|
public static GetOnTargetBins ( IEnumerable |
||
| bins | IEnumerable |
|
| manifest | NexteraManifest | |
| return | IEnumerable |
|
/// <summary>
/// Assumes the bins are sorted by genomic coordinates
/// </summary>
/// <param name="bins">Bins whose counts are to be normalized</param>
/// <param name="countsByGC">An array of lists. Each array element (0-100) will hold a list of counts whose bins have the same GC content.</param>
/// <param name="counts">Will hold all of the autosomal counts present in 'bins'</param>
public static void GetCountsByGC(List <SampleGenomicBin> bins, NexteraManifest manifest, out List <float>[] countsByGC, out List <float> counts)
{
countsByGC = new List <float> [numberOfGCbins];
counts = new List <float>(bins.Count);
// Initialize the lists
for (int i = 0; i < countsByGC.Length; i++)
{
countsByGC[i] = new List <float>();
}
foreach (SampleGenomicBin bin in manifest == null ? bins : EnrichmentUtilities.GetOnTargetBins(bins, manifest))
{
if (!GenomeMetadata.SequenceMetadata.IsAutosome(bin.GenomicBin.Chromosome))
{
continue;
}
// Put the observed count in the GC-appropriate list.
countsByGC[bin.GenomicBin.GC].Add(bin.Count);
// Add to the global list of counts.
counts.Add(bin.Count);
}
}
private void initialize()
{
IEnumerable <SampleGenomicBin> onTargetBins = manifest == null ? bins : EnrichmentUtilities.GetOnTargetBins(bins, manifest);
List <double> x = new List <double>();
List <double> y = new List <double>();
withoutChrY = new List <int>();
int i = 0; // index into x and y
foreach (var bin in onTargetBins)
{
double count = countTransformer(bin.Count); // Variance stablization
if (!double.IsInfinity(count))
{
x.Add(bin.GenomicBin.GC);
y.Add(count);
string chrom = bin.GenomicBin.Chromosome.ToLower();
bool isChrY = chrom == "chry" || chrom == "y";
if (!isChrY)
{
withoutChrY.Add(i);
}
i++;
}
}
gcs = x.ToArray();
counts = y.ToArray();
}
/// <summary>
/// Remove bins with extreme GC content.
/// </summary>
/// <param name="bins">Genomic bins in from which we filter out GC content outliers.</param>
/// <param name="threshold">Minimum number of bins with the same GC content required to keep a bin.</param>
///
/// The rationale of this function is that a GC normalization is performed by computing the median count
/// for each possible GC value. If that count is small, then the corresponding normalization constant
/// is unstable and we shouldn't use these data.
static List <SampleGenomicBin> RemoveBinsWithExtremeGC(List <SampleGenomicBin> bins, int threshold, NexteraManifest manifest = null)
{
// Will hold outlier-removed bins.
List <SampleGenomicBin> stripped = new List <SampleGenomicBin>();
// used to count the number of bins with each possible GC content (0-100)
int[] counts = new int[EnrichmentUtilities.numberOfGCbins];
double totalCount = 0;
foreach (SampleGenomicBin bin in manifest == null ? bins : EnrichmentUtilities.GetOnTargetBins(bins, manifest))
{
// We only count autosomal bins because these are the ones we computed normalization factor upon.
if (!GenomeMetadata.SequenceMetadata.IsAutosome(bin.GenomicBin.Chromosome))
{
continue;
}
counts[bin.GenomicBin.GC]++;
totalCount++;
}
int averageCountPerGC = Math.Max(minNumberOfBinsPerGCForWeightedMedian, (int)(totalCount / counts.Length));
threshold = Math.Min(threshold, averageCountPerGC);
foreach (SampleGenomicBin bin in bins)
{
// Remove outlier (not a lot of bins with the same GC content)
if (counts[bin.GenomicBin.GC] < threshold)
{
continue;
}
stripped.Add(bin);
}
return(stripped);
}
