/// <summary>
/// Setup: Model various copy ploidies.
/// </summary>
public void InitializePloidies()
{
Console.WriteLine("{0} Initialize ploidy models...", DateTime.Now);
this.AllPloidies = new List<SegmentPloidy>();
double diploidPredictedMAF = CanvasCommon.Utilities.EstimateDiploidMAF(2, this.MeanCoverage);
for (int copyNumber = 0; copyNumber <= MaximumCopyNumber; copyNumber++)
{
for (int majorCount = copyNumber; majorCount * 2 >= copyNumber; majorCount--)
{
SegmentPloidy ploidy = new SegmentPloidy();
ploidy.CopyNumber = copyNumber;
ploidy.MajorChromosomeCount = majorCount;
ploidy.ID = AllPloidies.Count;
AllPloidies.Add(ploidy);
if (copyNumber == 0)
{
ploidy.MinorAlleleFrequency = 0.01; // should reflect sequencing error rate
continue;
}
float VF = majorCount / (float)copyNumber;
ploidy.MinorAlleleFrequency = (VF < 0.5 ? VF : 1 - VF);
if (majorCount * 2 == copyNumber)
{
ploidy.MinorAlleleFrequency = CanvasCommon.Utilities.EstimateDiploidMAF(copyNumber, this.MeanCoverage);
}
}
}
Console.WriteLine("{0} Ploidy models prepared.", DateTime.Now);
}
/// <summary>
/// Initialize model points by subsampling from existing segment Coverage and MAF values.
/// Use distanceThreshold to ensure that both large and small cluster components get subsampled
/// </summary>
protected List<ModelPoint> InitializeModelPoints(List<SegmentInfo> segments, int numClusters, double distanceThreshold)
{
List<ModelPoint> modelPoints = new List<ModelPoint>();
List<SegmentInfo> usableSegments = new List<SegmentInfo>();
List<SegmentInfo> usedSegments = new List<SegmentInfo>();
foreach (SegmentInfo segment in segments) {
if (segment.Cluster != -1 && segment.MAF >= 0)
usableSegments.Add(segment);
}
Random rnd = new Random();
int lastIndex = rnd.Next(1, usableSegments.Count);
usedSegments.Add(usableSegments[lastIndex]);
int counter = 1;
double attempts = 0;
while (counter < numClusters)
{
int newIndex = rnd.Next(1, usableSegments.Count);
attempts += 1.0;
double distance = GetModelDistance(usableSegments[lastIndex].Coverage, usableSegments[newIndex].Coverage, usableSegments[lastIndex].MAF, usableSegments[newIndex].MAF);
if (distance > distanceThreshold || attempts/usableSegments.Count > 0.3) // escape outlier minima
{
usedSegments.Add(usableSegments[newIndex]);
counter++;
lastIndex = newIndex;
attempts = 0;
}
}
// Initialize model points with coverage and MAF values from subsampled segments
for (int i = 0; i < numClusters; i++)
{
ModelPoint point = new ModelPoint();
point.Coverage = usedSegments[i].Coverage;
point.MAF = usedSegments[i].MAF;
SegmentPloidy ploidy = new SegmentPloidy();
ploidy.CopyNumber = 2;
ploidy.MajorChromosomeCount = 1;
point.Ploidy = ploidy;
point.Cluster = i+1;
modelPoints.Add(point);
}
return modelPoints;
}
