private static ISampleMap <List <CanvasSegment> > MergeSegments(ISampleMap <List <CanvasSegment> > segments, int minimumCallSize, int qScoreThreshold)
{
int nSegments = segments.First().Value.Count;
var copyNumbers = new List <List <int> >(nSegments);
var qscores = new List <double>(nSegments);
foreach (int segmentIndex in Enumerable.Range(0, nSegments))
{
copyNumbers.Add(segments.Select(s => s.Value[segmentIndex].CopyNumber).ToList());
qscores.Add(segments.Select(s => s.Value[segmentIndex].QScore).Average());
}
if (copyNumbers == null && qscores != null || copyNumbers != null & qscores == null)
{
throw new ArgumentException("Both copyNumbers and qscores arguments must be specified.");
}
if (copyNumbers != null && copyNumbers.Count != nSegments)
{
throw new ArgumentException("Length of copyNumbers list should be equal to the number of segments.");
}
if (qscores != null && qscores.Count != nSegments)
{
throw new ArgumentException("Length of qscores list should be equal to the number of segments.");
}
var mergedSegments = new SampleMap <List <CanvasSegment> >();
foreach (var sampleSegments in segments)
{
var mergedSegmentsThisSample = CanvasSegment.MergeSegments(sampleSegments.Value.ToList(),
minimumCallSize, 10000, copyNumbers, qscores, qScoreThreshold);
mergedSegments.Add(sampleSegments.Key, mergedSegmentsThisSample);
}
return(mergedSegments);
}
public void TestMergeSegments()
{
// Construct several segments, and invoke CanvasSegment.MergeSegments, and ensure that the expected
// merges (and no others) occurred.
List <CanvasSegment> allSegments = new List <CanvasSegment>();
List <SampleGenomicBin> counts = new List <SampleGenomicBin>();
// Chr1 gets five segments and we should merge to three:
CanvasSegment seg = new CanvasSegment("chr1", 1000000, 2000000, counts);
seg.CopyNumber = 2;
allSegments.Add(seg);
seg = new CanvasSegment("chr1", 2000000, 2000100, counts);
seg.CopyNumber = 3;
allSegments.Add(seg);
seg = new CanvasSegment("chr1", 2000100, 3000000, counts);
seg.CopyNumber = 2;
allSegments.Add(seg);
seg = new CanvasSegment("chr1", 3000000, 3100000, counts);
seg.CopyNumber = 3;
allSegments.Add(seg);
seg = new CanvasSegment("chr1", 3100000, 4000000, counts);
seg.CopyNumber = 2;
allSegments.Add(seg);
// Chr2 gets segments with a large gap between, so can't merge:
seg = new CanvasSegment("chr2", 1000000, 2000000, counts);
seg.CopyNumber = 2;
allSegments.Add(seg);
seg = new CanvasSegment("chr2", 3000000, 3000100, counts);
seg.CopyNumber = 3;
allSegments.Add(seg);
seg = new CanvasSegment("chr2", 4000000, 5000000, counts);
seg.CopyNumber = 2;
allSegments.Add(seg);
// Chr3 has three segments that all merge to 1 big one:
seg = new CanvasSegment("chr3", 1000000, 2000000, counts);
seg.CopyNumber = 2;
allSegments.Add(seg);
seg = new CanvasSegment("chr3", 2000000, 3000000, counts);
seg.CopyNumber = 2;
allSegments.Add(seg);
seg = new CanvasSegment("chr3", 3000000, 4000000, counts);
seg.CopyNumber = 2;
allSegments.Add(seg);
var mergedSegments = CanvasSegment.MergeSegments(allSegments, 50000, 10000);
var segmentsByChromosome = CanvasSegment.GetSegmentsByChromosome(mergedSegments);
Assert.Equal(3, segmentsByChromosome["chr1"].Count);
Assert.Equal(3, segmentsByChromosome["chr2"].Count);
Assert.Single(segmentsByChromosome["chr3"]);
}
public void TestMergeSegments()
{
// Construct several segments, and invoke CanvasSegment.MergeSegments, and ensure that the expected
// merges (and no others) occurred.
List <CanvasSegment> allSegments = new List <CanvasSegment>();
List <float> counts = new List <float>();
// Chr1 gets five segments and we should merge to three:
CanvasSegment seg = new CanvasSegment("chr1", 1000000, 2000000, counts);
seg.CopyNumber = 2;
allSegments.Add(seg);
seg = new CanvasSegment("chr1", 2000000, 2000100, counts);
seg.CopyNumber = 3;
allSegments.Add(seg);
seg = new CanvasSegment("chr1", 2000100, 3000000, counts);
seg.CopyNumber = 2;
allSegments.Add(seg);
seg = new CanvasSegment("chr1", 3000000, 3100000, counts);
seg.CopyNumber = 3;
allSegments.Add(seg);
seg = new CanvasSegment("chr1", 3100000, 4000000, counts);
seg.CopyNumber = 2;
allSegments.Add(seg);
// Chr2 gets segments with a large gap between, so can't merge:
seg = new CanvasSegment("chr2", 1000000, 2000000, counts);
seg.CopyNumber = 2;
allSegments.Add(seg);
seg = new CanvasSegment("chr2", 3000000, 3000100, counts);
seg.CopyNumber = 3;
allSegments.Add(seg);
seg = new CanvasSegment("chr2", 4000000, 5000000, counts);
seg.CopyNumber = 2;
allSegments.Add(seg);
// Chr3 has three segments that all merge to 1 big one:
seg = new CanvasSegment("chr3", 1000000, 2000000, counts);
seg.CopyNumber = 2;
allSegments.Add(seg);
seg = new CanvasSegment("chr3", 2000000, 3000000, counts);
seg.CopyNumber = 2;
allSegments.Add(seg);
seg = new CanvasSegment("chr3", 3000000, 4000000, counts);
seg.CopyNumber = 2;
allSegments.Add(seg);
CanvasSegment.MergeSegments(ref allSegments, 50000, 10000);
Dictionary <string, List <CanvasSegment> > segmentsByChromosome = CanvasSegment.GetSegmentsByChromosome(allSegments);
Assert.AreEqual(segmentsByChromosome["chr1"].Count, 3);
Assert.AreEqual(segmentsByChromosome["chr2"].Count, 3);
Assert.AreEqual(segmentsByChromosome["chr3"].Count, 1);
}
public int CallVariants(string variantFrequencyFile, string inFile, string outFile, string ploidyVcfPath, string referenceFolder, string sampleName,
string truthDataPath)
{
if (!string.IsNullOrEmpty(truthDataPath))
{
_cnOracle = new CopyNumberOracle();
_cnOracle.LoadKnownCN(truthDataPath);
}
_segments = Segments.ReadSegments(_logger, new FileLocation(inFile));
_allSegments = _segments.AllSegments.ToList();
TempFolder = Path.GetDirectoryName(inFile);
if (_allSegments.Count == 0)
{
Console.WriteLine("CanvasDiploidCaller: No segments loaded; no CNV calls will be made.");
CanvasSegmentWriter.WriteSegments(outFile, _allSegments, _model?.DiploidCoverage, referenceFolder,
sampleName, null, null, QualityFilterThreshold, false, null, null);
return(0);
}
PloidyInfo ploidy = null;
if (!string.IsNullOrEmpty(ploidyVcfPath))
{
ploidy = PloidyInfo.LoadPloidyFromVcfFileNoSampleId(ploidyVcfPath);
}
// load MAF
var allelesByChromosome = CanvasIO.ReadFrequenciesWrapper(_logger, new FileLocation(variantFrequencyFile), _segments.IntervalsByChromosome);
_segments.AddAlleles(allelesByChromosome);
MeanCoverage = allelesByChromosome.SelectMany(x => x.Value).SelectMany(y => y.TotalCoverage).Average();
AggregateVariantCoverage(ref _allSegments);
// Create new models for different copy number states
InitializePloidies();
// Compute statistics on the copy number two regions
float[] diploidCounts = AggregateCounts(ref _allSegments);
_diploidCoverage = Utilities.Mean(diploidCounts);
_coverageWeightingFactor = CoverageWeighting / _diploidCoverage;
// new coverage model
_model = new CoverageModel {
DiploidCoverage = _diploidCoverage
};
List <SegmentInfo> segments = new List <SegmentInfo>();
foreach (CanvasSegment segment in _allSegments)
{
SegmentInfo info = new SegmentInfo {
Segment = segment
};
List <double> mafs = new List <double>();
foreach (float value in segment.Balleles.Frequencies)
{
mafs.Add(value > 0.5 ? 1 - value : value);
}
if (mafs.Count > 0)
{
info.Maf = Utilities.Median(mafs);
}
else
{
info.Maf = -1;
}
info.Coverage = Utilities.Median(segment.Counts);
info.Weight = _allSegments.Count > 100 ? segment.Length : segment.BinCount;
segments.Add(info);
}
AssignPloidyCallsDistance(_model);
CanvasSegment.AssignQualityScores(_allSegments, CanvasSegment.QScoreMethod.LogisticGermline, _germlineScoreParameters);
// Merge neighboring segments that got the same copy number call.
// merging segments requires quality scores so we do it after quality scores have been assigned
var mergedSegments = CanvasSegment.MergeSegments(_allSegments);
// recalculating qscores after merging segments improves performance!
CanvasSegment.AssignQualityScores(mergedSegments, CanvasSegment.QScoreMethod.LogisticGermline, _germlineScoreParameters);
CanvasSegment.SetFilterForSegments(QualityFilterThreshold, mergedSegments, CanvasFilter.SegmentSizeCutoff);
List <string> extraHeaders = new List <string>();
var coverageOutputPath = SingleSampleCallset.GetCoverageAndVariantFrequencyOutputPath(outFile);
CanvasSegment.WriteCoveragePlotData(mergedSegments, _model.DiploidCoverage, ploidy, coverageOutputPath, referenceFolder);
if (_cnOracle != null)
{
GenerateReportVersusKnownCopyNumber();
}
if (!string.IsNullOrEmpty(ploidy?.HeaderLine))
{
extraHeaders.Add(ploidy.HeaderLine);
}
CanvasSegmentWriter.WriteSegments(outFile, mergedSegments, _model.DiploidCoverage, referenceFolder, sampleName,
extraHeaders, ploidy, QualityFilterThreshold, false, null, null);
return(0);
}
public int CallVariants(string variantFrequencyFile, string inFile, string outFile, string ploidyBedPath, string referenceFolder, string sampleName,
string truthDataPath)
{
if (!string.IsNullOrEmpty(truthDataPath))
{
this.CNOracle = new CopyNumberOracle();
this.CNOracle.LoadKnownCN(truthDataPath);
}
this.Segments = CanvasSegment.ReadSegments(inFile);
this.TempFolder = Path.GetDirectoryName(inFile);
if (this.Segments.Count == 0)
{
Console.WriteLine("CanvasDiploidCaller: No segments loaded; no CNV calls will be made.");
CanvasSegment.WriteSegments(outFile, this.Segments, referenceFolder, sampleName, null, null);
return(0);
}
PloidyInfo ploidy = null;
if (!string.IsNullOrEmpty(ploidyBedPath))
{
ploidy = PloidyInfo.LoadPloidyFromBedFile(ploidyBedPath);
}
// load MAF
this.MeanCoverage = CanvasIO.LoadVariantFrequencies(variantFrequencyFile, this.Segments);
int medianVariantCoverage = AggregateVariantCoverage(ref this.Segments);
// Create new models for different copy number states
this.InitializePloidies();
// Compute statistics on the copy number two regions
float[] diploidCounts = AggregateCounts(ref this.Segments);
DiploidCoverage = CanvasCommon.Utilities.Mean(diploidCounts);
CoverageWeightingFactor = CoverageWeighting / DiploidCoverage;
// new coverage model
this.Model = new CoverageModel();
Model.DiploidCoverage = DiploidCoverage;
List <SegmentInfo> segments = new List <SegmentInfo>();
foreach (CanvasSegment segment in this.Segments)
{
SegmentInfo info = new SegmentInfo();
info.Segment = segment;
List <double> MAF = new List <double>();
foreach (float value in segment.VariantFrequencies)
{
MAF.Add(value > 0.5 ? 1 - value : value);
}
if (MAF.Count > 0)
{
info.MAF = CanvasCommon.Utilities.Median(MAF);
}
else
{
info.MAF = -1;
}
info.Coverage = CanvasCommon.Utilities.Median(segment.Counts);
if (this.Segments.Count > 100)
{
info.Weight = segment.End - segment.Begin;
}
else
{
info.Weight = segment.BinCount;
}
segments.Add(info);
}
// Assign copy number and major chromosome count for each segment
bool useGaussianMixtureModel = false; // For now, this is set false, since we saw weird performance on chrY (CANV-115):
if (useGaussianMixtureModel)
{
// optimize model covariance
double likelihood = FitGaussians(Model, segments);
AssignPloidyCallsGaussianMixture();
}
else
{
AssignPloidyCallsDistance(Model, segments, medianVariantCoverage);
}
// Merge neighboring segments that got the same copy number call.
CanvasSegment.MergeSegments(ref this.Segments);
CanvasSegment.AssignQualityScores(this.Segments, CanvasSegment.QScoreMethod.LogisticGermline);
List <string> extraHeaders = new List <string>();
string coverageOutputPath = CanvasCommon.Utilities.GetCoverageAndVariantFrequencyOutputPath(outFile);
CanvasSegment.WriteCoveragePlotData(this.Segments, Model.DiploidCoverage, ploidy, coverageOutputPath, referenceFolder);
if (this.CNOracle != null)
{
this.GenerateReportVersusKnownCN();
}
if (ploidy != null && !string.IsNullOrEmpty(ploidy.HeaderLine))
{
extraHeaders.Add(ploidy.HeaderLine);
}
CanvasSegment.WriteSegments(outFile, this.Segments, referenceFolder, sampleName, extraHeaders, ploidy);
return(0);
}