public void GetAltAllelesAndGenotypes_one_sample_ploidy_zero()
{
int[][] alleleCopyNumbers =
{
new[] { -1 }
};
var(altAlleleString, sampleGenotypes) = CanvasSegmentWriter.GetAltAllelesAndGenotypes(alleleCopyNumbers);
string[] expectedGenotypes = { "." };
Assert.Equal(".", altAlleleString);
Assert.Equal(expectedGenotypes, sampleGenotypes);
}
public void GetAltAllelesAndGenotypes_only_reference()
{
int[][] alleleCopyNumbers =
{
new[] { 1, 1 }, // Ref
new[] { 1, 1 } // Ref again
};
var(altAlleleString, sampleGenotypes) = CanvasSegmentWriter.GetAltAllelesAndGenotypes(alleleCopyNumbers);
string[] expectedGenotypes = { "0/0", "0/0" };
Assert.Equal(".", altAlleleString);
Assert.Equal(expectedGenotypes, sampleGenotypes);
}
public void GetAltAllelesAndGenotypes_hemizygous_regions()
{
int[][] alleleCopyNumbers =
{
new[] { 0 }, // Loss
new[] { 1 }, // Ref
new[] { 2 } // Gain
};
var(altAlleleString, sampleGenotypes) = CanvasSegmentWriter.GetAltAllelesAndGenotypes(alleleCopyNumbers);
string[] expectedGenotypes = { "1", "0", "2" };
Assert.Equal("<CN0>,<CN2>", altAlleleString);
Assert.Equal(expectedGenotypes, sampleGenotypes);
}
public void WriteHeaderAllAltCnTags_output_is_expected()
{
const int maxCopyNum = 3;
const string newLine = "\n";
string expected = string.Join(newLine, "##ALT=<ID=CN0,Description=\"Copy number allele: 0 copies\">", "##ALT=<ID=CN2,Description=\"Copy number allele: 2 copies\">", "##ALT=<ID=CN3,Description=\"Copy number allele: 3 copies\">") + newLine;
string output;
using (var stringWriter = new StringWriter())
{
stringWriter.NewLine = newLine;
CanvasSegmentWriter.WriteHeaderAllAltCnTags(new BgzipOrStreamWriter(stringWriter), maxCopyNum);
output = stringWriter.ToString();
}
Assert.Equal(expected, output);
}
public void GetAltAllelesAndGenotypes_various_genotypes()
{
int[][] alleleCopyNumbers =
{
new[] { -1, int.MaxValue }, // <DUP>
new[] { 1, 1 }, // Ref
new[] { 0, 1 }, // Loss
new[] { 1, 2 }, // Gain
new[] { 0, 3 }, // Gain
new[] { 2, 2 }, // Gain
new [] { 0, 2 }, // LOH
new [] { -1, -1 }, // CN=2, MCC null
new [] { -1 } // Ploidy == 0
};
var(altAlleleString, sampleGenotypes) = CanvasSegmentWriter.GetAltAllelesAndGenotypes(alleleCopyNumbers);
string[] expectedGenotypes = { "./4", "0/0", "0/1", "0/2", "1/3", "2/2", "1/2", "./.", "." };
Assert.Equal("<CN0>,<CN2>,<CN3>,<DUP>", altAlleleString);
Assert.Equal(expectedGenotypes, sampleGenotypes);
}
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);
}
internal int CallVariants(List <string> variantFrequencyFiles, List <string> segmentFiles,
IFileLocation outVcfFile, string ploidyBedPath, string referenceFolder, List <string> sampleNames, string commonCnvsBedPath, List <SampleType> sampleTypes)
{
// load files
// initialize data structures and classes
var fileCounter = 0;
var samplesInfo = new SampleMap <SampleMetrics>();
var sampleSegments = new SampleMap <Segments>();
var copyNumberModels = new SampleMap <ICopyNumberModel>();
var variantFrequencyFilesSampleList = new SampleMap <string>();
var kinships = new SampleMap <SampleType>();
foreach (string sampleName in sampleNames)
{
var sampleId = new SampleId(sampleName);
var segment = Segments.ReadSegments(_logger, new FileLocation(segmentFiles[fileCounter]));
segment.AddAlleles(CanvasIO.ReadFrequenciesWrapper(_logger, new FileLocation(variantFrequencyFiles[fileCounter]), segment.IntervalsByChromosome));
sampleSegments.Add(sampleId, segment);
var sampleInfo = SampleMetrics.GetSampleInfo(segment.AllSegments, ploidyBedPath, _callerParameters.NumberOfTrimmedBins, sampleId);
var copyNumberModel = _copyNumberModelFactory.CreateModel(_callerParameters.MaximumCopyNumber, sampleInfo.MaxCoverage, sampleInfo.MeanCoverage, sampleInfo.MeanMafCoverage);
samplesInfo.Add(sampleId, sampleInfo);
copyNumberModels.Add(sampleId, copyNumberModel);
variantFrequencyFilesSampleList.Add(sampleId, variantFrequencyFiles[fileCounter]);
kinships.Add(sampleId, sampleTypes[fileCounter]);
fileCounter++;
}
var segmentSetsFromCommonCnvs = CreateSegmentSetsFromCommonCnvs(variantFrequencyFilesSampleList,
_callerParameters.MinAlleleCountsThreshold, commonCnvsBedPath, sampleSegments);
var segmentsForVariantCalling = GetHighestLikelihoodSegments(segmentSetsFromCommonCnvs, samplesInfo, copyNumberModels).ToList();
PedigreeInfo pedigreeInfo = PedigreeInfo.GetPedigreeInfo(kinships, _callerParameters);
Parallel.ForEach(
segmentsForVariantCalling,
new ParallelOptions
{
MaxDegreeOfParallelism = Math.Min(Environment.ProcessorCount, _callerParameters.MaxCoreNumber)
},
segments => _variantCaller.CallVariant(segments, samplesInfo, copyNumberModels, pedigreeInfo)
);
var variantCalledSegments = new SampleMap <List <CanvasSegment> >();
foreach (var key in samplesInfo.SampleIds)
{
variantCalledSegments.Add(key, segmentsForVariantCalling.Select(segment => segment[key]).ToList());
}
var mergedVariantCalledSegments = MergeSegments(variantCalledSegments, _callerParameters.MinimumCallSize, _qualityFilterThreshold);
FilterExcessivelyShortSegments(mergedVariantCalledSegments);
var outputFolder = outVcfFile.Directory;
foreach (var sampleId in samplesInfo.SampleIds)
{
var coverageOutputPath = SingleSampleCallset.GetCoverageAndVariantFrequencyOutput(outputFolder,
sampleId.ToString());
CanvasSegment.WriteCoveragePlotData(mergedVariantCalledSegments[sampleId], samplesInfo[sampleId].MeanCoverage,
samplesInfo[sampleId].Ploidy, coverageOutputPath, referenceFolder);
}
bool isPedigreeInfoSupplied = pedigreeInfo != null && pedigreeInfo.HasFullPedigree();
var denovoQualityThreshold = isPedigreeInfoSupplied ? (int?)_deNovoQualityFilterThreshold : null;
var ploidies = samplesInfo.Select(info => info.Value.Ploidy).ToList();
var diploidCoverage = samplesInfo.Select(info => info.Value.MeanCoverage).ToList();
var names = samplesInfo.SampleIds.Select(id => id.ToString()).ToList();
CanvasSegmentWriter.WriteMultiSampleSegments(outVcfFile.FullName, mergedVariantCalledSegments, diploidCoverage, referenceFolder, names,
null, ploidies, _qualityFilterThreshold, denovoQualityThreshold, CanvasFilter.SegmentSizeCutoff, isPedigreeInfoSupplied);
foreach (var sampleId in samplesInfo.SampleIds)
{
var outputVcfPath = SingleSampleCallset.GetVcfOutput(outputFolder, sampleId.ToString());
var sampleMetrics = samplesInfo[sampleId];
var segments = mergedVariantCalledSegments[sampleId];
CanvasSegmentWriter.WriteSegments(outputVcfPath.FullName, segments,
sampleMetrics.MeanCoverage, referenceFolder, sampleId.ToString(), null,
sampleMetrics.Ploidy, _qualityFilterThreshold, isPedigreeInfoSupplied, denovoQualityThreshold, null);
var visualizationTemp = outputFolder.CreateSubdirectory($"VisualizationTemp{sampleId}");
var normalizationFactor = NormalizationCalculator.ComputeNormalizationFactor(segments);
var bigWig = _coverageBigWigWriter.Write(segments, visualizationTemp, normalizationFactor);
bigWig?.MoveTo(SingleSampleCallset.GetCoverageBigWig(outputFolder, sampleId.ToString()));
var copyNumberBedGraph = SingleSampleCallset.GetCopyNumberBedGraph(outputFolder, sampleId.ToString());
_copyNumberBedGraphWriter.Write(segments, sampleMetrics.Ploidy, copyNumberBedGraph);
var partitionBedgraphHeader = "track type=bedGraph visibility=full autoScale=on graphType=points";
var originalSegments = sampleSegments[sampleId];
_partitionCoverageBedGraphWriter.Write(originalSegments.AllSegments, SingleSampleCallset.GetPartitionBedGraph(outputFolder, sampleId.ToString()), normalizationFactor, partitionBedgraphHeader);
}
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.");
CanvasSegmentWriter.WriteSegments(outFile, this.Segments, Model?.DiploidCoverage, referenceFolder,
sampleName, null, null, QualityFilterThreshold, isPedigreeInfoSupplied: false);
return(0);
}
PloidyInfo ploidy = null;
if (!string.IsNullOrEmpty(ploidyBedPath))
{
ploidy = PloidyInfo.LoadPloidyFromBedFile(ploidyBedPath);
}
// load MAF
this.MeanCoverage = CanvasIO.LoadFrequencies(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.Alleles.Frequencies)
{
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);
}
CanvasSegment.AssignQualityScores(this.Segments, 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
CanvasSegment.MergeSegments(ref this.Segments);
// recalculating qscores after merging segments improves performance!
CanvasSegment.AssignQualityScores(this.Segments, CanvasSegment.QScoreMethod.LogisticGermline, germlineScoreParameters);
CanvasSegment.FilterSegments(QualityFilterThreshold, Segments);
List <string> extraHeaders = new List <string>();
string coverageOutputPath = SingleSampleCallset.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);
}
CanvasSegmentWriter.WriteSegments(outFile, this.Segments, Model.DiploidCoverage, referenceFolder, sampleName,
extraHeaders, ploidy, QualityFilterThreshold, isPedigreeInfoSupplied: false);
return(0);
}