public CanvasCallset(
IFileLocation bam,
string sampleName,
IDirectoryLocation wholeGenomeFastaFolder,
IDirectoryLocation outputFolder,
IFileLocation kmerFasta,
IFileLocation filterBed,
IFileLocation ploidyBed,
IFileLocation normalVcfPath,
bool isDbSnpVcf,
IEnumerable<IFileLocation> normalBamPaths,
NexteraManifest manifest,
IFileLocation somaticVcfPath,
IFileLocation outputVcfPath)
{
Bam = new Bam(bam);
SampleName = sampleName;
WholeGenomeFastaFolder = wholeGenomeFastaFolder;
OutputFolder = outputFolder;
KmerFasta = kmerFasta;
FilterBed = filterBed;
PloidyBed = ploidyBed;
NormalVcfPath = normalVcfPath;
IsDbSnpVcf = isDbSnpVcf;
Manifest = manifest;
SomaticVcfPath = somaticVcfPath;
OutputVcfPath = outputVcfPath;
NormalBamPaths = normalBamPaths.Select(file => new Bam(file));
var genomeSizeXml = WholeGenomeFastaFolder.GetFileLocation("GenomeSize.xml");
GenomeMetadata = new GenomeMetadata();
GenomeMetadata.Deserialize(genomeSizeXml.FullName);
}
public static void WriteTargetBed(NexteraManifest manifest, BgzipOrStreamWriter writer, GenomeMetadata genome)
{
List<NexteraManifest.ManifestRegion> tempRegions = manifest.Regions;
if (genome != null)
{
tempRegions = new List<NexteraManifest.ManifestRegion>(manifest.Regions);
Dictionary<string, int> chromsomeIndexLookup = new Dictionary<string, int>();
//generate chromsome index lookup and sort
for (int chromosomeIndex = 0; chromosomeIndex < genome.Sequences.Count; chromosomeIndex++)
{
GenomeMetadata.SequenceMetadata sequence = genome.Sequences[chromosomeIndex];
chromsomeIndexLookup[sequence.Name] = chromosomeIndex;
}
tempRegions.Sort((a, b) => a.CompareTo(b, chromsomeIndexLookup));
}
foreach (NexteraManifest.ManifestRegion region in tempRegions)
{
TargetInterval interval = region.GetTargetInterval();
writer.WriteLine(string.Join("\t", new[]
{
interval.ReferenceName,
(interval.Begin - 1).ToString(CultureInfo.InvariantCulture),
interval.End.ToString(CultureInfo.InvariantCulture),
region.Name //region name is needed for PUMA metrics outputs to generate .coverage.csv file
}));
}
}
public RawRatioCalculator(NexteraManifest manifest, double minReferenceCount = 1,
double maxReferecneCount = double.PositiveInfinity)
{
_manifest = manifest;
_minReferenceCount = minReferenceCount;
_maxReferenceCount = maxReferecneCount;
}
/// <summary>
/// Assumes the bins are sorted by genomic coordinates
/// </summary>
/// <param name="bins">Bins whose counts are to be normalized</param>
/// <param name="countsByGC">An array of lists. Each array element (0-100) will hold a list of counts whose bins have the same GC content.</param>
/// <param name="counts">Will hold all of the autosomal counts present in 'bins'</param>
public static void GetCountsByGC(List <SampleGenomicBin> bins, NexteraManifest manifest, out List <float>[] countsByGC, out List <float> counts)
{
countsByGC = new List <float> [numberOfGCbins];
counts = new List <float>(bins.Count);
// Initialize the lists
for (int i = 0; i < countsByGC.Length; i++)
{
countsByGC[i] = new List <float>();
}
foreach (SampleGenomicBin bin in manifest == null ? bins : EnrichmentUtilities.GetOnTargetBins(bins, manifest))
{
if (!GenomeMetadata.SequenceMetadata.IsAutosome(bin.GenomicBin.Chromosome))
{
continue;
}
// Put the observed count in the GC-appropriate list.
countsByGC[bin.GenomicBin.GC].Add(bin.Count);
// Add to the global list of counts.
counts.Add(bin.Count);
}
}
/// <summary>
/// Output bed file of regions. Each region spans both probes and the target interval
/// Note that the BED format uses:
/// 0-based start position (inclusive) and 1-based end position (inclusive)
/// which is equivalent to saying:
/// 0-based start position (inclusive) and 0-based end position (exclusive)
/// </summary>
public static void WriteRegionBed(NexteraManifest manifest, string outputPath, GenomeMetadata genome)
{
using (BgzipOrStreamWriter writer = new BgzipOrStreamWriter(outputPath))
{
WriteRegionBed(manifest, writer, genome);
}
}
private IFileLocation WriteManifest(NexteraManifest manifest, IDirectoryLocation sandbox)
{
var path = sandbox.GetFileLocation(manifest.Name);
NexteraManifestUtils.WriteNexteraManifests(manifest, path.FullName);
return(path);
}
private IFileLocation CreateDbSnpVcfForManifest(IFileLocation fullDbSnpVcf, NexteraManifest manifest, IDirectoryLocation sandBox)
{
IFileLocation targetedDbSnpVcf = sandBox.GetFileLocation($"{manifest.Name}_{fullDbSnpVcf.Name}");
Isas.Manifests.NexteraManifest.VcfUtilities.IntersectVcfWithManifest(fullDbSnpVcf.FullName, targetedDbSnpVcf.FullName, manifest);
return(targetedDbSnpVcf);
}
private static void LoadBinCounts(string binnedPath, NexteraManifest manifest, out List <double> binCounts,
out List <int> onTargetIndices)
{
binCounts = new List <double>();
onTargetIndices = new List <int>();
var regionsByChrom = manifest.GetManifestRegionsByChromosome();
string currChrom = null;
List <NexteraManifest.ManifestRegion> regions = null; // 1-based regions
int regionIndex = -1;
bool onTarget = false;
using (GzipReader reader = new GzipReader(binnedPath))
{
string line;
string[] toks;
int binIdx = 0;
while ((line = reader.ReadLine()) != null)
{
toks = line.Split('\t');
string chrom = toks[0];
int start = int.Parse(toks[1]); // 0-based, inclusive
int stop = int.Parse(toks[2]); // 0-based, exclusive
if (currChrom != chrom)
{
currChrom = chrom;
onTarget = false;
if (!regionsByChrom.ContainsKey(currChrom))
{
regions = null;
}
else
{
regions = regionsByChrom[currChrom];
regionIndex = 0;
}
}
while (regions != null && regionIndex < regions.Count && regions[regionIndex].End < start + 1)
{
regionIndex++;
}
if (regions != null && regionIndex < regions.Count && regions[regionIndex].Start <= stop) // overlap
{
onTarget = true;
}
else
{
onTarget = false;
}
if (onTarget)
{
onTargetIndices.Add(binIdx);
}
binCounts.Add(double.Parse(toks[3]));
binIdx++;
}
}
}
public CanvasCallset(
IFileLocation bam,
string sampleName,
IDirectoryLocation wholeGenomeFastaFolder,
IDirectoryLocation outputFolder,
IFileLocation kmerFasta,
IFileLocation filterBed,
IFileLocation ploidyBed,
IFileLocation normalVcfPath,
bool isDbSnpVcf,
IEnumerable <IFileLocation> normalBamPaths,
NexteraManifest manifest,
IFileLocation somaticVcfPath,
IFileLocation outputVcfPath)
{
Bam = new Bam(bam);
SampleName = sampleName;
WholeGenomeFastaFolder = wholeGenomeFastaFolder;
OutputFolder = outputFolder;
KmerFasta = kmerFasta;
FilterBed = filterBed;
PloidyBed = ploidyBed;
NormalVcfPath = normalVcfPath;
IsDbSnpVcf = isDbSnpVcf;
Manifest = manifest;
SomaticVcfPath = somaticVcfPath;
OutputVcfPath = outputVcfPath;
NormalBamPaths = normalBamPaths.Select(file => new Bam(file));
var genomeSizeXml = WholeGenomeFastaFolder.GetFileLocation("GenomeSize.xml");
GenomeMetadata = new GenomeMetadata();
GenomeMetadata.Deserialize(genomeSizeXml.FullName);
}
private static void GetWeightedAverageBinCount(IEnumerable <string> binnedPaths, string mergedBinnedPath,
NexteraManifest manifest = null)
{
int sampleCount = binnedPaths.Count();
if (sampleCount == 1) // copy file
{
if (File.Exists(binnedPaths.First()))
{
if (File.Exists(mergedBinnedPath))
{
File.Delete(mergedBinnedPath);
}
File.Copy(binnedPaths.First(), mergedBinnedPath);
}
}
else // merge normal samples
{
double[] weights = new double[sampleCount];
List <double>[] binCountsBySample = new List <double> [sampleCount];
for (int sampleIndex = 0; sampleIndex < sampleCount; sampleIndex++)
{
string binnedPath = binnedPaths.ElementAt(sampleIndex);
var binCounts = new BinCounts(binnedPath, manifest: manifest);
List <double> counts = binCounts.AllCounts;
// If a manifest is available, get the median of bins overlapping the targeted regions only.
// For small panels, there could be a lot of bins with zero count and the median would be 0 if taken over all the bins, resulting in division by zero.
double median = binCounts.OnTargetMedianBinCount;
weights[sampleIndex] = median > 0 ? 1.0 / median : 0;
binCountsBySample[sampleIndex] = counts;
}
double weightSum = weights.Sum();
for (int i = 0; i < sampleCount; i++)
{
weights[i] /= weightSum;
} // so weights sum to 1
// Computed weighted average of bin counts across samples
using (GzipReader reader = new GzipReader(binnedPaths.First()))
using (GzipWriter writer = new GzipWriter(mergedBinnedPath))
{
string line;
string[] toks;
int lineIdx = 0;
while ((line = reader.ReadLine()) != null)
{
toks = line.Split('\t');
double weightedBinCount = 0;
for (int i = 0; i < sampleCount; i++)
{
weightedBinCount += weights[i] * binCountsBySample[i][lineIdx];
}
toks[3] = String.Format("{0}", weightedBinCount);
writer.WriteLine(String.Join("\t", toks));
lineIdx++;
}
}
}
}
/// <summary>
/// Pick the best normal control that has the smallest mean squared log-ratios (LR2s).
/// </summary>
/// <param name="tumorBinnedPath"></param>
/// <param name="normalBinnedPaths"></param>
/// <param name="bestBinnedPath"></param>
/// <param name="manifest"></param>
private static void GetBestLR2BinCount(string tumorBinnedPath, IEnumerable<string> normalBinnedPaths, string bestBinnedPath,
NexteraManifest manifest = null)
{
int bestNormalSampleIndex = 0;
int normalSampleCount = normalBinnedPaths.Count();
if (normalSampleCount > 1) // find the best normal
{
List<double[]> binCountsByNormalSample = new List<double[]>();
for (int normalSampleIndex = 0; normalSampleIndex < normalSampleCount; normalSampleIndex++)
{
string normalBinnedPath = normalBinnedPaths.ElementAt(normalSampleIndex);
var binCounts = new BinCounts(normalBinnedPath, manifest: manifest);
List<double> counts = binCounts.OnTargetCounts;
double median = binCounts.OnTargetMedianBinCount;
// If a manifest is available, get the median of bins overlapping the targeted regions only.
// For small panels, there could be a lot of bins with zero count and the median would be 0 if taken over all the bins, resulting in division by zero.
double weight = median > 0 ? 1.0 / median : 0;
binCountsByNormalSample.Add(counts.Select(cnt => cnt * weight).ToArray());
}
double[] tumorBinCounts;
{
var binCounts = new BinCounts(tumorBinnedPath, manifest: manifest);
List<double> counts = binCounts.OnTargetCounts;
double tumorMedian = binCounts.OnTargetMedianBinCount;
double tumorWeight = tumorMedian > 0 ? 1.0 / tumorMedian : 0;
tumorBinCounts = counts.Select(cnt => cnt * tumorWeight).ToArray();
}
// Find the best normal sample
bestNormalSampleIndex = -1;
double minMeanSquaredLogRatios = double.PositiveInfinity;
for (int normalSampleIndex = 0; normalSampleIndex < normalSampleCount; normalSampleIndex++)
{
// Get the sum of squared log ratios
var result = GetMeanSquaredLogRatios(tumorBinCounts, binCountsByNormalSample[normalSampleIndex]);
double meanSquaredLogRatios = result.Item1;
int ignoredBinCount = result.Item2;
// TODO: Skip a (bad) normal sample if too many bins were ignored.
// Donavan's script skips a normal sample if more than 100 log ratios is NA.
// The cut-off is likely panel-dependent.
if (meanSquaredLogRatios < minMeanSquaredLogRatios)
{
minMeanSquaredLogRatios = meanSquaredLogRatios;
bestNormalSampleIndex = normalSampleIndex;
}
}
}
// copy file
string srcBinnedPath = normalBinnedPaths.ElementAt(bestNormalSampleIndex);
if (File.Exists(srcBinnedPath))
{
if (File.Exists(bestBinnedPath)) { File.Delete(bestBinnedPath); }
File.Copy(srcBinnedPath, bestBinnedPath);
}
}
public PCAReferenceGenerator(IFileLocation sampleBinnedFile, IFileLocation pcaModelFile,
NexteraManifest manifest, double minBinCount = 1, double maxBinCount = double.PositiveInfinity)
{
_sampleBinnedFile = sampleBinnedFile;
_model = new PCAModel(pcaModelFile);
_manifest = manifest;
_minBinCount = minBinCount;
_maxBinCount = maxBinCount;
_ratioCalculator = new RawRatioCalculator(manifest, _minBinCount, _maxBinCount);
}
private void LoadBinCounts(IEnumerable <SampleGenomicBin> bins, NexteraManifest manifest)
{
if (manifest == null)
{
Counts = bins.Select(bin => (double)bin.Count).ToList();
}
else
{
LoadBinCounts(bins, manifest, out Counts, out OnTargetIndices);
}
}
private void LoadBinCounts(string binnedPath, NexteraManifest manifest)
{
if (manifest == null)
{
LoadBinCounts(binnedPath, out Counts);
}
else
{
LoadBinCounts(binnedPath, manifest, out Counts, out OnTargetIndices);
}
}
public LoessGCNormalizer(IEnumerable<GenomicBin> bins, NexteraManifest manifest, int robustnessIter = 2,
Func<float, double> countTransformer = null, Func<double, float> invCountTransformer = null)
{
this.bins = bins;
this.manifest = manifest;
if (robustnessIter >= 0) { this.robustnessIter = robustnessIter; }
if (countTransformer != null && invCountTransformer != null)
{
this.countTransformer = countTransformer;
this.invCountTransformer = invCountTransformer;
}
initialize();
}
public WeightedAverageReferenceGenerator(IEnumerable <IFileLocation> controlBinnedFiles, NexteraManifest manifest)
{
foreach (var binnedFile in controlBinnedFiles)
{
if (!binnedFile.Exists)
{
throw new FileNotFoundException(binnedFile.FullName + " does not exist.");
}
}
_controlBinnedFiles = controlBinnedFiles;
_manifest = manifest;
}
private static void LoadBinCounts(IEnumerable <SampleGenomicBin> bins, NexteraManifest manifest,
out List <double> binCounts, out List <int> onTargetIndices)
{
binCounts = new List <double>();
onTargetIndices = new List <int>();
var regionsByChrom = manifest.GetManifestRegionsByChromosome();
string currChrom = null;
List <NexteraManifest.ManifestRegion> regions = null; // 1-based regions
int regionIndex = -1;
bool onTarget = false;
int binIdx = 0;
foreach (var bin in bins)
{
if (currChrom != bin.GenomicBin.Chromosome)
{
currChrom = bin.GenomicBin.Chromosome;
onTarget = false;
if (!regionsByChrom.ContainsKey(currChrom))
{
regions = null;
}
else
{
regions = regionsByChrom[currChrom];
regionIndex = 0;
}
}
while (regions != null && regionIndex < regions.Count && regions[regionIndex].End < bin.Start + 1)
{
regionIndex++;
}
if (regions != null && regionIndex < regions.Count && regions[regionIndex].Start <= bin.Stop) // overlap
{
onTarget = true;
}
else
{
onTarget = false;
}
if (onTarget)
{
onTargetIndices.Add(binIdx);
}
binCounts.Add(bin.Count);
binIdx++;
}
}
private static void GetWeightedAverageBinCount(IEnumerable<string> binnedPaths, string mergedBinnedPath,
NexteraManifest manifest = null)
{
int sampleCount = binnedPaths.Count();
if (sampleCount == 1) // copy file
{
if (File.Exists(binnedPaths.First()))
{
if (File.Exists(mergedBinnedPath)) { File.Delete(mergedBinnedPath); }
File.Copy(binnedPaths.First(), mergedBinnedPath);
}
}
else // merge normal samples
{
double[] weights = new double[sampleCount];
List<double>[] binCountsBySample = new List<double>[sampleCount];
for (int sampleIndex = 0; sampleIndex < sampleCount; sampleIndex++)
{
string binnedPath = binnedPaths.ElementAt(sampleIndex);
var binCounts = new BinCounts(binnedPath, manifest: manifest);
List<double> counts = binCounts.AllCounts;
// If a manifest is available, get the median of bins overlapping the targeted regions only.
// For small panels, there could be a lot of bins with zero count and the median would be 0 if taken over all the bins, resulting in division by zero.
double median = binCounts.OnTargetMedianBinCount;
weights[sampleIndex] = median > 0 ? 1.0 / median : 0;
binCountsBySample[sampleIndex] = counts;
}
double weightSum = weights.Sum();
for (int i = 0; i < sampleCount; i++) { weights[i] /= weightSum; } // so weights sum to 1
// Computed weighted average of bin counts across samples
using (GzipReader reader = new GzipReader(binnedPaths.First()))
using (GzipWriter writer = new GzipWriter(mergedBinnedPath))
{
string line;
string[] toks;
int lineIdx = 0;
while ((line = reader.ReadLine()) != null)
{
toks = line.Split('\t');
double weightedBinCount = 0;
for (int i = 0; i < sampleCount; i++) { weightedBinCount += weights[i] * binCountsBySample[i][lineIdx]; }
toks[3] = String.Format("{0}", weightedBinCount);
writer.WriteLine(String.Join("\t", toks));
lineIdx++;
}
}
}
}
public CanvasTumorNormalEnrichmentInput(
Bam tumorBam,
Bam normalBam,
Vcf normalVcf,
Vcf somaticVcf,
GenomeMetadata genomeMetadata,
NexteraManifest nexteraManifest)
{
TumorBam = tumorBam;
NormalBam = normalBam;
NormalVcf = normalVcf;
SomaticVcf = somaticVcf;
GenomeMetadata = genomeMetadata;
NexteraManifest = nexteraManifest;
}
public LoessGCNormalizer(IEnumerable <SampleGenomicBin> bins, NexteraManifest manifest, int robustnessIter = 2,
Func <float, double> countTransformer = null, Func <double, float> invCountTransformer = null)
{
this.bins = bins;
this.manifest = manifest;
if (robustnessIter >= 0)
{
this.robustnessIter = robustnessIter;
}
if (countTransformer != null && invCountTransformer != null)
{
this.countTransformer = countTransformer;
this.invCountTransformer = invCountTransformer;
}
initialize();
}
public CanvasEnrichmentInput(Bam bam, GenomeMetadata genomeMetadata,
IEnumerable <Bam> controlBamPaths,
NexteraManifest nexteraManifest,
CanvasEnrichmentPrecomputedControl precomputedControl,
SamplePloidyInfo ploidyInfo,
IFileLocation predefinedBinsFile,
CanvasPcaModels pcaModels)
{
Bam = bam;
GenomeMetadata = genomeMetadata;
NexteraManifest = nexteraManifest;
PrecomputedControl = precomputedControl;
NormalBamPaths = new ReadOnlyCollection <Bam>(controlBamPaths.ToList());
PloidyInfo = ploidyInfo;
PredefinedBinsFile = predefinedBinsFile;
PcaModels = pcaModels;
}
private CanvasCallset GetCallset(ILogger logger)
{
AnalysisDetails analysisDetails = new AnalysisDetails(CommonOptions.OutputDirectory, CommonOptions.WholeGenomeFasta, CommonOptions.KmerFasta, CommonOptions.FilterBed, SingleSampleCommonOptions.PloidyVcf, null);
IFileLocation outputVcfPath = CommonOptions.OutputDirectory.GetFileLocation("CNV.vcf.gz");
var manifest = new NexteraManifest(_manifest.FullName, null, logger.Error);
CanvasCallset callSet = new CanvasCallset(
_tumorNormalOptions.TumorBam,
SingleSampleCommonOptions.SampleName,
SingleSampleCommonOptions.BAlleleSites,
SingleSampleCommonOptions.IsDbSnpVcf,
new[] { _normalBam },
manifest,
_tumorNormalOptions.SomaticVcf,
outputVcfPath,
analysisDetails);
return(callSet);
}
public CanvasCallset(
IFileLocation bam,
string sampleName,
IFileLocation normalVcfPath,
bool isDbSnpVcf,
IEnumerable <IFileLocation> normalBamPaths,
NexteraManifest manifest,
IFileLocation somaticVcfPath,
IFileLocation outputVcfPath,
AnalysisDetails analysisDetails)
{
SingleSampleCallset = new SingleSampleCallset(new Bam(bam), sampleName, normalVcfPath, isDbSnpVcf, analysisDetails.OutputFolder, outputVcfPath);
Manifest = manifest;
SomaticVcfPath = somaticVcfPath;
AnalysisDetails = analysisDetails;
NormalBamPaths = normalBamPaths.Select(file => new Bam(file));
}
private static void GetBinRatio(string tumorBinnedPath, string normalBinnedPath, string ratioBinnedPath,
string ploidyBedPath, NexteraManifest manifest = null)
{
PloidyInfo referencePloidy = String.IsNullOrEmpty(ploidyBedPath) ? null : PloidyInfo.LoadPloidyFromBedFile(ploidyBedPath);
double tumorMedian = (new BinCounts(tumorBinnedPath, manifest: manifest)).OnTargetMedianBinCount;
double normalMedian = (new BinCounts(normalBinnedPath, manifest: manifest)).OnTargetMedianBinCount;
double librarySizeFactor = (tumorMedian > 0 && normalMedian > 0) ? normalMedian / tumorMedian : 1;
using (GzipReader tumorReader = new GzipReader(tumorBinnedPath))
using (GzipReader normalReader = new GzipReader(normalBinnedPath))
using (GzipWriter writer = new GzipWriter(ratioBinnedPath))
{
string normalLine;
string tumorLine;
string[] normalToks;
string[] tumorToks;
double normalCount;
double tumorCount;
double ratio;
while ((normalLine = normalReader.ReadLine()) != null)
{
tumorLine = tumorReader.ReadLine();
normalToks = normalLine.Split('\t');
tumorToks = tumorLine.Split('\t');
normalCount = double.Parse(normalToks[3]);
tumorCount = double.Parse(tumorToks[3]);
// The weighted average count of a bin could be less than 1.
// Using these small counts for coverage normalization creates large ratios.
// It would be better to just drop these bins so we don't introduce too much noise into segmentation and CNV calling.
if (normalCount < 1)
{
continue;
} // skip the bin
string chrom = normalToks[0];
int start = int.Parse(normalToks[1]);
int end = int.Parse(normalToks[2]);
// get the normal ploidy from intervalsWithPloidyByChrom
double factor = CanvasDiploidBinRatioFactor * GetPloidy(referencePloidy, chrom, start, end) / 2.0;
ratio = tumorCount / normalCount * factor * librarySizeFactor;
normalToks[3] = String.Format("{0}", ratio);
writer.WriteLine(String.Join("\t", normalToks));
}
}
}
public BestLR2ReferenceGenerator(IFileLocation sampleBinnedFile, IEnumerable <IFileLocation> controlBinnedFiles,
NexteraManifest manifest)
{
if (!sampleBinnedFile.Exists)
{
throw new FileNotFoundException(sampleBinnedFile.FullName + " does not exist.");
}
foreach (var binnedFile in controlBinnedFiles)
{
if (!binnedFile.Exists)
{
throw new FileNotFoundException(binnedFile.FullName + " does not exist.");
}
}
_sampleBinnedFile = sampleBinnedFile;
_controlBinnedFiles = controlBinnedFiles;
_manifest = manifest;
}
/// <summary>
/// Perform GC normalization depending on the mode
/// </summary>
/// <param name="bins">Bins whose counts are to be normalized</param>
/// <param name="manifest"></param>
/// <param name="mode">GC normalization mode</param>
static void NormalizeByGC(List <SampleGenomicBin> bins, NexteraManifest manifest, CanvasGCNormalizationMode mode)
{
switch (mode)
{
case CanvasGCNormalizationMode.MedianByGC:
NormalizeByGC(bins, manifest: manifest);
break;
case CanvasGCNormalizationMode.LOESS:
var normalizer = new LoessGCNormalizer(bins, manifest, robustnessIter: 0,
countTransformer: x => (double)Math.Log(x),
invCountTransformer: x => (float)Math.Exp(x));
normalizer.Normalize();
break;
default:
throw new Illumina.Common.IlluminaException("Unsupported Canvas GC normalization mode: " + mode.ToString());
}
}
public CanvasCallset(
SingleSampleCallset singleSampleCallset,
AnalysisDetails analysisDetails,
IEnumerable <IFileLocation> normalBamPaths,
NexteraManifest manifest,
IFileLocation somaticVcfPath)
{
SingleSampleCallset = singleSampleCallset;
Manifest = manifest;
if (somaticVcfPath != null)
{
SomaticVcfPath = somaticVcfPath;
}
AnalysisDetails = analysisDetails;
if (normalBamPaths != null)
{
NormalBamPaths = normalBamPaths.Select(file => new Bam(file));
}
}
private CanvasCallset GetCallset(ILogger logger)
{
IFileLocation outputVcfPath = CommonOptions.OutputDirectory.GetFileLocation("CNV.vcf.gz");
var manifest = new NexteraManifest(_manifest.FullName, null, logger.Error);
CanvasCallset callSet = new CanvasCallset(
_tumorNormalOptions.TumorBam,
CommonOptions.SampleName,
CommonOptions.WholeGenomeFasta,
CommonOptions.OutputDirectory,
CommonOptions.KmerFasta,
CommonOptions.FilterBed,
CommonOptions.PloidyBed,
CommonOptions.BAlleleSites,
CommonOptions.IsDbSnpVcf,
new[] { _normalBam },
manifest,
_tumorNormalOptions.SomaticVcf,
outputVcfPath);
return callSet;
}
private CanvasCallset GetCallset(ILogger logger)
{
IFileLocation outputVcfPath = CommonOptions.OutputDirectory.GetFileLocation("CNV.vcf.gz");
var manifest = new NexteraManifest(_somaticEnrichmentOptions.Manifest.FullName, null, logger.Error);
CanvasCallset callSet = new CanvasCallset(
_somaticEnrichmentOptions.Bam,
CommonOptions.SampleName,
CommonOptions.WholeGenomeFasta,
CommonOptions.OutputDirectory,
CommonOptions.KmerFasta,
CommonOptions.FilterBed,
CommonOptions.PloidyBed,
CommonOptions.BAlleleSites,
CommonOptions.IsDbSnpVcf,
_somaticEnrichmentOptions.ControlBams,
manifest,
null,
outputVcfPath);
return callSet;
}
/// <summary>
/// Assumes the bins are sorted by genomic coordinates
/// </summary>
/// <param name="bins">Bins whose counts are to be normalized</param>
/// <param name="countsByGC">An array of lists. Each array element (0-100) will hold a list of counts whose bins have the same GC content.</param>
/// <param name="counts">Will hold all of the autosomal counts present in 'bins'</param>
public static void GetCountsByGC(List<GenomicBin> bins, NexteraManifest manifest, out List<float>[] countsByGC, out List<float> counts)
{
countsByGC = new List<float>[numberOfGCbins];
counts = new List<float>(bins.Count);
// Initialize the lists
for (int i = 0; i < countsByGC.Length; i++)
countsByGC[i] = new List<float>();
foreach (GenomicBin bin in manifest == null ? bins : EnrichmentUtilities.GetOnTargetBins(bins, manifest))
{
if (!GenomeMetadata.SequenceMetadata.IsAutosome(bin.Chromosome)) { continue; }
// Put the observed count in the GC-appropriate list.
countsByGC[bin.GC].Add(bin.Count);
// Add to the global list of counts.
counts.Add(bin.Count);
}
}
/// <summary>
/// Get the on-target bins by intersecting the manifest.
/// </summary>
/// <param name="bins"></param>
/// <param name="manifest"></param>
/// <returns></returns>
public static IEnumerable<GenomicBin> GetOnTargetBins(IEnumerable<GenomicBin> bins, NexteraManifest manifest)
{
var regionsByChrom = manifest.GetManifestRegionsByChromosome();
string currChrom = null;
List<NexteraManifest.ManifestRegion> regions = null; // 1-based regions
int regionIndex = -1;
bool offTarget = true;
foreach (GenomicBin bin in bins) // 0-based bins
{
if (currChrom != bin.Chromosome)
{
currChrom = bin.Chromosome;
offTarget = true;
if (!regionsByChrom.ContainsKey(currChrom))
{
regions = null;
}
else
{
regions = regionsByChrom[currChrom];
regionIndex = 0;
}
}
while (regions != null && regionIndex < regions.Count && regions[regionIndex].End < bin.Start + 1)
{
regionIndex++;
}
if (regions != null && regionIndex < regions.Count && regions[regionIndex].Start <= bin.Stop) // overlap
{
offTarget = false;
}
else
{
offTarget = true;
}
if (offTarget) { continue; } // ignore off-target bins
yield return bin;
}
}
private CanvasCallset GetCallset(ILogger logger)
{
IFileLocation outputVcfPath = CommonOptions.OutputDirectory.GetFileLocation("CNV.vcf.gz");
var manifest = new NexteraManifest(_somaticEnrichmentOptions.Manifest.FullName, null, logger.Error);
CanvasCallset callSet = new CanvasCallset(
_somaticEnrichmentOptions.Bam,
CommonOptions.SampleName,
CommonOptions.WholeGenomeFasta,
CommonOptions.OutputDirectory,
CommonOptions.KmerFasta,
CommonOptions.FilterBed,
CommonOptions.PloidyBed,
CommonOptions.BAlleleSites,
CommonOptions.IsDbSnpVcf,
_somaticEnrichmentOptions.ControlBams,
manifest,
null,
outputVcfPath);
return(callSet);
}
public static void WriteRegionBed(NexteraManifest manifest, BgzipOrStreamWriter writer, GenomeMetadata genome)
{
List<NexteraManifest.ManifestRegion> tempRegions = manifest.Regions;
if (genome != null)
{
tempRegions = new List<NexteraManifest.ManifestRegion>(manifest.Regions);
Dictionary<string, int> chromsomeIndexLookup = new Dictionary<string, int>();
//generate chromsome index lookup and sort
for (int chromosomeIndex = 0; chromosomeIndex < genome.Sequences.Count; chromosomeIndex++)
{
GenomeMetadata.SequenceMetadata sequence = genome.Sequences[chromosomeIndex];
chromsomeIndexLookup[sequence.Name] = chromosomeIndex;
}
tempRegions.Sort((a, b) => a.CompareTo(b, chromsomeIndexLookup));
}
foreach (NexteraManifest.ManifestRegion region in tempRegions)
{
writer.WriteLine(string.Format("{0}\t{1}\t{2}", region.Chromosome, region.Start - 1, region.End));
}
}
/// <summary>
/// Perform a simple GC normalization.
/// </summary>
/// <param name="bins">Bins whose counts are to be normalized.</param>
/// <param name="manifest"></param>
static void NormalizeByGC(List <SampleGenomicBin> bins, NexteraManifest manifest = null)
{
// DebugPrintCountsByGC(bins, "CountsByGC-Before.txt");
// An array of lists. Each array element (0-100) will hold a list of counts whose bins have the same GC content.
List <float>[] countsByGC;
// Will hold all of the autosomal counts present in 'bins'
List <float> counts;
EnrichmentUtilities.GetCountsByGC(bins, manifest, out countsByGC, out counts);
double globalMedian = Utilities.Median(counts);
double?[] medians = new double?[countsByGC.Length];
// Compute the median count for each GC bin
for (int gcBinIndex = 0; gcBinIndex < countsByGC.Length; gcBinIndex++)
{
if (countsByGC[gcBinIndex].Count >= defaultMinNumberOfBinsPerGC)
{
medians[gcBinIndex] = Utilities.Median(countsByGC[gcBinIndex]);
}
else
{
List <Tuple <float, float> > weightedCounts = GetWeightedCounts(countsByGC, gcBinIndex);
medians[gcBinIndex] = Utilities.WeightedMedian(weightedCounts);
}
}
// Divide each count by the median count of bins with the same GC content
for (int gcBinIndex = 0; gcBinIndex < bins.Count; gcBinIndex++)
{
double?median = medians[bins[gcBinIndex].GenomicBin.GC];
if (median != null && median > 0)
{
bins[gcBinIndex].Count = (float)(globalMedian * (double)bins[gcBinIndex].Count / median);
}
}
// DebugPrintCountsByGC(bins, "CountsByGC-After.txt");
}
private CanvasCallset GetCallset(ILogger logger)
{
AnalysisDetails analysisDetails = new AnalysisDetails(CommonOptions.OutputDirectory, CommonOptions.WholeGenomeFasta,
CommonOptions.KmerFasta, CommonOptions.FilterBed, SingleSampleCommonOptions.PloidyVcf, null);
IFileLocation outputVcfPath = CommonOptions.OutputDirectory.GetFileLocation("CNV.vcf.gz");
var manifest = new NexteraManifest(_somaticEnrichmentOptions.Manifest.FullName, null, logger.Error);
// TODO: refactor and remove the following two lines
manifest.CanvasControlBinnedPath = _somaticEnrichmentOptions.ControlBinned?.FullName;
manifest.CanvasBinSize = _somaticEnrichmentOptions.ControlBinSize;
CanvasCallset callSet = new CanvasCallset(
_somaticEnrichmentOptions.Bam,
SingleSampleCommonOptions.SampleName,
SingleSampleCommonOptions.BAlleleSites,
SingleSampleCommonOptions.IsDbSnpVcf,
_somaticEnrichmentOptions.ControlBams,
manifest,
null,
outputVcfPath,
analysisDetails);
return(callSet);
}
public static int Run(CanvasNormalizeParameters parameters)
{
NexteraManifest manifest = string.IsNullOrEmpty(parameters.manifestPath) ? null : new NexteraManifest(parameters.manifestPath, null, Console.WriteLine);
switch (parameters.normalizationMode)
{
case CanvasNormalizeMode.BestLR2:
GetBestLR2BinCount(parameters.tumorBedPath, parameters.normalBedPaths, parameters.weightedAverageNormalBedPath,
manifest: manifest);
break;
case CanvasNormalizeMode.WeightedAverage:
GetWeightedAverageBinCount(parameters.normalBedPaths, parameters.weightedAverageNormalBedPath, manifest: manifest);
break;
default:
throw new Exception(string.Format("Invalid CanvasNormalize mode '{0}'", parameters.normalizationMode));
}
GetBinRatio(parameters.tumorBedPath, parameters.weightedAverageNormalBedPath, parameters.outBedPath,
parameters.ploidyBedPath, manifest: manifest);
return(0);
}
public BinCounts(string binnedPath, NexteraManifest manifest = null)
{
BinnedPath = binnedPath;
Manifest = manifest;
LoadBinCounts();
}
public static void WriteNexteraManifests(NexteraManifest manifest, TextWriter writer)
{
writer.WriteLine("#{0}: {1}", "Manifest Type", "Regions");
writer.WriteLine("#{0}: {1}", "Target Region Count", manifest.Regions.Count);
writer.WriteLine("#{0}: {1}", "Date", DateTime.Now.ToShortDateString());
writer.WriteLine("[Header]");
//writer.WriteLine("Manifest Version\t1.0");
if (!string.IsNullOrEmpty(manifest.GenomeName))
{
writer.WriteLine("ReferenceGenome\t" + manifest.GenomeName);
}
writer.WriteLine("[Regions]");
List<string> headers = new List<string>();
if (manifest.ColumnNames != null && manifest.ColumnNames.Length > 0)
{
foreach (int columnNumber in manifest.ColumnNumbers)
{
if (columnNumber >= 0 && columnNumber < manifest.ColumnNames.Length)
{
headers.Add(manifest.ColumnNames[columnNumber]);
}
}
writer.WriteLine(string.Join("\t", headers.ToArray()));
}
if (manifest.Regions != null && manifest.Regions.Count > 0)
{
foreach (NexteraManifest.ManifestRegion region in manifest.Regions)
{
NexteraManifest.ManifestRegion tmpRegion = new NexteraManifest.ManifestRegion(region);
List<string> line = new List<string>();
foreach (string header in headers)
{
switch (header.ToLowerInvariant())
{
case "name":
line.Add(tmpRegion.Name);
break;
case "chromosome":
line.Add(tmpRegion.Chromosome);
break;
case "start":
case "amplicon start":
line.Add(tmpRegion.Start.ToString());
break;
case "end":
case "amplicon end":
line.Add(tmpRegion.End.ToString());
break;
case "startprobelength":
case "upstream probe length":
line.Add(tmpRegion.StartProbeLength.ToString());
break;
case "endprobelength":
case "downstream probe length":
line.Add(tmpRegion.EndProbeLength.ToString());
break;
case "groupname":
case "group name":
case "group":
case "ip group":
line.Add(tmpRegion.GroupName);
break;
}
}
writer.WriteLine(string.Join("\t", line.ToArray()));
}
}
}
public static NexteraManifest GetUpdatedNexteraManifestsWithNewRegions(NexteraManifest manifest, List<RegionStatistics> regionStats)
{
NexteraManifest nexteraManifestsWithNewRegions = new NexteraManifest(manifest);
// create a dictionary for new regions
Dictionary<string, RegionStatistics> regionStatsLookup = new Dictionary<string, RegionStatistics>();
foreach (RegionStatistics regionStat in regionStats)
{
regionStatsLookup.Add(regionStat.RegionName, regionStat);
}
//update the regions
if (nexteraManifestsWithNewRegions.Regions != null && nexteraManifestsWithNewRegions.Regions.Count > 0)
{
var newRegions = new List<NexteraManifest.ManifestRegion>();
foreach (NexteraManifest.ManifestRegion region in nexteraManifestsWithNewRegions.Regions)
{
NexteraManifest.ManifestRegion tmpRegion = new NexteraManifest.ManifestRegion(region);
if (!regionStatsLookup.ContainsKey(region.Name))
continue;
tmpRegion.Start = regionStatsLookup[region.Name].StartPosition;
tmpRegion.End = regionStatsLookup[region.Name].EndPosition;
newRegions.Add(tmpRegion);
}
nexteraManifestsWithNewRegions.Regions = newRegions;
}
return nexteraManifestsWithNewRegions;
}
private static void GetBinRatio(string tumorBinnedPath, string normalBinnedPath, string ratioBinnedPath,
string ploidyBedPath, NexteraManifest manifest = null)
{
PloidyInfo referencePloidy = String.IsNullOrEmpty(ploidyBedPath) ? null : PloidyInfo.LoadPloidyFromBedFile(ploidyBedPath);
double tumorMedian = (new BinCounts(tumorBinnedPath, manifest: manifest)).OnTargetMedianBinCount;
double normalMedian = (new BinCounts(normalBinnedPath, manifest: manifest)).OnTargetMedianBinCount;
double librarySizeFactor = (tumorMedian > 0 && normalMedian > 0) ? normalMedian / tumorMedian : 1;
using (GzipReader tumorReader = new GzipReader(tumorBinnedPath))
using (GzipReader normalReader = new GzipReader(normalBinnedPath))
using (GzipWriter writer = new GzipWriter(ratioBinnedPath))
{
string normalLine;
string tumorLine;
string[] normalToks;
string[] tumorToks;
double normalCount;
double tumorCount;
double ratio;
while ((normalLine = normalReader.ReadLine()) != null)
{
tumorLine = tumorReader.ReadLine();
normalToks = normalLine.Split('\t');
tumorToks = tumorLine.Split('\t');
normalCount = double.Parse(normalToks[3]);
tumorCount = double.Parse(tumorToks[3]);
// The weighted average count of a bin could be less than 1.
// Using these small counts for coverage normalization creates large ratios.
// It would be better to just drop these bins so we don't introduce too much noise into segmentation and CNV calling.
if (normalCount < 1) { continue; } // skip the bin
string chrom = normalToks[0];
int start = int.Parse(normalToks[1]);
int end = int.Parse(normalToks[2]);
// get the normal ploidy from intervalsWithPloidyByChrom
double factor = CanvasDiploidBinRatioFactor * GetPloidy(referencePloidy, chrom, start, end) / 2.0;
ratio = tumorCount / normalCount * factor * librarySizeFactor;
normalToks[3] = String.Format("{0}", ratio);
writer.WriteLine(String.Join("\t", normalToks));
}
}
}
/// <summary>
/// Calculates how many possible alignments corresponds to the desired number of observed alignments per bin.
/// </summary>
/// <param name="countsPerBin">Desired number of observed alignments per bin.</param>
/// <param name="possibleAlignments">BitArrays of possible alignments (unique mers).</param>
/// <param name="observedAlignments">BitArrays storing the observed alignments.</param>
/// <returns>Median alignment rate observed on the autosomes.</returns>
static int CalculateNumberOfPossibleAlignmentsPerBin(int countsPerBin, Dictionary<string, BitArray> possibleAlignments,
Dictionary<string, HitArray> observedAlignments, NexteraManifest manifest = null)
{
List<double> rates = new List<double>();
Dictionary<string, List<NexteraManifest.ManifestRegion>> manifestRegionsByChrom = null;
if (manifest != null)
{
manifestRegionsByChrom = manifest.GetManifestRegionsByChromosome();
}
List<ThreadStart> tasks = new List<ThreadStart>();
foreach (string chr in possibleAlignments.Keys)
{
// We don't want to include the sex chromosomes because they may not be copy number 2
if (!GenomeMetadata.SequenceMetadata.IsAutosome(chr))
continue;
HitArray observed = observedAlignments[chr];
BitArray possible = possibleAlignments[chr];
List<NexteraManifest.ManifestRegion> regions = null;
if (manifestRegionsByChrom != null)
{
if (!manifestRegionsByChrom.ContainsKey(chr)) { continue; }
regions = manifestRegionsByChrom[chr];
}
tasks.Add(new ThreadStart(() =>
{
int numberObserved = observed.CountSetBits(regions);
int numberPossible = CountSetBits(possible, regions);
double rate = numberObserved / (double)numberPossible;
lock (rates)
{
rates.Add(rate);
}
}));
}
Console.WriteLine("Launch CalculateNumberOfPossibleAlignmentsPerBin jobs...");
Console.Out.WriteLine();
//Parallel.ForEach(tasks, t => { t.Invoke(); }); //todo allow controling degree of parallelism
Illumina.SecondaryAnalysis.Utilities.DoWorkParallelThreads(tasks);
Console.WriteLine("CalculateNumberOfPossibleAlignmentsPerBin jobs complete.");
Console.Out.WriteLine();
double medianRate = CanvasCommon.Utilities.Median(rates);
return (int)(countsPerBin / medianRate);
}
/// <summary>
/// Perform variance stabilization by GC bins.
/// </summary>
/// <param name="bins">Bins whose counts are to be normalized.</param>
static bool NormalizeVarianceByGC(List<GenomicBin> bins, NexteraManifest manifest = null)
{
// DebugPrintCountsByGC(bins, "CountsByGCVariance-Before.txt");
// An array of lists. Each array element (0-100) will hold a list of counts whose bins have the same GC content.
List<float>[] countsByGC;
// Will hold all of the autosomal counts present in 'bins'
List<float> counts;
EnrichmentUtilities.GetCountsByGC(bins, manifest, out countsByGC, out counts);
// Estimate quartiles of all bins genomewide
var globalQuartiles = Utilities.Quartiles(counts);
// Will hold interquartile range (IQR) separately for each GC bin
List<float> localIQR = new List<float>(countsByGC.Length);
// Will hold quartiles separately for each GC bin
List<Tuple<float, float, float>> localQuartiles = new List<Tuple<float, float, float>>(countsByGC.Length);
// calculate interquartile range (IQR) for GC bins and populate localQuartiles list
for (int i = 0; i < countsByGC.Length; i++)
{
if (countsByGC[i].Count == 0)
{
localIQR.Add(-1f);
localQuartiles.Add(new Tuple<float, float, float>(-1f, -1f, -1f));
}
else if (countsByGC[i].Count >= defaultMinNumberOfBinsPerGC)
{
localQuartiles.Add(Utilities.Quartiles(countsByGC[i]));
localIQR.Add(localQuartiles[i].Item3 - localQuartiles[i].Item1);
}
else
{
List<Tuple<float, float>> weightedCounts = GetWeightedCounts(countsByGC, i);
double[] quartiles = Utilities.WeightedQuantiles(weightedCounts, new List<float>() { 0.25f, 0.5f, 0.75f });
localQuartiles.Add(new Tuple<float, float, float>((float)quartiles[0], (float)quartiles[1], (float)quartiles[2]));
localIQR.Add((float)(quartiles[2] - quartiles[0]));
}
}
// Identify if particular GC bins have IQR twice as large as IQR genomewide
float globalIQR = globalQuartiles.Item3 - globalQuartiles.Item1;
// Holder for GC bins with large IQR (compared to genomewide IQR)
int significantIQRcounter = 0;
for (int i = 10; i < 90; i++)
{
if (globalIQR < localIQR[i] * 2f)
significantIQRcounter++;
}
if (significantIQRcounter <= 0)
return false;
// Divide each count by the median count of bins with the same GC content
foreach (GenomicBin bin in bins)
{
var scaledLocalIqr = localIQR[bin.GC] * 0.8f;
if (globalIQR >= scaledLocalIqr) continue;
// ratio of GC bins and global IQRs
float iqrRatio = scaledLocalIqr / globalIQR;
var medianGCCount = localQuartiles[bin.GC].Item2;
bin.Count = medianGCCount + (bin.Count - medianGCCount) / iqrRatio;
}
// DebugPrintCountsByGC(bins, "CountsByGCVariance-After.txt");
return true;
}
public static void WriteNexteraManifests(NexteraManifest manifest, string path)
{
using (StreamWriter writer = new StreamWriter(path))
{
WriteNexteraManifests(manifest, writer);
}
}
/// <summary>
/// Get the on-target bins by intersecting the manifest.
/// </summary>
/// <param name="bins"></param>
/// <param name="manifest"></param>
/// <returns></returns>
public static IEnumerable <SampleGenomicBin> GetOnTargetBins(IEnumerable <SampleGenomicBin> bins, NexteraManifest manifest)
{
var regionsByChrom = manifest.GetManifestRegionsByChromosome();
string currChrom = null;
List <NexteraManifest.ManifestRegion> regions = null; // 1-based regions
int regionIndex = -1;
bool offTarget = true;
foreach (SampleGenomicBin bin in bins) // 0-based bins
{
if (currChrom != bin.GenomicBin.Chromosome)
{
currChrom = bin.GenomicBin.Chromosome;
offTarget = true;
if (!regionsByChrom.ContainsKey(currChrom))
{
regions = null;
}
else
{
regions = regionsByChrom[currChrom];
regionIndex = 0;
}
}
while (regions != null && regionIndex < regions.Count && regions[regionIndex].End < bin.Start + 1)
{
regionIndex++;
}
if (regions != null && regionIndex < regions.Count && regions[regionIndex].Start <= bin.Stop) // overlap
{
offTarget = false;
}
else
{
offTarget = true;
}
if (offTarget)
{
continue;
} // ignore off-target bins
yield return(bin);
}
}
/// <summary>
/// Perform a simple GC normalization.
/// </summary>
/// <param name="bins">Bins whose counts are to be normalized.</param>
/// <param name="manifest"></param>
static void NormalizeByGC(List<GenomicBin> bins, NexteraManifest manifest = null)
{
// DebugPrintCountsByGC(bins, "CountsByGC-Before.txt");
// An array of lists. Each array element (0-100) will hold a list of counts whose bins have the same GC content.
List<float>[] countsByGC;
// Will hold all of the autosomal counts present in 'bins'
List<float> counts;
EnrichmentUtilities.GetCountsByGC(bins, manifest, out countsByGC, out counts);
double globalMedian = Utilities.Median(counts);
double?[] medians = new double?[countsByGC.Length];
// Compute the median count for each GC bin
for (int gcBinIndex = 0; gcBinIndex < countsByGC.Length; gcBinIndex++)
{
if (countsByGC[gcBinIndex].Count >= defaultMinNumberOfBinsPerGC)
{
medians[gcBinIndex] = Utilities.Median(countsByGC[gcBinIndex]);
}
else
{
List<Tuple<float, float>> weightedCounts = GetWeightedCounts(countsByGC, gcBinIndex);
medians[gcBinIndex] = Utilities.WeightedMedian(weightedCounts);
}
}
// Divide each count by the median count of bins with the same GC content
for (int gcBinIndex = 0; gcBinIndex < bins.Count; gcBinIndex++)
{
double? median = medians[bins[gcBinIndex].GC];
if (median != null && median > 0)
bins[gcBinIndex].Count = (float)(globalMedian * (double)bins[gcBinIndex].Count / median);
}
// DebugPrintCountsByGC(bins, "CountsByGC-After.txt");
}
static int Main(string[] args)
{
Utilities.LogCommandLine(args);
string inFile = null;
string outFile = null;
bool doGCnorm = false;
bool doSizeFilter = false;
bool doOutlierRemoval = false;
string ffpeOutliersFile = null;
string manifestFile = null;
CanvasGCNormalizationMode gcNormalizationMode = CanvasGCNormalizationMode.MedianByGC;
string modeDescription = String.Format("gc normalization mode. Available modes: {0}. Default: {1}",
String.Join(", ", Enum.GetValues(typeof(CanvasGCNormalizationMode)).Cast <CanvasGCNormalizationMode>()),
gcNormalizationMode);
bool needHelp = false;
OptionSet p = new OptionSet()
{
{ "i|infile=", "input file - usually generated by CanvasBin", v => inFile = v },
{ "o|outfile=", "text file to output containing cleaned bins", v => outFile = v },
{ "g|gcnorm", "perform GC normalization", v => doGCnorm = v != null },
{ "s|filtsize", "filter out genomically large bins", v => doSizeFilter = v != null },
{ "r|outliers", "filter outlier points", v => doOutlierRemoval = v != null },
{ "f|ffpeoutliers=", "filter regions of FFPE biases", v => ffpeOutliersFile = v },
{ "t|manifest=", "Nextera manifest file", v => manifestFile = v },
{ "w|weightedmedian=", "Minimum number of bins per GC required to calculate weighted median", v => minNumberOfBinsPerGCForWeightedMedian = int.Parse(v) },
{ "m|mode=", modeDescription, v => gcNormalizationMode = Utilities.ParseCanvasGCNormalizationMode(v) },
{ "h|help", "show this message and exit", v => needHelp = v != null },
};
List <string> extraArgs = p.Parse(args);
if (needHelp)
{
ShowHelp(p);
return(0);
}
if (inFile == null || outFile == null)
{
ShowHelp(p);
return(0);
}
// Does the input file exist?
if (!File.Exists(inFile))
{
Console.WriteLine("CanvasClean.exe: File {0} does not exist! Exiting.", inFile);
return(1);
}
List <SampleGenomicBin> bins = CanvasIO.ReadFromTextFile(inFile);
if (doOutlierRemoval)
{
bins = RemoveOutliers(bins);
}
if (doSizeFilter)
{
bins = RemoveBigBins(bins);
}
// do not run FFPE outlier removal on targeted/low coverage data
if (ffpeOutliersFile != null && bins.Count < 50000)
{
ffpeOutliersFile = null;
}
// estimate localSD metric to use in doFFPEOutlierRemoval later and write to a text file
double LocalSD = -1.0;
if (ffpeOutliersFile != null)
{
LocalSD = getLocalStandardDeviation(bins);
CanvasIO.WriteLocalSDToTextFile(ffpeOutliersFile, LocalSD);
}
if (doGCnorm)
{
NexteraManifest manifest = manifestFile == null ? null : new NexteraManifest(manifestFile, null, Console.WriteLine);
List <SampleGenomicBin> strippedBins = gcNormalizationMode == CanvasGCNormalizationMode.MedianByGC
? RemoveBinsWithExtremeGC(bins, defaultMinNumberOfBinsPerGC, manifest: manifest)
: bins;
if (strippedBins.Count == 0)
{
Console.Error.WriteLine("Warning in CanvasClean: Coverage too low to perform GC correction; proceeding without GC correction");
}
else
{
bins = strippedBins;
NormalizeByGC(bins, manifest, gcNormalizationMode);
// Use variance normalization only on large exome panels and whole genome sequencing
// The treshold is set to 10% of an average number of bins on CanvasClean data
if (ffpeOutliersFile != null && bins.Count > 500000)
{
bool isNormalizeVarianceByGC = NormalizeVarianceByGC(bins, manifest: manifest);
// If normalization by variance was run (isNormalizeVarianceByGC), perform mean centering by using NormalizeByGC
if (isNormalizeVarianceByGC)
{
NormalizeByGC(bins, manifest, gcNormalizationMode);
}
}
}
}
if (ffpeOutliersFile != null)
{
// threshold 20 is derived to separate FF and noisy FFPE samples (derived from a training set of approx. 40 samples)
List <SampleGenomicBin> LocalMadstrippedBins = RemoveBinsWithExtremeLocalSD(bins, LocalSD, 20, outFile);
bins = LocalMadstrippedBins;
}
CanvasIO.WriteToTextFile(outFile, bins);
return(0);
}
/// <summary>
/// Bin alignments.
/// </summary>
/// <param name="referenceFile">Reference fasta file.</param>
/// <param name="binSize">Desired number of alignments per bin.</param>
/// <param name="possibleAlignments">BitArrays of possible alignments.</param>
/// <param name="observedAlignments">BitArrays of observed alignments.</param>
/// <param name="predefinedBins">Pre-defined bins. null if not available.</param>
/// <returns>A list of bins.</returns>
static List<GenomicBin> BinCounts(string referenceFile, int binSize, CanvasCoverageMode coverageMode, NexteraManifest manifest,
Dictionary<string, BitArray> possibleAlignments,
Dictionary<string, HitArray> observedAlignments,
Dictionary<string, Int16[]> fragmentLengths,
Dictionary<string, List<GenomicBin>> predefinedBins,
string outFile)
{
bool debugGCCorrection = false; // write value of GC bins and correction factor
Dictionary<string, GenericRead> fastaEntries = new Dictionary<string, GenericRead>();
List<string> chromosomes = new List<string>();
Int16 meanFragmentSize = 0;
Int16 meanFragmentCutoff = 3;
if (coverageMode == CanvasCoverageMode.GCContentWeighted)
meanFragmentSize = MeanFragmentSize(fragmentLengths);
using (FastaReader reader = new FastaReader(referenceFile))
{
GenericRead fastaEntry = new GenericRead();
// Loop through each chromosome in the reference.
while (reader.GetNextEntry(ref fastaEntry))
{
chromosomes.Add(fastaEntry.Name);
fastaEntries[fastaEntry.Name] = fastaEntry;
fastaEntry = new GenericRead();
}
}
// calculate GC content of the forward read at every position along the genome
Dictionary<string, byte[]> readGCContent = new Dictionary<string, byte[]>();
if (coverageMode == CanvasCoverageMode.GCContentWeighted)
{
byte gcCap = (byte)numberOfGCbins;
List<ThreadStart> normalizationTasks = new List<ThreadStart>();
foreach (KeyValuePair<string, Int16[]> fragmentLengthsKVP in fragmentLengths)
{
string chr = fragmentLengthsKVP.Key;
GenericRead fastaEntry = fastaEntries[chr];
normalizationTasks.Add(new ThreadStart(() =>
{
// contains GC content of the forward read at every position for current chr
byte[] gcContent = new byte[fastaEntry.Bases.Length];
int gcCounter = 0;
// Iteratively calculate GC content of "reads" using fasta genome reference
for (int pos = 0; pos < fastaEntry.Bases.Length - meanFragmentSize * meanFragmentCutoff - 1; pos++)
{
Int16 currentFragment = 0;
if (fragmentLengthsKVP.Value[pos] == 0)
currentFragment = meanFragmentSize;
else
currentFragment = Convert.ToInt16(Math.Min(fragmentLengthsKVP.Value[pos], meanFragmentSize * meanFragmentCutoff));
for (int i = pos; i < pos + currentFragment; i++)
{
switch (fastaEntry.Bases[i])
{
case 'C':
case 'c':
case 'G':
case 'g':
gcCounter++;
break;
default:
break;
}
}
if (gcCounter < 0)
gcCounter = 0;
gcContent[pos] = (byte)Math.Min(100 * gcCounter / currentFragment, gcCap);
gcCounter = 0;
}
lock (readGCContent)
{
readGCContent[chr] = gcContent;
}
}));
}
Console.WriteLine("{0} Launching normalization tasks.", DateTime.Now);
Console.Out.Flush();
//Parallel.ForEach(normalizationTasks, t => { t.Invoke(); });
Illumina.SecondaryAnalysis.Utilities.DoWorkParallelThreads(normalizationTasks);
Console.WriteLine("{0} Normalization tasks complete.", DateTime.Now);
Console.Out.Flush();
}
// populate observed and expected read GC bin vectors
float[] observedVsExpectedGC = new float[0];
if (coverageMode == CanvasCoverageMode.GCContentWeighted)
observedVsExpectedGC = ComputeObservedVsExpectedGC(observedAlignments, readGCContent, manifest, debugGCCorrection, outFile);
Dictionary<string, List<GenomicBin>> perChromosomeBins = new Dictionary<string, List<GenomicBin>>();
List<ThreadStart> binningTasks = new List<ThreadStart>();
foreach (KeyValuePair<string, GenericRead> fastaEntryKVP in fastaEntries)
{
string chr = fastaEntryKVP.Key;
if (!possibleAlignments.ContainsKey(chr)) continue;
if (predefinedBins != null && !predefinedBins.ContainsKey(chr)) continue;
BinTaskArguments args = new BinTaskArguments();
args.FastaEntry = fastaEntryKVP.Value;
args.Chromosome = chr;
args.PossibleAlignments = possibleAlignments[chr];
args.ObservedAlignments = observedAlignments[chr];
args.CoverageMode = coverageMode;
perChromosomeBins[chr] = predefinedBins == null ? new List<GenomicBin>() : predefinedBins[chr];
args.Bins = perChromosomeBins[chr];
args.BinSize = binSize;
if (coverageMode == CanvasCoverageMode.GCContentWeighted)
args.ReadGCContent = readGCContent[chr];
else
args.ReadGCContent = null;
args.ObservedVsExpectedGC = observedVsExpectedGC;
binningTasks.Add(new ThreadStart(() => { BinCountsForChromosome(args); }));
}
Console.WriteLine("{0} Launch BinCountsForChromosome jobs...", DateTime.Now);
Console.Out.WriteLine();
//Parallel.ForEach(binningTasks, t => { t.Invoke(); });
Illumina.SecondaryAnalysis.Utilities.DoWorkParallelThreads(binningTasks);
Console.WriteLine("{0} Completed BinCountsForChromosome jobs.", DateTime.Now);
Console.Out.WriteLine();
List<GenomicBin> finalBins = new List<GenomicBin>();
foreach (string chr in chromosomes)
{
if (!perChromosomeBins.ContainsKey(chr)) continue;
finalBins.AddRange(perChromosomeBins[chr]);
}
return finalBins;
}
private static void LoadBinCounts(string binnedPath, NexteraManifest manifest, out List<double> binCounts,
out List<int> onTargetIndices)
{
binCounts = new List<double>();
onTargetIndices = new List<int>();
var regionsByChrom = manifest.GetManifestRegionsByChromosome();
string currChrom = null;
List<NexteraManifest.ManifestRegion> regions = null; // 1-based regions
int regionIndex = -1;
bool onTarget = false;
using (GzipReader reader = new GzipReader(binnedPath))
{
string line;
string[] toks;
int binIdx = 0;
while ((line = reader.ReadLine()) != null)
{
toks = line.Split('\t');
string chrom = toks[0];
int start = int.Parse(toks[1]); // 0-based, inclusive
int stop = int.Parse(toks[2]); // 0-based, exclusive
if (currChrom != chrom)
{
currChrom = chrom;
onTarget = false;
if (!regionsByChrom.ContainsKey(currChrom))
{
regions = null;
}
else
{
regions = regionsByChrom[currChrom];
regionIndex = 0;
}
}
while (regions != null && regionIndex < regions.Count && regions[regionIndex].End < start + 1)
{
regionIndex++;
}
if (regions != null && regionIndex < regions.Count && regions[regionIndex].Start <= stop) // overlap
{
onTarget = true;
}
else
{
onTarget = false;
}
if (onTarget) { onTargetIndices.Add(binIdx); }
binCounts.Add(double.Parse(toks[3]));
binIdx++;
}
}
}
public LSNormRatioCalculator(NexteraManifest manifest)
{
_manifest = manifest;
}
/// <summary>
/// Pick the best normal control that has the smallest mean squared log-ratios (LR2s).
/// </summary>
/// <param name="tumorBinnedPath"></param>
/// <param name="normalBinnedPaths"></param>
/// <param name="bestBinnedPath"></param>
/// <param name="manifest"></param>
private static void GetBestLR2BinCount(string tumorBinnedPath, IEnumerable <string> normalBinnedPaths, string bestBinnedPath,
NexteraManifest manifest = null)
{
int bestNormalSampleIndex = 0;
int normalSampleCount = normalBinnedPaths.Count();
if (normalSampleCount > 1) // find the best normal
{
List <double[]> binCountsByNormalSample = new List <double[]>();
for (int normalSampleIndex = 0; normalSampleIndex < normalSampleCount; normalSampleIndex++)
{
string normalBinnedPath = normalBinnedPaths.ElementAt(normalSampleIndex);
var binCounts = new BinCounts(normalBinnedPath, manifest: manifest);
List <double> counts = binCounts.OnTargetCounts;
double median = binCounts.OnTargetMedianBinCount;
// If a manifest is available, get the median of bins overlapping the targeted regions only.
// For small panels, there could be a lot of bins with zero count and the median would be 0 if taken over all the bins, resulting in division by zero.
double weight = median > 0 ? 1.0 / median : 0;
binCountsByNormalSample.Add(counts.Select(cnt => cnt * weight).ToArray());
}
double[] tumorBinCounts;
{
var binCounts = new BinCounts(tumorBinnedPath, manifest: manifest);
List <double> counts = binCounts.OnTargetCounts;
double tumorMedian = binCounts.OnTargetMedianBinCount;
double tumorWeight = tumorMedian > 0 ? 1.0 / tumorMedian : 0;
tumorBinCounts = counts.Select(cnt => cnt * tumorWeight).ToArray();
}
// Find the best normal sample
bestNormalSampleIndex = -1;
double minMeanSquaredLogRatios = double.PositiveInfinity;
for (int normalSampleIndex = 0; normalSampleIndex < normalSampleCount; normalSampleIndex++)
{
// Get the sum of squared log ratios
var result = GetMeanSquaredLogRatios(tumorBinCounts, binCountsByNormalSample[normalSampleIndex]);
double meanSquaredLogRatios = result.Item1;
int ignoredBinCount = result.Item2;
// TODO: Skip a (bad) normal sample if too many bins were ignored.
// Donavan's script skips a normal sample if more than 100 log ratios is NA.
// The cut-off is likely panel-dependent.
if (meanSquaredLogRatios < minMeanSquaredLogRatios)
{
minMeanSquaredLogRatios = meanSquaredLogRatios;
bestNormalSampleIndex = normalSampleIndex;
}
}
}
// copy file
string srcBinnedPath = normalBinnedPaths.ElementAt(bestNormalSampleIndex);
if (File.Exists(srcBinnedPath))
{
if (File.Exists(bestBinnedPath))
{
File.Delete(bestBinnedPath);
}
File.Copy(srcBinnedPath, bestBinnedPath);
}
}
/// <summary>
/// Perform GC normalization depending on the mode
/// </summary>
/// <param name="bins">Bins whose counts are to be normalized</param>
/// <param name="manifest"></param>
/// <param name="mode">GC normalization mode</param>
static void NormalizeByGC(List<GenomicBin> bins, NexteraManifest manifest, CanvasGCNormalizationMode mode)
{
switch (mode)
{
case CanvasGCNormalizationMode.MedianByGC:
NormalizeByGC(bins, manifest: manifest);
break;
case CanvasGCNormalizationMode.LOESS:
var normalizer = new LoessGCNormalizer(bins, manifest, robustnessIter: 0,
countTransformer: x => (double)Math.Log(x),
invCountTransformer: x => (float)Math.Exp(x));
normalizer.Normalize();
break;
default:
throw new ApplicationException("Unsupported Canvas GC normalization mode: " + mode.ToString());
}
}
/// <summary>
/// Perform variance stabilization by GC bins.
/// </summary>
/// <param name="bins">Bins whose counts are to be normalized.</param>
static bool NormalizeVarianceByGC(List <SampleGenomicBin> bins, NexteraManifest manifest = null)
{
// DebugPrintCountsByGC(bins, "CountsByGCVariance-Before.txt");
// An array of lists. Each array element (0-100) will hold a list of counts whose bins have the same GC content.
List <float>[] countsByGC;
// Will hold all of the autosomal counts present in 'bins'
List <float> counts;
EnrichmentUtilities.GetCountsByGC(bins, manifest, out countsByGC, out counts);
// Estimate quartiles of all bins genomewide
var globalQuartiles = Utilities.Quartiles(counts);
// Will hold interquartile range (IQR) separately for each GC bin
List <float> localIQR = new List <float>(countsByGC.Length);
// Will hold quartiles separately for each GC bin
List <Tuple <float, float, float> > localQuartiles = new List <Tuple <float, float, float> >(countsByGC.Length);
// calculate interquartile range (IQR) for GC bins and populate localQuartiles list
for (int i = 0; i < countsByGC.Length; i++)
{
if (countsByGC[i].Count == 0)
{
localIQR.Add(-1f);
localQuartiles.Add(new Tuple <float, float, float>(-1f, -1f, -1f));
}
else if (countsByGC[i].Count >= defaultMinNumberOfBinsPerGC)
{
localQuartiles.Add(Utilities.Quartiles(countsByGC[i]));
localIQR.Add(localQuartiles[i].Item3 - localQuartiles[i].Item1);
}
else
{
List <Tuple <float, float> > weightedCounts = GetWeightedCounts(countsByGC, i);
double[] quartiles = Utilities.WeightedQuantiles(weightedCounts, new List <float>()
{
0.25f, 0.5f, 0.75f
});
localQuartiles.Add(new Tuple <float, float, float>((float)quartiles[0], (float)quartiles[1], (float)quartiles[2]));
localIQR.Add((float)(quartiles[2] - quartiles[0]));
}
}
// Identify if particular GC bins have IQR twice as large as IQR genomewide
float globalIQR = globalQuartiles.Item3 - globalQuartiles.Item1;
// Holder for GC bins with large IQR (compared to genomewide IQR)
int significantIQRcounter = 0;
for (int i = 10; i < 90; i++)
{
if (globalIQR < localIQR[i] * 2f)
{
significantIQRcounter++;
}
}
if (significantIQRcounter <= 0)
{
return(false);
}
// Divide each count by the median count of bins with the same GC content
foreach (SampleGenomicBin bin in bins)
{
var scaledLocalIqr = localIQR[bin.GenomicBin.GC] * 0.8f;
if (globalIQR >= scaledLocalIqr)
{
continue;
}
// ratio of GC bins and global IQRs
float iqrRatio = scaledLocalIqr / globalIQR;
var medianGCCount = localQuartiles[bin.GenomicBin.GC].Item2;
bin.Count = medianGCCount + (bin.Count - medianGCCount) / iqrRatio;
}
// DebugPrintCountsByGC(bins, "CountsByGCVariance-After.txt");
return(true);
}
/// <summary>
/// Remove bins with extreme GC content.
/// </summary>
/// <param name="bins">Genomic bins in from which we filter out GC content outliers.</param>
/// <param name="threshold">Minimum number of bins with the same GC content required to keep a bin.</param>
///
/// The rationale of this function is that a GC normalization is performed by computing the median count
/// for each possible GC value. If that count is small, then the corresponding normalization constant
/// is unstable and we shouldn't use these data.
static List<GenomicBin> RemoveBinsWithExtremeGC(List<GenomicBin> bins, int threshold, NexteraManifest manifest = null)
{
// Will hold outlier-removed bins.
List<GenomicBin> stripped = new List<GenomicBin>();
// used to count the number of bins with each possible GC content (0-100)
int[] counts = new int[EnrichmentUtilities.numberOfGCbins];
double totalCount = 0;
foreach (GenomicBin bin in manifest == null ? bins : EnrichmentUtilities.GetOnTargetBins(bins, manifest))
{
// We only count autosomal bins because these are the ones we computed normalization factor upon.
if (!GenomeMetadata.SequenceMetadata.IsAutosome(bin.Chromosome))
continue;
counts[bin.GC]++;
totalCount++;
}
int averageCountPerGC = Math.Max(minNumberOfBinsPerGCForWeightedMedian, (int)(totalCount / counts.Length));
threshold = Math.Min(threshold, averageCountPerGC);
foreach (GenomicBin bin in bins)
{
// Remove outlier (not a lot of bins with the same GC content)
if (counts[bin.GC] < threshold)
continue;
stripped.Add(bin);
}
return stripped;
}
/// <summary>
/// Computes fragment-based GC normalization correction factor
/// </summary>
/// <returns>An array of observed vs expected GC counts.</returns>
static float[] ComputeObservedVsExpectedGC(Dictionary<string, HitArray> observedAlignments,
Dictionary<string, byte[]> readGCContent, NexteraManifest manifest,
bool debugGC, string outFile)
{
Dictionary<string, List<NexteraManifest.ManifestRegion>> regionsByChrom = null;
if (manifest != null)
{
regionsByChrom = manifest.GetManifestRegionsByChromosome();
}
long[] expectedReadCountsByGC = new long[numberOfGCbins];
long[] observedReadCountsByGC = new long[numberOfGCbins];
foreach (KeyValuePair<string, byte[]> chromosomeReadGCContent in readGCContent)
{
string chr = chromosomeReadGCContent.Key;
if (!observedAlignments.ContainsKey(chr)) { continue; }
if (manifest == null) // look at the entire genome
{
for (int i = 0; i < chromosomeReadGCContent.Value.Length; i++)
{
expectedReadCountsByGC[chromosomeReadGCContent.Value[i]]++;
observedReadCountsByGC[chromosomeReadGCContent.Value[i]] += observedAlignments[chr].Data[i];
}
}
else // look at only the targeted regions
{
if (!regionsByChrom.ContainsKey(chr)) { continue; }
int i = -1;
foreach (var region in regionsByChrom[chr])
{
if (i < region.Start) // avoid overlapping targeted regions
{
i = region.Start - 1; // i is 0-based; manifest coordinates are 1-based.
}
for (; i < chromosomeReadGCContent.Value.Length && i < region.End; i++)
{
expectedReadCountsByGC[chromosomeReadGCContent.Value[i]]++;
observedReadCountsByGC[chromosomeReadGCContent.Value[i]] += observedAlignments[chr].Data[i];
}
}
}
}
// calculate ratio of observed to expected read counts for each read GC bin
float[] observedVsExpectedGC = new float[numberOfGCbins];
for (int i = 0; i < numberOfGCbins; i++)
observedVsExpectedGC[i] = 1;
long sumObserved = 0;
long sumExpected = 0;
foreach (long gcContent in observedReadCountsByGC)
sumObserved += gcContent;
foreach (long gcContent in expectedReadCountsByGC)
sumExpected += gcContent;
for (int binIndex = 0; binIndex < numberOfGCbins; binIndex++)
{
if (expectedReadCountsByGC[binIndex] == 0)
expectedReadCountsByGC[binIndex] = 1;
if (observedReadCountsByGC[binIndex] == 0)
observedReadCountsByGC[binIndex] = 1;
observedVsExpectedGC[binIndex] = ((float)observedReadCountsByGC[binIndex] / (float)expectedReadCountsByGC[binIndex]) * ((float)sumExpected / (float)sumObserved);
}
if (debugGC)
{
using (GzipWriter writer = new GzipWriter(outFile + ".gcstat"))
{
for (int binIndex = 0; binIndex < numberOfGCbins; binIndex++)
{
writer.WriteLine(string.Format("{0}\t{1}\t{2}", expectedReadCountsByGC[binIndex], observedReadCountsByGC[binIndex], observedVsExpectedGC[binIndex]));
}
}
}
return observedVsExpectedGC;
}
