public BamRewriter(IBamReader bamReader, IBamWriter bamWriter, IAlignmentPairFilter filter,
IReadPairHandler pairHandler, long?bufferSize = 100000, bool getUnpaired = false, string chrFilter = null)
{
_bamReader = bamReader;
_bamWriter = bamWriter;
_filter = filter;
_pairHandler = pairHandler;
_bufferSize = bufferSize;
_getUnpaired = getUnpaired;
_chrFilter = chrFilter;
_alignmentBuffer = new List <BamAlignment>();
OnLog = message => Console.WriteLine(message);
}
private IEnumerable <BamAlignment> TryStitch(ReadPair readPair, IReadPairHandler pairHandler, out PairClassification classification)
{
// TODO if we end up allowing NM calculation in here, this will become true.
const bool allowStitchingOnImperfectReads = false;
IEnumerable <BamAlignment> bamAlignmentList = pairHandler.ExtractReads(readPair);
var bamAlignmentList2 = bamAlignmentList.ToList();
if (bamAlignmentList2.Count == 1)
{
readPair.Stitched = true;
classification = PairClassification.PerfectStitched;
if (allowStitchingOnImperfectReads)
{
var stitchedResult = bamAlignmentList2[0];
int?nm = 0;
//TODO handle this if it is a hit on performance. Making it simple for now because the previous logic where we were lazy evaluating was a bit skewed
var containsImperfections = ReadContainsImperfections(stitchedResult, _trustSoftclips);
//nm = stitchedResult.GetIntTag("NM"); // TODO reinstate this if stitched read has proper NM
var numMismatchesInR1 = readPair.Read1.GetIntTag("NM");
var numMismatchesInR2 = readPair.Read2.GetIntTag("NM");
if (containsImperfections ||
(nm > 0 || numMismatchesInR1 > 0 || numMismatchesInR2 > 0))
{
classification = PairClassification.ImperfectStitched;
if (numMismatchesInR1 <= NumMismatchesToBeConsideredLikelySnvInStitched &&
numMismatchesInR2 <= NumMismatchesToBeConsideredLikelySnvInStitched &&
!containsImperfections)
{
classification = PairClassification.SingleMismatchStitched;
}
else if (nm >= _numMismatchesToBeConsideredMessy ||
numMismatchesInR1 >= _numMismatchesToBeConsideredMessy ||
numMismatchesInR2 >= _numMismatchesToBeConsideredMessy)
{
classification = PairClassification.MessyStitched;
}
}
}
foreach (var alignment in bamAlignmentList)
{
foreach (var tag in _tagsToKeepFromR1)
{
var r1Tag = readPair.Read1.GetStringTag(tag);
if (r1Tag != null)
{
alignment.ReplaceOrAddStringTag(tag, r1Tag);
}
}
}
}
else
{
classification = PairClassification.FailStitch;
}
return(bamAlignmentList);
}
public PairResult GetBamAlignmentsAndClassification(ReadPair readPair, IReadPairHandler pairHandler)
{
if (readPair.PairStatus == PairStatus.Duplicate)
{
// TODO hasIndels and numMismatches and split don't have meaning in an unusable dup, but it's a bit misleading to set them..
// Also, it's kind of silly to extract those alignments if we're going to set it to unusable anyway.
var alignments = readPair.GetAlignments().ToList();
return(new PairResult(alignments: alignments, readPair: readPair, classification: PairClassification.Duplicate, hasIndels: false,
isSplit: false, numMismatchesInSingleton: 0, softclipLengthForIndelRead: 0)
{
IsReputableIndelContaining = false
});
}
var classification = PairClassification.Unknown;
IEnumerable <BamAlignment> bamAlignmentList = null;
int?numMismatchesInR1 = null;
int?numMismatchesInR2 = null;
var r1HasIndels = OverlappingIndelHelpers.ReadContainsIndels(readPair.Read1);
var r2HasIndels = OverlappingIndelHelpers.ReadContainsIndels(readPair.Read2);
var hasIndels = r1HasIndels || r2HasIndels;
if (IsCompletedPairedPair(readPair))
{
if (BothReadsHighQuality(readPair))
{
numMismatchesInR1 = readPair.Read1.GetIntTag("NM");
numMismatchesInR2 = readPair.Read2.GetIntTag("NM");
var tryStitch = true;
if (hasIndels)
{
if (numMismatchesInR1 == null && numMismatchesInR2 == null)
{
Logger.WriteWarningToLog(
$"Found indel-containing read without NM: '{readPair.Name}', likely indicating that NM is not set on any read. Consider preprocessing the BAM to calculate NM tags for best results.");
}
return(HandleIndelPairIfStitchUnallowed(readPair, numMismatchesInR1 ?? 0,
numMismatchesInR2 ?? 0, r1HasIndels, r2HasIndels));
}
else
{
// TODO if not realigning anything (or not realigning imperfects), go ahead and stitch immediately
// TODO why are we using this bool multiple times
// ^ Because there's no point checking for imperfectinos if we don't care about softclips -- we already know there are no indels because this is in the else of if(hasIndels)
if (!_trustSoftclips && (ReadContainsImperfections(readPair.Read1, _trustSoftclips) ||
ReadContainsImperfections(readPair.Read2, _trustSoftclips)))
{
tryStitch = false;
classification =
ClassifySoftclipContainingPairGivenSoftclipDistrust(readPair, numMismatchesInR1,
numMismatchesInR2);
bamAlignmentList = readPair.GetAlignments();
}
else
{
if (numMismatchesInR1 == null)
{
numMismatchesInR1 = readPair.Read1.GetIntTag("NM");
}
if (numMismatchesInR2 == null)
{
numMismatchesInR2 = readPair.Read2.GetIntTag("NM");
}
if (numMismatchesInR1 >= _numMismatchesToBeConsideredMessy ||
numMismatchesInR2 >= _numMismatchesToBeConsideredMessy)
{
classification = PairClassification.UnstitchMessy;
tryStitch = false;
if (numMismatchesInR1 <= 1 || numMismatchesInR2 <= 1)
{
// One of the reads is clean
tryStitch = false;
if (numMismatchesInR1 <= 1)
{
// R1 is the clean one.
if (readPair.Read2.IsReverseStrand())
{
classification = PairClassification.UnstitchReverseMessy;
}
else
{
classification = PairClassification.UnstitchForwardMessy;
}
}
else
{
if (readPair.Read1.IsReverseStrand())
{
classification = PairClassification.UnstitchReverseMessy;
}
else
{
classification = PairClassification.UnstitchForwardMessy;
}
}
}
bamAlignmentList = readPair.GetAlignments();
}
else if (numMismatchesInR1 + numMismatchesInR2 == 0)
{
classification = PairClassification.UnstitchPerfect;
bamAlignmentList = readPair.GetAlignments();
}
else if (numMismatchesInR1 <= 1 && numMismatchesInR2 <= 1)
{
classification = PairClassification.UnstitchSingleMismatch;
bamAlignmentList = readPair.GetAlignments();
}
else
{
classification = PairClassification.UnstitchImperfect;
bamAlignmentList = readPair.GetAlignments();
}
}
classification = AdjustClassificationForMultimapper(readPair, classification);
}
if (classification == PairClassification.UnstitchMessySuspiciousRead)
{
tryStitch = false;
}
if (_skipStitch)
{
tryStitch = false;
}
if (classification != PairClassification.UnstitchPerfect)
{
//For now we can't stitch anything else because we can't properly calculate NM!!
tryStitch = false;
}
if (!tryStitch)
{
if (bamAlignmentList == null)
{
bamAlignmentList = readPair.GetAlignments().ToList();
classification = PairClassification.Unstitchable;
}
}
else
{
bamAlignmentList = TryStitch(readPair, pairHandler, out classification);
}
}
else if (OneReadIsHighQuality(readPair))
{
classification = PairClassification.Split;
if (hasIndels)
{
numMismatchesInR1 = numMismatchesInR1 ?? readPair.Read1?.GetIntTag("NM") ?? 0;
numMismatchesInR2 = numMismatchesInR2 ?? readPair.Read2?.GetIntTag("NM") ?? 0;
return(HandlePairContainingIndels(readPair, r1HasIndels, r2HasIndels, numMismatchesInR1.Value,
numMismatchesInR2.Value, true, PairClassification.Split, true));
}
}
else
{
classification = PairClassification.Unusable;
bamAlignmentList = readPair.GetAlignments().ToList();
}
}
else
{
numMismatchesInR1 = numMismatchesInR1 ?? readPair.Read1?.GetIntTag("NM") ?? 0;
numMismatchesInR2 = numMismatchesInR2 ?? readPair.Read2?.GetIntTag("NM") ?? 0;
return(ClassifyIncompletePair(readPair, r1HasIndels, r2HasIndels, numMismatchesInR1.Value, numMismatchesInR2.Value));
}
// TODO - not sure why I originally had this double-check on whether pairs were split? shouldn't this already be evident from the pair status?
//var isSplit = bamAlignmentList?.Count() > 0 && bamAlignmentList?.Select(x => x.RefID).Distinct().Count() > 1;
var isSplit = false;
if (isSplit || classification == PairClassification.Split || readPair.PairStatus == PairStatus.SplitChromosomes ||
readPair.PairStatus == PairStatus.MateNotFound || readPair.PairStatus == PairStatus.MateUnmapped)
{
return(HandleSplitNonIndelPair(readPair, bamAlignmentList, hasIndels, isSplit));
}
var pr = new PairResult(bamAlignmentList.ToList(), readPair, classification, hasIndels, isSplit);
if (classification == PairClassification.UnstitchMessy || classification == PairClassification.UnstitchMessySuspiciousRead)
{
if (_checkMd && HasSuspiciousMd(readPair, numMismatchesInR1, numMismatchesInR2, pr))
{
classification = PairClassification.UnstitchMessySuspiciousMd;
pr.Classification = classification;
}
}
return(pr);
}
public PostRealignmentStitcher(IReadPairHandler stitchedPairHandler, IStatusHandler statusHandler, List <string> tagsToKeepFromR1 = null)
{
_stitchedPairHandler = stitchedPairHandler;
_statusHandler = statusHandler;
_tagsToKeepFromR1 = tagsToKeepFromR1 ?? new List <string>();
}
