public static int?AnyIndelCoveredInMate(IEnumerable <IndelSite> readIndelPositions,
BamAlignment readWithoutIndels, BamAlignment readWithIndels, int anchorSize = 0)
{
if (readIndelPositions == null || !readIndelPositions.Any())
{
return(null);
}
if (readWithIndels.IsReverseStrand())
{
readIndelPositions = readIndelPositions.Reverse();
}
foreach (var indelPosition in readIndelPositions)
{
var coveredInR1 =
readWithoutIndels.ContainsPosition(indelPosition.PreviousMappedPosition - anchorSize, readWithIndels.RefID) &&
readWithoutIndels.ContainsPosition(indelPosition.NextMappedPosition + anchorSize, readWithIndels.RefID);
if (coveredInR1)
{
return(indelPosition.PreviousMappedPosition);
}
}
return(null);
}
private static void AddReadLevelIndelMetrics(BamAlignment bamAlignment, bool isReputable, bool stitched, IndelEvidence indelMetrics,
bool isRepeat)
{
indelMetrics.Observations++;
if (stitched)
{
indelMetrics.Stitched++;
}
else
{
if (bamAlignment.IsReverseStrand())
{
indelMetrics.Reverse++;
}
else
{
indelMetrics.Forward++;
}
}
if (isReputable)
{
indelMetrics.ReputableSupport++;
}
if (isRepeat)
{
indelMetrics.IsRepeat++;
}
if (!bamAlignment.IsMateMapped() || bamAlignment.MateRefID != bamAlignment.RefID)
{
indelMetrics.IsSplit++;
}
}
public override void MoveToNextRecord()
{
_isEnd = !_bamReader.GetNextAlignment(ref _currentBamAlignment, false);
if (_isEnd)
{
return;
}
// No memory allocation
_currentSerializedAlignment.RefID = _currentBamAlignment.RefID;
_currentSerializedAlignment.Position = _currentBamAlignment.Position;
_currentSerializedAlignment.AlignmentFlag = _currentBamAlignment.AlignmentFlag;
_currentSerializedAlignment.FragmentLength = _currentBamAlignment.FragmentLength;
_currentSerializedAlignment.MapQuality = _currentBamAlignment.MapQuality;
_currentSerializedAlignment.MatePosition = _currentBamAlignment.MatePosition;
_currentSerializedAlignment.MateRefID = _currentBamAlignment.MateRefID;
_currentSerializedAlignment.IsReverseStrand = _currentBamAlignment.IsReverseStrand();
_currentSerializedAlignment.Name = _currentBamAlignment.Name;
}
private static List <BamAlignment> IndelsDisagreeWithStrongMate(List <IndelSite> r1IndelPositions,
List <IndelSite> r2IndelPositions, BamAlignment read1,
BamAlignment read2, out bool disagree, int mismatchesAllowed = 1, int r1IndelAdjustment = 0,
int r2IndelAdjustment = 0, bool softclipWeakOne = true, int?r1Nm = null, int?r2Nm = null)
{
var checkBoth = true;
// TODO maybe also check if one of the reads has ins AND del
// TODO if we've grabbed this info here, propagate it out so we don't do it twice
// TODO indel adjustment should only actually remove insertions, no??
var read1Nm = r1Nm ?? read1.GetIntTag("NM");
var read2Nm = r2Nm ?? read2.GetIntTag("NM");
var read1AdjustedNm = read1Nm - r1IndelAdjustment;
var read2AdjustedNm = read2Nm - r2IndelAdjustment;
disagree = false;
var r1IndelPositionsUnique = r1IndelPositions != null && r2IndelPositions != null?GetUniqueIndelSites(r1IndelPositions, r2IndelPositions) : r1IndelPositions;
var r2IndelPositionsUnique = r1IndelPositions != null && r2IndelPositions != null?GetUniqueIndelSites(r2IndelPositions, r1IndelPositions) : r2IndelPositions;
// No sense doing further checks if there's nothing to disagree over...
if (r1IndelPositionsUnique.Any() || r2IndelPositionsUnique.Any())
{
var r1AdjustedClean = read1AdjustedNm <= mismatchesAllowed;
var r2AdjustedClean = read2AdjustedNm <= mismatchesAllowed;
var r1Clean = read1Nm <= mismatchesAllowed;
var r2Clean = read2Nm <= mismatchesAllowed;
var r1NumIndels = r1IndelPositions?.Count;
var r2NumIndels = r2IndelPositions?.Count;
var r1IsGood = r1AdjustedClean && (r1Clean || r1NumIndels <= 1);
var r2IsGood = r2AdjustedClean && (r2Clean || r2NumIndels <= 1);
if ((read1Nm != null && read2Nm != null) && (r1IsGood || r2IsGood))
{
if (r1IsGood)
{
var disagreeingPos = AnyIndelCoveredInMate(r2IndelPositionsUnique, read1, read2);
if (disagreeingPos != null)
{
disagree = true;
if (softclipWeakOne && !r2IsGood)
{
SoftclipAfterIndel(read2, read2.IsReverseStrand(), disagreeingPos.Value);
}
}
else
{
if (checkBoth)
{
disagreeingPos = AnyIndelCoveredInMate(r1IndelPositionsUnique, read2, read1);
if (disagreeingPos != null)
{
disagree = true;
}
}
}
}
else
{
var disagreeingPos = AnyIndelCoveredInMate(r1IndelPositionsUnique, read2, read1);
if (disagreeingPos != null)
{
disagree = true;
if (softclipWeakOne && !r1IsGood)
{
SoftclipAfterIndel(read1, read1.IsReverseStrand(), disagreeingPos.Value);
}
}
else
{
if (checkBoth)
{
disagreeingPos = AnyIndelCoveredInMate(r2IndelPositionsUnique, read1, read2);
if (disagreeingPos != null)
{
disagree = true;
}
}
}
}
}
}
// If both are good, and they disagree, should still say they disagree?
return(new List <BamAlignment>()
{
read1, read2
});
}
public bool AddSerializedAlignment(ref BamAlignment al)
{
byte[] byteArray = null;
int smallOffset = 0;
if (!_serializedAlignments.GetByteArrayAndOffset(
_offset,
ref byteArray,
ref smallOffset))
{
// Out of space
return(false);
}
int smallOffsetInitial = smallOffset;
if (!SerializeAlignment(ref al, ref byteArray, ref smallOffset))
{
// It didn't fit in the subarray. Try the next one.
// This math moves to the next array. For example, say the
// max size for 1 array is 1000, and we were at 1987.
// 1000 - (1987 - 1000 * (1987/1000)) = 13.
// 1987 + 13 = 2000
// 2000 is the first element of array number 2. (indexes start at 0).
_offset += _serializedAlignments.GetMaxArraySize() - (_offset - _serializedAlignments.GetMaxArraySize() * (_offset / _serializedAlignments.GetMaxArraySize()));
if (!_serializedAlignments.GetByteArrayAndOffset(
_offset,
ref byteArray,
ref smallOffset))
{
// Out of space
return(false);
}
smallOffsetInitial = smallOffset;
if (!SerializeAlignment(ref al, ref byteArray, ref smallOffset))
{
// We just checked that we have space. This should never fail.
throw new InvalidOperationException("Error: Check available memory. Serialization of alignment failed.");
}
}
int alignmentSize = smallOffset - smallOffsetInitial;
if (_bamAlignmentList.Count == 1000)
{
// Assume the first 1000 records are representative of the typical size
// Add 20% to ensure a memory reallocation is unlikely.
_bamAlignmentList.Capacity = (int)(1.2 * _serializedAlignments.SizeInBytes / (_offset / 1000));
}
_bamAlignmentList.Add(new SerializedBamAlignment(
_offset,
alignmentSize,
al.RefID,
al.Position,
al.AlignmentFlag,
al.FragmentLength,
al.MapQuality,
al.MatePosition,
al.MateRefID,
al.IsReverseStrand()));
_offset += (UInt64)alignmentSize;
return(true);
}
public void AddAlignment(BamAlignment alignment, ReadNumber readNumber = ReadNumber.NA)
{
var alignmentCopy = new BamAlignment(alignment);
if (alignmentCopy.IsPrimaryAlignment() && !alignmentCopy.IsSupplementaryAlignment())
{
if (FragmentSize == 0)
{
FragmentSize = Math.Abs(alignmentCopy.FragmentLength);
// Can be either F1R2 or F2R1
NormalPairOrientation = (!alignmentCopy.IsReverseStrand() && alignmentCopy.IsMateReverseStrand()) ||
(alignmentCopy.IsReverseStrand() && !alignmentCopy.IsMateReverseStrand());
if (NormalPairOrientation)
{
if (alignmentCopy.RefID == alignmentCopy.MateRefID)
{
if (!alignmentCopy.IsReverseStrand())
{
if (alignmentCopy.Position > alignmentCopy.MatePosition)
{
// RF
NormalPairOrientation = false;
}
}
else
{
if (alignmentCopy.MatePosition > alignmentCopy.Position)
{
// RF
NormalPairOrientation = false;
}
}
}
}
}
NumPrimaryReads++;
bool useForPos = true;
if (useForPos)
{
if (alignmentCopy.Position > MaxPosition)
{
MaxPosition = alignment.Position;
}
if (alignmentCopy.Position < MinPosition)
{
MinPosition = alignment.Position;
}
}
if (readNumber == ReadNumber.NA)
{
if (Read1 != null && Read2 != null)
{
throw new InvalidDataException($"Already have both primary alignments for {alignment.Name}.");
}
if (Read1 == null)
{
Read1 = alignmentCopy;
}
else
{
Read2 = alignmentCopy;
}
}
else if (readNumber == ReadNumber.Read1)
{
if (Read1 != null)
{
throw new InvalidDataException($"Already have a read 1 primary alignment for {alignment.Name}.");
}
Read1 = alignmentCopy;
}
else if (readNumber == ReadNumber.Read2)
{
if (Read2 != null)
{
throw new InvalidDataException($"Already have a read 2 primary alignment for {alignment.Name}.");
}
Read2 = alignmentCopy;
}
}
else if (alignmentCopy.IsSupplementaryAlignment())
{
switch (readNumber)
{
case ReadNumber.Read1:
if (Read1SupplementaryAlignments == null)
{
Read1SupplementaryAlignments = new List <BamAlignment>();
}
Read1SupplementaryAlignments.Add(alignmentCopy);
break;
case ReadNumber.Read2:
if (Read2SupplementaryAlignments == null)
{
Read2SupplementaryAlignments = new List <BamAlignment>();
}
Read2SupplementaryAlignments.Add(alignmentCopy);
break;
case ReadNumber.NA:
if (Read1SupplementaryAlignments == null)
{
Read1SupplementaryAlignments = new List <BamAlignment>();
}
Read1SupplementaryAlignments.Add(alignmentCopy);
break;
default:
throw new ArgumentOutOfRangeException(nameof(readNumber), readNumber, null);
}
}
else
{
switch (readNumber)
{
case ReadNumber.Read1:
if (Read1SecondaryAlignments == null)
{
Read1SecondaryAlignments = new List <BamAlignment>();
}
Read1SecondaryAlignments.Add(alignmentCopy);
break;
case ReadNumber.Read2:
if (Read2SecondaryAlignments == null)
{
Read2SecondaryAlignments = new List <BamAlignment>();
}
Read2SecondaryAlignments.Add(alignmentCopy);
break;
case ReadNumber.NA:
if (Read1SecondaryAlignments == null)
{
Read1SecondaryAlignments = new List <BamAlignment>();
}
Read1SecondaryAlignments.Add(alignmentCopy);
break;
default:
throw new ArgumentOutOfRangeException(nameof(readNumber), readNumber, null);
}
}
// Set as improper once we add any alignment that is flagged as improper
if (!alignmentCopy.IsProperPair())
{
IsImproper = true;
}
}
/// <summary>
/// Reads in a bam file and marks within the BitArrays which genomic mers are present.
/// </summary>
/// <param name="bamFile">bam file read alignments from.</param>
/// <param name="observedAlignments">Dictioanry of BitArrays, one for each chromosome, to store the alignments in.</param>
static void LoadObservedAlignmentsBAM(string bamFile, bool isPairedEnd, string chromosome, CanvasCoverageMode coverageMode, HitArray observed, Int16[] fragmentLengths)
{
// Sanity check: The .bai file must exist, in order for us to seek to our target chromosome!
string indexPath = bamFile + ".bai";
if (!File.Exists(indexPath))
{
throw new Exception(string.Format("Fatal error: Bam index not found at {0}", indexPath));
}
using (BamReader reader = new BamReader(bamFile))
{
int desiredRefIndex = -1;
desiredRefIndex = reader.GetReferenceIndex(chromosome);
if (desiredRefIndex == -1)
{
throw new ApplicationException(
string.Format("Unable to retrieve the reference sequence index for {0} in {1}.", chromosome,
bamFile));
}
bool result = reader.Jump(desiredRefIndex, 0);
if (!result)
{
// Note: This is not necessarily an error, it just means that there *are* no reads for this chromosome in this
// .bam file. That is not uncommon e.g. for truseq amplicon.
return;
}
int readCount = 0;
int keptReadCount = 0;
string header = reader.GetHeader();
BamAlignment alignment = new BamAlignment();
while (reader.GetNextAlignment(ref alignment, true))
{
readCount++;
// Flag check - Require reads to be aligned, passing filter, non-duplicate:
if (!alignment.IsMapped())
{
continue;
}
if (alignment.IsFailedQC())
{
continue;
}
if (alignment.IsDuplicate())
{
continue;
}
if (alignment.IsReverseStrand())
{
continue;
}
if (!alignment.IsMainAlignment())
{
continue;
}
// Require the alignment to start with 35 bases of non-indel:
if (alignment.CigarData[0].Type != 'M' || alignment.CigarData[0].Length < 35)
{
continue;
}
if (isPairedEnd && !alignment.IsProperPair())
{
continue;
}
int refID = alignment.RefID;
// quit if the current reference index is different from the desired reference index
if (refID != desiredRefIndex)
{
break;
}
if (refID == -1)
{
continue;
}
keptReadCount++;
if (coverageMode == CanvasCoverageMode.Binary)
{
observed.Data[alignment.Position] = 1;
}
else
{
observed.Set(alignment.Position);
}
// store fragment size, make sure it's within Int16 range and is positive (simplification for now)
if (coverageMode == CanvasCoverageMode.GCContentWeighted)
{
fragmentLengths[alignment.Position] = Convert.ToInt16(Math.Max(Math.Min(Int16.MaxValue, alignment.FragmentLength), 0));
}
}
Console.WriteLine("Kept {0} of {1} total reads", keptReadCount, readCount);
}
}
/// <summary>
/// Reads in a bam file and marks within the BitArrays which genomic mers are present.
/// </summary>
/// <param name="bamFile">bam file read alignments from.</param>
/// <param name="observedAlignments">Dictioanry of BitArrays, one for each chromosome, to store the alignments in.</param>
static void LoadObservedAlignmentsBAM(string bamFile, bool isPairedEnd, string chromosome, CanvasCoverageMode coverageMode, HitArray observed, Int16[] fragmentLengths)
{
// Sanity check: The .bai file must exist, in order for us to seek to our target chromosome!
string indexPath = bamFile + ".bai";
if (!File.Exists(indexPath))
{
throw new Exception(string.Format("Fatal error: Bam index not found at {0}", indexPath));
}
using (BamReader reader = new BamReader(bamFile))
{
int desiredRefIndex = -1;
desiredRefIndex = reader.GetReferenceIndex(chromosome);
if (desiredRefIndex == -1)
{
throw new ApplicationException(
string.Format("Unable to retrieve the reference sequence index for {0} in {1}.", chromosome,
bamFile));
}
bool result = reader.Jump(desiredRefIndex, 0);
if (!result)
{
// Note: This is not necessarily an error, it just means that there *are* no reads for this chromosome in this
// .bam file. That is not uncommon e.g. for truseq amplicon.
return;
}
int readCount = 0;
int keptReadCount = 0;
string header = reader.GetHeader();
BamAlignment alignment = new BamAlignment();
while (reader.GetNextAlignment(ref alignment, true))
{
readCount++;
// Flag check - Require reads to be aligned, passing filter, non-duplicate:
if (!alignment.IsMapped()) continue;
if (alignment.IsFailedQC()) continue;
if (alignment.IsDuplicate()) continue;
if (alignment.IsReverseStrand()) continue;
if (!alignment.IsMainAlignment()) continue;
// Require the alignment to start with 35 bases of non-indel:
if (alignment.CigarData[0].Type != 'M' || alignment.CigarData[0].Length < 35) continue;
if (isPairedEnd && !alignment.IsProperPair()) continue;
int refID = alignment.RefID;
// quit if the current reference index is different from the desired reference index
if (refID != desiredRefIndex)
break;
if (refID == -1)
continue;
keptReadCount++;
if (coverageMode == CanvasCoverageMode.Binary)
{
observed.Data[alignment.Position] = 1;
}
else
{
observed.Set(alignment.Position);
}
// store fragment size, make sure it's within Int16 range and is positive (simplification for now)
if (coverageMode == CanvasCoverageMode.GCContentWeighted)
fragmentLengths[alignment.Position] = Convert.ToInt16(Math.Max(Math.Min(Int16.MaxValue, alignment.FragmentLength), 0));
}
Console.WriteLine("Kept {0} of {1} total reads", keptReadCount, readCount);
}
}