/// <summary>
/// Initializes a new instance of the <see cref="Bio.Variant.ContinuousFrequencySNPGenotype"/> class.
/// </summary>
/// <param name="res">Res.</param>
/// <param name="pileup">Pileup.</param>
public ContinuousFrequencySNPGenotype(GenotypeCallResult res, PileUp pileup = null)
{
ResultType = res;
if (pileup != null)
{
pOriginalPosition = pileup.Position;
InsertionOffset = pileup.InsertionOffSet;
}
}
private static ContinuousFrequencySNPGenotype callGenotype(PileUp pu)
{
//If it looks like a deletion, skip it
if (pileupHasTooManyIndels(pu))
{
return(new ContinuousFrequencySNPGenotype(GenotypeCallResult.TooManyGaps, pu));
}
// Otherwise, drop gaps, ambiguous bases and low scoring reads.
var filteredBases = pu.Bases.Where(z => z.Base != BaseAndQuality.N_BASE_INDEX &&
z.Base != BaseAndQuality.GAP_BASE_INDEX && z.PhredScore > 17).ToArray();
if (filteredBases.Length == 0)
{
return(new ContinuousFrequencySNPGenotype(GenotypeCallResult.NoData));
}
// initialize the continuous frequency based on counts of bases.
var base_pair_counts = new int[BasePairFrequencies.NUM_BASES];
foreach (var bp in filteredBases)
{
base_pair_counts [bp.Base]++;
}
var freqs = base_pair_counts.Select(x => x / (double)filteredBases.Length).ToArray();
var theta = new BasePairFrequencies(freqs);
//if only one base has data or if
//if we are not doing EM optimization, we are done.
if (base_pair_counts.Count(x => x > 0) > 1 && DO_EM_ESTIMATION)
{
//first make an NumReads * Num_Bases Matrix
double[][] conditionalProbs = new double[filteredBases.Length][];
for (int i = 0; i < filteredBases.Length; i++)
{
conditionalProbs [i] = new double[BasePairFrequencies.NUM_BASES];
}
double likDif = double.MaxValue;
double last_lik = double.MinValue;
while (likDif > 1e-3)
{
double lik = 0;
for (int i = 0; i < conditionalProbs.Length; i++)
{
lik += updateConditionalProbabilities(theta, conditionalProbs [i], filteredBases [i]);
}
likDif = lik - last_lik;
last_lik = lik;
//now update thetas by summing the conditional probability of each value
Array.Clear(theta.Frequencies, 0, BasePairFrequencies.NUM_BASES);
for (int i = 0; i < conditionalProbs.Length; i++)
{
var cur = conditionalProbs [i];
for (int j = 0; j < cur.Length; j++)
{
theta.Frequencies [j] += cur [j];
}
}
for (int j = 0; j < theta.Frequencies.Length; j++)
{
theta.Frequencies [j] /= (double)conditionalProbs.Length;
}
}
}
return(new ContinuousFrequencySNPGenotype(theta, base_pair_counts, pu));
}
private static bool pileupHasTooManyIndels(PileUp up)
{
var freq = up.Bases.Count(z => z.Base == BaseAndQuality.GAP_BASE_INDEX) / (double)up.Bases.Count;
return(freq >= MIN_DELETION_PERCENTAGE_NEEDED_TO_CALL);
}
public ContinuousFrequencySNPGenotype(BasePairFrequencies frequencies, int[] originalReadCounts, PileUp pileup = null) :
this(GenotypeCallResult.GenotypeCalled, pileup)
{
Frequencies = frequencies;
OriginalBasePairCounts = originalReadCounts;
var freq = frequencies.Frequencies;
this.indexMax = 0;
var max = freq [0];
for (int i = 1; i < freq.Length; i++)
{
if (freq [i] > max)
{
max = freq [i];
indexMax = i;
}
}
}
