/// <summary>
/// Step 1: Load the normal het SNVs of interest.
/// </summary>
protected void LoadVariants(string vcfPath)
{
Console.WriteLine("{0} Loading variants of interest from {1}", DateTime.Now, vcfPath);
this.Variants = new List<VcfVariant>();
int overallCount = 0;
int countThisChromosome = 0;
using (VcfReader reader = new VcfReader(vcfPath, requireGenotypes: false))
{
VcfVariant variant = new VcfVariant();
while (true)
{
bool result = reader.GetNextVariant(out variant);
if (!result) break;
overallCount++;
if (variant.ReferenceName != this.Chromosome)
{
// Shortcut: If we've seen records for the desired chromosome, then as soon as we hit another chromosome,
// we can abort:
if (countThisChromosome > 0) break;
continue;
}
countThisChromosome++;
// Single-allele SNVs only:
if (variant.VariantAlleles.Length != 1 || variant.VariantAlleles[0].Length != 1 || variant.ReferenceAllele.Length != 1) continue;
// PF variants only:
if ((variant.GenotypeColumns != null && variant.GenotypeColumns.Any()) && variant.Filters != "PASS") continue; // FILTER may not say PASS for a dbSNP VCF file
if (variant.GenotypeColumns != null && variant.GenotypeColumns.Any()) // not available if we use a dbSNP VCF file
{
if (!variant.GenotypeColumns[0].ContainsKey("GT")) continue; // no genotype - we don't know if it's a het SNV.
string genotype = variant.GenotypeColumns[0]["GT"];
if (genotype != "0/1" && genotype != "1/0") continue;
// Also require they have a high enough quality score:
if (variant.GenotypeColumns[0].ContainsKey("GQX")) // Note: Allow no GQX field, in case we want to use another caller (e.g. Pisces) and not crash
{
float GQX = float.Parse(variant.GenotypeColumns[0]["GQX"]);
if (GQX < 30) continue;
}
}
// Note: Let's NOT require the variant be in dbSNP. Maybe we didn't do annotation, either because
// we chose not to or because we're on a reference without annotation available.
//if (variant.Identifier == ".") continue;
// Remember all the variants that pass all our tests:
this.Variants.Add(variant);
variant = new VcfVariant();
}
}
Console.WriteLine("Retained {0} variants, out of {1} records for {2}", this.Variants.Count, countThisChromosome, this.Chromosome);
}
protected IEnumerable<CNVCall> GetCnvCallsFromVcf(string vcfPath)
{
using (VcfReader reader = new VcfReader(vcfPath, false))
{
foreach (VcfVariant variant in reader.GetVariants())
{
int end;
int CN = GetCopyNumber(variant, out end);
yield return new CNVCall(variant.ReferenceName, variant.ReferencePosition, end, CN);
}
}
}
protected IEnumerable<CNVCall> GetCnvCallsFromVcf(string vcfPath, bool includePassingOnly)
{
using (VcfReader reader = new VcfReader(vcfPath, false))
{
foreach (VcfVariant variant in reader.GetVariants())
{
int end;
int CN = GetCopyNumber(variant, out end);
if (includePassingOnly && variant.Filters != "PASS") continue;
yield return new CNVCall(variant.ReferenceName, variant.ReferencePosition, end, CN);
}
}
}
/// <summary>
/// Load the somatic SNV output (from Strelka), and use it to dervie an estimate of overall purity.
/// Reference: https://ukch-confluence.illumina.com/display/collab/Estimate+purity+of+tumour+samples+using+somatic+SNVs
/// </summary>
protected double EstimatePurityFromSomaticSNVs()
{
Dictionary<string, List<CanvasSegment>> segmentsByChromosome = CanvasSegment.GetSegmentsByChromosome(Segments);
int recordCount = 0;
List<float> variantFrequencies = new List<float>();
using (VcfReader reader = new VcfReader(this.SomaticVCFPath))
{
foreach (VcfVariant variant in reader.GetVariants())
{
recordCount++;
// Skip non-PF variants:
if (variant.Filters != "PASS") continue;
// Skip everything but SNVs:
if (variant.ReferenceAllele.Length > 1 || variant.VariantAlleles.Length != 1 || variant.VariantAlleles[0].Length != 1
|| variant.VariantAlleles[0] == ".")
{
continue;
}
string refTagName = string.Format("{0}U", variant.ReferenceAllele[0]);
string altTagName = string.Format("{0}U", variant.VariantAlleles[0][0]);
string[] counts = variant.Genotypes.Last()[refTagName].Split(',');
int refCount = 0;
foreach (string bit in counts) refCount += int.Parse(bit);
int altCount = 0;
counts = variant.Genotypes.Last()[altTagName].Split(',');
foreach (string bit in counts) altCount += int.Parse(bit);
float VF = altCount / (float)(altCount + refCount);
if (VF >= 0.5) continue;
variantFrequencies.Add(VF);
}
}
Console.WriteLine(">>>Loaded {0} somatic variants; saved {1} somatic SNV frequencies", recordCount, variantFrequencies.Count);
// Don't bother providing an estimate if we have very few events:
if (variantFrequencies.Count < 100)
{
return double.NaN;
}
double mean = variantFrequencies.Average();
double estimatedPurity = Math.Min(1, mean * 2);
Console.WriteLine(">>>Estimated tumor purity of {0} from somatic SNV calls", estimatedPurity);
return estimatedPurity;
}
public VcfWriter(string filePath, VcfReader reader)
{
HeaderLines = reader.HeaderLines.ToList();
Samples = reader.Samples.ToList();
Open(filePath);
}
/// <summary>
/// Load a list of all variants in a file. This is memory-intensive; don't do this for whole-genome vcf files!
/// </summary>
public static List<VcfVariant> GetAllVariantsInFile(string vcfPath)
{
List<VcfVariant> allVariants = new List<VcfVariant>();
using (VcfReader reader = new VcfReader(vcfPath))
{
foreach (VcfVariant variant in reader.GetVariants())
{
allVariants.Add(variant);
}
}
return allVariants;
}