private void ProcessNeighborhood(VcfNeighborhood neighborhood)
{
Logger.WriteToLog("Processing Neighborhood {0}.", neighborhood.Id);
try
{
var clusterer = _factory.CreateNeighborhoodClusterer();
var veadGroupSource = _factory.CreateVeadGroupSource();
var collapsedReads = veadGroupSource.GetVeadGroups(neighborhood);
//(1) Get CLUSTERS
var clusters = clusterer.ClusterVeadGroups(collapsedReads.ToList(), neighborhood.Id);
//clean out vg, we dont need them any more
veadGroupSource = null;
collapsedReads = null;
bool crushNbhdVariantsToSamePositon = !_factory.Options.VcfWritingParams.AllowMultipleVcfLinesPerLoci;
//(2) Turn clusters into MNV candidates
neighborhood.CreateMnvsFromClusters(clusters.Clusters,
_factory.Options.BamFilterParams.MinimumBaseCallQuality, _factory.Options.VariantCallingParams.MaximumVariantQScore,
crushNbhdVariantsToSamePositon);
neighborhood.SetGenotypesAndPruneExcessAlleles();
// (3) Variant call the candidates
var variantCaller = _factory.CreateVariantCaller();
variantCaller.CallMNVs(neighborhood);
variantCaller.CallRefs(neighborhood);
//wait untill vcf is ready to write...
}
catch (Exception ex)
{
Logger.WriteToLog("Error processing neighborhood {0}", neighborhood.Id);
Logger.WriteExceptionToLog(ex);
}
}
public void AddMnvsFromClusters()
{
//TODO even with mock cluster this takes too much setting up.
var nbhd = new VcfNeighborhood(new VariantCallingParameters(), 0, "chr1", new VariantSite(120), new VariantSite(121), "T");
var vead = PhasedVariantTestUtilities.CreateVeadFromStringArray("r1", new[, ] {
{ "C", "A" }, { "G", "A" }, { "T", "A" }
});
var vead2 = PhasedVariantTestUtilities.CreateVeadFromStringArray("r2", new[, ] {
{ "C", "A" }, { "G", "A" }, { "T", "A" }
});
var vead3 = PhasedVariantTestUtilities.CreateVeadFromStringArray("r3", new[, ] {
{ "C", "A" }, { "G", "A" }, { "T", "A" }
});
var veads = new List <Vead> {
vead, vead2, vead3
};
vead.SiteResults[0].VcfReferencePosition = 1;
vead.SiteResults[1].VcfReferencePosition = 2;
vead.SiteResults[2].VcfReferencePosition = 3;
vead2.SiteResults[0].VcfReferencePosition = 1;
vead2.SiteResults[1].VcfReferencePosition = 2;
vead2.SiteResults[2].VcfReferencePosition = 3;
vead3.SiteResults[0].VcfReferencePosition = 1;
vead3.SiteResults[1].VcfReferencePosition = 2;
vead3.SiteResults[2].VcfReferencePosition = 3;
nbhd.ReferenceSequence = "CGT";
var mockCluster = new Mock <ICluster>();
mockCluster.Setup(c => c.CountsAtSites).Returns(new[] { 10, 3, 5 });
var consensus = PhasedVariantTestUtilities.CreateVeadGroup(veads);
mockCluster.Setup(c => c.GetConsensusSites()).Returns(consensus.SiteResults);
mockCluster.Setup(c => c.GetVeadGroups()).Returns(new List <VeadGroup>()
{
consensus
});
nbhd.CreateMnvsFromClusters(new List <ICluster>()
{
mockCluster.Object
}, 20, 100);
var allele = nbhd.CandidateVariants.First();
Assert.Equal(6, allele.TotalCoverage);
Assert.Equal(6, allele.AlleleSupport);
Assert.Equal("CGT", allele.ReferenceAllele);
Assert.Equal("AAA", allele.AlternateAllele);
int[] depths = new int[0];
int[] nocalls = new int[0];
nbhd.DepthAtSites(new List <ICluster>()
{
mockCluster.Object
}, out depths, out nocalls);
Assert.Equal(3, depths.Length);
Assert.Equal(3, depths[0]);
Assert.Equal(3, depths[1]);
Assert.Equal(3, depths[2]);
}
//this unit test was made after we found bug ScyllaLoosingRefCalls_PICS-723.
//We had a 1/. GT reported when it should be 1/0.
//The reason for this is that all the refs (the "0"s) got incorrectly sucked up.
//Ie, MNV ACG-> AG claimed 50 refs, so we (incorrectly) subtracted 50 refs from it.
//The bug is that the ref counts got subtractedfrom the exact same mnv that claimed them.
// This should never happen, and was not the intent of the alg.
//
//The affected mehtod is: CreateMnvsFromClusters in VcfNbhd
public void CreateMnvsFromClusters_TakeUpRefCount()
{
var originalVcfVariant1 = TestHelper.CreateDummyAllele("chr1", 123, "ACG", "AT", 1000, 156);
var originalVcfVariant2 = TestHelper.CreateDummyAllele("chr1", 123, "A", "TTTTTT", 1000, 200);
var originalVcfVariant3 = TestHelper.CreateDummyAllele("chr1", 123, "AC", "TT", 1000, 100);
var vs1 = new VariantSite(originalVcfVariant1);
var vs2 = new VariantSite(originalVcfVariant2);
var caller = new VariantCaller(new VariantCallingParameters(), new BamFilterParameters());
var nbhd = new VcfNeighborhood(new VariantCallingParameters(), 0, "chr1", vs1, vs2, "");
nbhd.SetRangeOfInterest();
nbhd.AddAcceptedPhasedVariant(
new CalledAllele(AlleleCategory.Snv)
{
Chromosome = "chr1",
ReferencePosition = 123,
ReferenceAllele = "A",
AlternateAllele = "T",
VariantQscore = 100,
TotalCoverage = 1000,
AlleleSupport = 200
});
nbhd.AddAcceptedPhasedVariant(
new CalledAllele(AlleleCategory.Mnv)
{
Chromosome = "chr1",
ReferencePosition = 123,
ReferenceAllele = "ACG",
AlternateAllele = "AT",
VariantQscore = 100,
TotalCoverage = 1000,
AlleleSupport = 300
});
nbhd.AddAcceptedPhasedVariant(
new CalledAllele(AlleleCategory.Insertion)
{
Chromosome = "chr1",
ReferencePosition = 123,
ReferenceAllele = "A",
AlternateAllele = "AAAAA",
VariantQscore = 100,
TotalCoverage = 1000,
AlleleSupport = 250
});
//default behavior, nothing gets sucked up
nbhd.UsedRefCountsLookup = new Dictionary <int, SuckedUpRefRecord>()
{
};
vs1.VcfReferencePosition = 123;
var vead = new Vead("dummy", new VariantSite[] { vs1 });
var vg = new VeadGroup(vead);
var fakeCluster = new Cluster("test", new List <VeadGroup>()
{
vg
});
fakeCluster.ResetConsensus();
nbhd.CreateMnvsFromClusters(new List <Cluster> {
fakeCluster
},
20, 100);
caller.CallMNVs(nbhd);
caller.CallRefs(nbhd);
var acceptedMNVs = nbhd.CalledVariants;
var acceptedRefs = nbhd.CalledRefs;
Assert.Equal(2, acceptedMNVs.Count);
Assert.Equal(3, acceptedMNVs[123].Count);
Assert.Equal(1, acceptedRefs.Count);
//check the ref counts on all the MNVs. Nothing should be sucked up.
Assert.Equal(1000 - 200, acceptedMNVs[123][0].ReferenceSupport); //total depth - allele suport. overly simple for now)
Assert.Equal(1000 - 300, acceptedMNVs[123][1].ReferenceSupport); //total depth - allele suport. overly simple for now)
Assert.Equal(1000 - 250, acceptedMNVs[123][2].ReferenceSupport); //total depth - allele suport. overly simple for now)
// now variant 0 will suck up 100 ref calls:
var suckedUpRefRecord100 = new SuckedUpRefRecord()
{
Counts = 100, AlleleThatClaimedIt = nbhd.CandidateVariants[0]
};
nbhd.UsedRefCountsLookup = new Dictionary <int, SuckedUpRefRecord>()
{
{ 123, suckedUpRefRecord100 }
};
nbhd.CreateMnvsFromClusters(new List <Cluster> {
fakeCluster
},
20, 100);
caller.CallMNVs(nbhd);
caller.CallRefs(nbhd);
acceptedMNVs = nbhd.CalledVariants;
acceptedRefs = nbhd.CalledRefs;
//check the ref counts on all the MNVs. refs should only be taken up by the first one
Assert.Equal(1000 - 200, acceptedMNVs[123][0].ReferenceSupport); //total depth - allele suport. overly simple for now)
//old result - has bug
//Assert.Equal(1000 - 300, acceptedMNVs[123][1].ReferenceSupport); //total depth - allele suport - sucked up ref)
//Assert.Equal(1000 - 250, acceptedMNVs[123][2].ReferenceSupport); //total depth - allele suport - sucked up ref)
//new result, fixed
Assert.Equal(1000 - 300 - 100, acceptedMNVs[123][1].ReferenceSupport); //total depth - allele suport - sucked up ref)
Assert.Equal(1000 - 250 - 100, acceptedMNVs[123][2].ReferenceSupport); //total depth - allele suport - sucked up ref)
}