both left and right
        if (l.fwd>4 & ((l.fwd > l.inv) | (l.fwd == l.inv & err.fwd <= err.inv))) { #in case of equal length choose forward match

            curseq.Dnew <- curseq.D.fwdmatch

            #find new inserted elements
            curseq.insVD <- curseq.D.fwdmatchVD
            curseq.insDJ <- curseq.D.fwdmatchDJ

            #find new deleted elements
            curseq.delVD <- Dfwd.leftVD
            curseq.delDJ <- Dfwd.leftDJ

            #set errors
            output$errD <- err.fwd

            #set to forward or unknown
            if ((l.fwd > l.inv) | (l.fwd == l.inv & err.fwd < err.inv)) {
                output$D.fwdorinv <- "fwd"
            } else {
                output$D.fwdorinv <- "unknown"			
            }
        } else if (l.inv>4) {

            curseq.Dnew <- curseq.D.invmatch

            #find new inserted elements
            curseq.insVD <- curseq.D.invmatchVD
            curseq.insDJ <- curseq.D.invmatchDJ

            #find new deleted elements
            curseq.delVD <- Dinv.leftVD
            curseq.delDJ <- Dinv.leftDJ

            #set errors
            output$errD <- err.inv

            #set to inverted
            output$D.fwdorinv <- "inv"	

        } else {

            #too short D fragment matches, so consider D to be empty
            curseq.Dnew <- ""

            #find new inserted elements
            if (curseq.V=="" & curseq.J!="") {
                curseq.insVD <- substr(curseq.D, 1, nchar(curseq.D))
                curseq.insDJ <- ""
            } else if (curseq.J=="" & curseq.V=="") {
                curseq.insVD <- ""
                curseq.insDJ <- substr(curseq.D, 1, nchar(curseq.D))
            } else {
                curseq.insVD <- substr(curseq.D, 1, floor(nchar(curseq.D)/2))
                curseq.insDJ <- substr(curseq.D, floor(nchar(curseq.D)/2)+1, nchar(curseq.D))
            }

            #find new deleted elements
            curseq.delVD <- substr(ref_seq$Dfwd, 1, nchar(ref_seq$Dfwd)/2)
            curseq.delDJ <- substr(ref_seq$Dfwd, nchar(ref_seq$Dfwd)/2+1, nchar(ref_seq$Dfwd))

            #set errors
            output$errD <- 0

            #set to inverted
            output$D.fwdorinv <- "short"	

        }

        #now the D element is known (including errors allowed at both sides), and the deletions at both sites
        output$D <- curseq.Dnew
        output$delDl <- curseq.delVD
        output$delDr <- curseq.delDJ


        #find out whether V part can be extended if 1 error is assumed
        curseq.VinsVD <- paste(curseq.V, curseq.insVD, sep="")
        curseq.VinsVD.s <- seqinr::s2c(curseq.VinsVD)

        #find overlapping part assuming that we start from the V side and move towards D
        #note that first mismatch position is at overlapV+1
        #below we look for a second mismatch position
        maxvalfromV <- min(length(ref_seq$Vpart.s), length(curseq.VinsVD.s))
        mismatchposV <- overlapV+1
        if ((maxvalfromV - overlapV) > 2) {
            for (overlapVerr in (overlapV+2):maxvalfromV) {
                if (curseq.VinsVD.s[overlapVerr] != ref_seq$Vpart.s[overlapVerr]) {
                    mismatchposV <- overlapVerr
                    break
                }
            }
            #if match occurs until the end of the relevant sequence part, remember this
            if (overlapVerr == maxvalfromV & curseq.VinsVD.s[overlapVerr] == ref_seq$Vpart.s[overlapVerr]) {
                mismatchposV <- overlapVerr+1
            }
        }

        #if there is no second mismatch position closeby, we assume there was an error and extend V
        if (mismatchposV > (overlapV+3)) {
            curseq.Vnew <- substr(curseq.VinsVD, 1, mismatchposV-1)
            curseq.insVDnewfromV <- substr(curseq.VinsVD, mismatchposV, nchar(curseq.VinsVD))
            curseq.delV <- substr(ref_seq$Vpart, mismatchposV, nchar(ref_seq$Vpart))
            errorsV=1
        } else { #otherwise no extension
            curseq.Vnew <- curseq.V
            curseq.insVDnewfromV <- curseq.insVD
            curseq.delV <- substr(ref_seq$Vpart, overlapV+1, nchar(ref_seq$Vpart))
            errorsV=0
        }

        #now we know the V part, the deletions from V and the insertions between V and D
        output$V <- curseq.Vnew
        output$insVD <- curseq.insVDnewfromV
        output$delV <- curseq.delV
        output$errV <- errorsV

        #find out whether J part can be extended if 1 error is assumed
        curseq.JinsDJ <- paste(curseq.insDJ, curseq.J, sep="")
        curseq.JinsDJ.s <- seqinr::s2c(curseq.JinsDJ)

        #find overlapping part assuming that we start from the J side and move towards D
        #note that first mismatch position is at overlapJ+1 (when seen from the back of the sequence)
        #below we look for a second mismatch position
        lenJ <- length(ref_seq$Jpart.s)
        lenJinsDJ <- length(curseq.JinsDJ.s)
        maxvalfromJ <- min(lenJ, lenJinsDJ)
        mismatchposJ <- overlapJ+1
        if ((maxvalfromJ - overlapJ) > 2) {
            for (overlapJerr in (overlapJ+2):maxvalfromJ) {
                if (curseq.JinsDJ.s[lenJinsDJ - overlapJerr + 1] != ref_seq$Jpart.s[lenJ - overlapJerr + 1]) {
                    mismatchposJ <- overlapJerr
                    break
                }
            }
            #if match occurs until the end of the relevant sequence part, remember this
            if (overlapJerr == maxvalfromJ & curseq.JinsDJ.s[lenJinsDJ - overlapJerr + 1] == ref_seq$Jpart.s[lenJ - overlapJerr + 1]) {
                mismatchposJ <- overlapJerr+1
            }
        }

        #if there is no second mismatch position closeby, we assume there was an error and extend V
        if (mismatchposJ > (overlapJ+3)) {
            curseq.Jnew <- substr(curseq.JinsDJ, lenJinsDJ - mismatchposJ + 2, lenJinsDJ)
            curseq.insDJnewfromJ <- substr(curseq.JinsDJ, 1, lenJinsDJ - mismatchposJ + 1)
            curseq.delJ <- substr(ref_seq$Jpart, 1, lenJ - mismatchposJ + 1)
            errorsJ=1
        } else { #otherwise no extension
            curseq.Jnew <- curseq.J
            curseq.insDJnewfromJ <- curseq.insDJ
            curseq.delJ <- substr(ref_seq$Jpart, 1, lenJ - overlapJ)
            errorsJ=0
        }

        #now we know the J part, the deletions between D and J and the insertions between V and D
        output$J <- curseq.Jnew
        output$insDJ <- curseq.insDJnewfromJ
        output$delJ <- curseq.delJ
        output$errJ <- errorsJ

    } else { #empty D element


        output$V <- curseq.V
        output$J <- curseq.J
        output$D <- ""
        output$insVD <- ""
        output$insDJ <- ""
        output$D.fwdorinv <- "short"

        #arbitrary choice of forward element
        output$delDl <- substr(ref_seq$Dfwd, 1, nchar(ref_seq$Dfwd)/2)
        output$delDr <- substr(ref_seq$Dfwd, nchar(ref_seq$Dfwd)/2+1, nchar(ref_seq$Dfwd))

        #deleted parts of V and J are simply the missing part of the original V and J
        output$delV <- substr(ref_seq$Vpart, nchar(curseq.V)+1, nchar(ref_seq$Vpart))
        output$delJ <- substr(ref_seq$Jpart, 1, nchar(ref_seq$Jpart)-nchar(curseq.J))

        #no errors at all in this case
        output$errV <- 0
        output$errD <- 0
        output$errJ <- 0
    }
    output
}