public void CallThroughAnEmptyNbhd()
{
var originalVcfVariant = TestHelper.CreateDummyAllele("chr1", 123, "A", "T", 1000, 156);
var originalVcfVariant2 = TestHelper.CreateDummyAllele("chr1", 124, "A", "T", 1000, 156);
var vs1 = new VariantSite(originalVcfVariant);
var vs2 = new VariantSite(originalVcfVariant2);
var caller = new VariantCaller(new VariantCallingParameters(), new BamFilterParameters());
//since there is an alt at position 124 ( a call of 156 alt / 1000 total, that means 844 original ref calls.
//Of which we said, 100 will get sucked up. So that leaves 744 / 1000 calls for a reference.
//So, we can still make a confident ref call. (we will call it 0/., since we know its not a homozygous ref)
var nbhd = new VcfNeighborhood(new VariantCallingParameters(), 0, "chr1", vs1, vs2, "");
nbhd.SetRangeOfInterest();
caller.CallMNVs(nbhd);
caller.CallRefs(nbhd);
var acceptedMNVs = nbhd.CalledVariants;
var acceptedRefs = nbhd.CalledRefs;
Assert.Equal(0, acceptedMNVs.Count);
Assert.Equal(2, acceptedRefs.Count);
Assert.Equal(Genotype.RefAndNoCall, acceptedRefs[123].Genotype);
Assert.Equal(Genotype.RefAndNoCall, acceptedRefs[124].Genotype);
Assert.Equal(123, acceptedRefs[123].ReferencePosition);
Assert.Equal(124, acceptedRefs[124].ReferencePosition);
}
public void CheckAddingFilters()
{
var originalVcfVariant = TestHelper.CreateDummyAllele("chr1", 123, "A", "T", 1000, 156);
var originalVcfVariant2 = TestHelper.CreateDummyAllele("chr1", 124, "A", "T", 1000, 156);
var vs1 = new VariantSite(originalVcfVariant);
var vs2 = new VariantSite(originalVcfVariant2);
var variantCallingParameters = new VariantCallingParameters();
//Set up filters so calls are sure to trigger them.
variantCallingParameters.LowDepthFilter = 2000;
variantCallingParameters.MinimumFrequencyFilter = 0.80F;
variantCallingParameters.MinimumVariantQScoreFilter = 300;
var caller = new VariantCaller(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 = 500
});
nbhd.UsedRefCountsLookup = new Dictionary <int, SuckedUpRefRecord>()
{
};
caller.CallMNVs(nbhd);
caller.CallRefs(nbhd);
var acceptedMNVs = nbhd.CalledVariants;
var acceptedRefs = nbhd.CalledRefs;
Assert.Equal(1, acceptedMNVs.Count);
Assert.Equal(1, acceptedMNVs[123].Count);
Assert.True(acceptedMNVs[123][0].Filters.Contains(FilterType.LowDepth));
Assert.True(acceptedMNVs[123][0].Filters.Contains(FilterType.LowVariantFrequency));
Assert.True(acceptedMNVs[123][0].Filters.Contains(FilterType.LowVariantQscore));
Assert.Equal(2, acceptedRefs.Count);
Assert.True(acceptedRefs[123].Filters.Contains(FilterType.LowDepth));
Assert.True(acceptedRefs[123].Filters.Contains(FilterType.LowVariantQscore));
//note reference calls dont win the "LowVariantFrequency" flag.
}
public void VarCallsBecomeRefsAndNulls()
{
var originalVcfVariant = TestHelper.CreateDummyAllele("chr1", 123, "A", "T", 1000, 156);
var originalVcfVariant2 = TestHelper.CreateDummyAllele("chr1", 124, "A", "T", 1000, 156);
var vs1 = new VariantSite(originalVcfVariant);
var vs2 = new VariantSite(originalVcfVariant2);
var vcParams = new VariantCallingParameters();
vcParams.Validate();
var caller = new VariantCaller(vcParams, new BamFilterParameters());
//since there is an alt at position 124 ( a call of 156 alt / 1000 total, that means 844 original ref calls.
//Of which we said, 100 will get sucked up. So that leaves 744 / 1000 calls for a reference.
//So, we can still make a confident ref call.
var nbhd = new VcfNeighborhood(vcParams, 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 = 500
});
nbhd.UsedRefCountsLookup = new Dictionary <int, SuckedUpRefRecord>()
{
};
caller.CallMNVs(nbhd);
caller.CallRefs(nbhd);
var acceptedMNVs = nbhd.CalledVariants;
var acceptedRefs = nbhd.CalledRefs;
Assert.Equal(1, acceptedMNVs.Count);
Assert.Equal(1, acceptedMNVs[123].Count);
Assert.Equal(2, acceptedRefs.Count);
var vcfVariant2asRef = new VcfVariant()
{
ReferenceName = "chr1",
ReferencePosition = 124,
ReferenceAllele = "A",
VariantAlleles = new[] { "." },
Genotypes = new List <Dictionary <string, string> >()
{
new Dictionary <string, string>()
{
{ "GT", "0/." }, { "DP", "1000" }, { "AD", "844" }
}
},
};
VcfMergerTests.CheckVariantsMatch(originalVcfVariant, acceptedMNVs[123][0]);
VcfMergerTests.CheckVariantsMatch(vcfVariant2asRef, acceptedRefs[124]);
// If one has been sucked up and there are refs remaining, we should output it as a ref.
var suckedUpRefRecord100 = new SuckedUpRefRecord()
{
Counts = 100, AlleleThatClaimedIt = new CalledAllele()
};
nbhd.UsedRefCountsLookup = new Dictionary <int, SuckedUpRefRecord>()
{
{ 124, suckedUpRefRecord100 }
};
caller.CallMNVs(nbhd);
caller.CallRefs(nbhd);
acceptedMNVs = nbhd.CalledVariants;
acceptedRefs = nbhd.CalledRefs;
Assert.Equal(1, acceptedMNVs.Count);
Assert.Equal(1, acceptedMNVs[123].Count);
Assert.Equal(2, acceptedRefs.Count);
vcfVariant2asRef = new VcfVariant()
{
ReferenceName = "chr1",
ReferencePosition = 124,
ReferenceAllele = "A",
VariantAlleles = new[] { "." },
Genotypes = new List <Dictionary <string, string> >()
{
new Dictionary <string, string>()
{
{ "GT", "0/." }, { "DP", "1000" }, { "AD", "744" }
}
},
};
VcfMergerTests.CheckVariantsMatch(originalVcfVariant, acceptedMNVs[123][0]);
VcfMergerTests.CheckVariantsMatch(vcfVariant2asRef, acceptedRefs[124]);
// If one has been sucked up all the way
// we should output it as a null.
var suckedUpRefRecord1000 = new SuckedUpRefRecord()
{
Counts = 1000, AlleleThatClaimedIt = new CalledAllele()
};
nbhd.UsedRefCountsLookup = new Dictionary <int, SuckedUpRefRecord>()
{
{ 124, suckedUpRefRecord1000 }
};
caller.CallMNVs(nbhd);
caller.CallRefs(nbhd);
acceptedMNVs = nbhd.CalledVariants;
acceptedRefs = nbhd.CalledRefs;
Assert.Equal(1, acceptedMNVs.Count);
Assert.Equal(1, acceptedMNVs[123].Count);
Assert.Equal(2, acceptedRefs.Count);
var vcfVariant2asNull = new VcfVariant()
{
ReferenceName = "chr1",
ReferencePosition = 124,
ReferenceAllele = "A",
VariantAlleles = new[] { "." },
Genotypes = new List <Dictionary <string, string> >()
{
new Dictionary <string, string>()
{
{ "GT", "./." }, { "DP", "1000" }, { "AD", "0" }
}
},
};
VcfMergerTests.CheckVariantsMatch(originalVcfVariant, acceptedMNVs[123][0]);
VcfMergerTests.CheckVariantsMatch(vcfVariant2asNull, acceptedRefs[124]);
}
public void CallAVariantInANewLocation()
{
//set up the original variants
var originalVcfVariant1 = TestHelper.CreateDummyAllele("chr1", 123, "A", "T", 1000, 156);
var originalVcfVariant2 = TestHelper.CreateDummyAllele("chr1", 124, "A", "T", 1000, 156);
var originalVcfVariant3 = TestHelper.CreateDummyAllele("chr1", 234, "A", "T", 1000, 156);
var originalVcfVariant4 = TestHelper.CreateDummyAllele("chr1", 234, "A", "T", 1000, 156);
var vs1 = new VariantSite(originalVcfVariant1);
var vs2 = new VariantSite(originalVcfVariant2);
var vs3 = new VariantSite(originalVcfVariant3);
var vs4 = new VariantSite(originalVcfVariant4);
var vcParams = new VariantCallingParameters();
vcParams.Validate();
var caller = new VariantCaller(vcParams, new BamFilterParameters());
var nbhd = new VcfNeighborhood(vcParams, 0, "chr1", vs1, vs2, "");
nbhd.AddVariantSite(vs3, "RRRRR"); //note, we do not add vs4, that is not going to get used for phasing. Sps it is a variant that failed filters.
nbhd.SetRangeOfInterest();
//now stage one candidate MNV:
var newMNV = new CalledAllele(AlleleCategory.Snv)
{
Chromosome = "chr1",
ReferencePosition = 129,
ReferenceAllele = "A",
AlternateAllele = "TT",
VariantQscore = 100,
TotalCoverage = 1000,
AlleleSupport = 500
};
nbhd.AddAcceptedPhasedVariant(newMNV);
var suckedUpRefRecord1000 = new SuckedUpRefRecord()
{
Counts = 1000, AlleleThatClaimedIt = new CalledAllele()
};
nbhd.UsedRefCountsLookup = new Dictionary <int, SuckedUpRefRecord>()
{
{ 124, suckedUpRefRecord1000 }
};
caller.CallMNVs(nbhd);
caller.CallRefs(nbhd);
var acceptedMNVs = nbhd.CalledVariants;
var acceptedRefs = nbhd.CalledRefs;
var vcfVariant0asRef = new VcfVariant()
{
ReferenceName = "chr1",
ReferencePosition = 123,
ReferenceAllele = "A",
VariantAlleles = new[] { "." },
Genotypes = new List <Dictionary <string, string> >()
{
new Dictionary <string, string>()
{
{ "GT", "0/." }
}
},
};
var vcfVariant3asRef = new VcfVariant()
{
ReferenceName = "chr1",
ReferencePosition = 234,
ReferenceAllele = "A",
VariantAlleles = new[] { "." },
Genotypes = new List <Dictionary <string, string> >()
{
new Dictionary <string, string>()
{
{ "GT", "0/." }
}
},
};
var vcfVariant2asNull = new VcfVariant()
{
ReferenceName = "chr1",
ReferencePosition = 124,
ReferenceAllele = "A",
VariantAlleles = new[] { "." },
Genotypes = new List <Dictionary <string, string> >()
{
new Dictionary <string, string>()
{
{ "GT", "./." }
}
},
};
Assert.Equal(1, acceptedMNVs.Count);
Assert.Equal(1, acceptedMNVs[129].Count);
Assert.Equal(3, acceptedRefs.Count);
VcfMergerTests.CheckVariantsMatch(vcfVariant0asRef, acceptedRefs[123]);
VcfMergerTests.CheckVariantsMatch(vcfVariant2asNull, acceptedRefs[124]);
VcfMergerTests.CheckVariantsMatch(newMNV, acceptedMNVs[129][0]);
VcfMergerTests.CheckVariantsMatch(vcfVariant3asRef, acceptedRefs[234]);
}
//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)
}