|
public static GetCountsByGC ( List |
||
| bins | List |
Bins whose counts are to be normalized |
| manifest | NexteraManifest | |
| countsByGC | List |
An array of lists. Each array element (0-100) will hold a list of counts whose bins have the same GC content. |
| counts | List |
Will hold all of the autosomal counts present in 'bins' |
| return | void | |
/// <summary>
/// Perform a simple GC normalization.
/// </summary>
/// <param name="bins">Bins whose counts are to be normalized.</param>
/// <param name="manifest"></param>
static void NormalizeByGC(List <SampleGenomicBin> bins, NexteraManifest manifest = null)
{
// DebugPrintCountsByGC(bins, "CountsByGC-Before.txt");
// An array of lists. Each array element (0-100) will hold a list of counts whose bins have the same GC content.
List <float>[] countsByGC;
// Will hold all of the autosomal counts present in 'bins'
List <float> counts;
EnrichmentUtilities.GetCountsByGC(bins, manifest, out countsByGC, out counts);
double globalMedian = Utilities.Median(counts);
double?[] medians = new double?[countsByGC.Length];
// Compute the median count for each GC bin
for (int gcBinIndex = 0; gcBinIndex < countsByGC.Length; gcBinIndex++)
{
if (countsByGC[gcBinIndex].Count >= defaultMinNumberOfBinsPerGC)
{
medians[gcBinIndex] = Utilities.Median(countsByGC[gcBinIndex]);
}
else
{
List <Tuple <float, float> > weightedCounts = GetWeightedCounts(countsByGC, gcBinIndex);
medians[gcBinIndex] = Utilities.WeightedMedian(weightedCounts);
}
}
// Divide each count by the median count of bins with the same GC content
for (int gcBinIndex = 0; gcBinIndex < bins.Count; gcBinIndex++)
{
double?median = medians[bins[gcBinIndex].GenomicBin.GC];
if (median != null && median > 0)
{
bins[gcBinIndex].Count = (float)(globalMedian * (double)bins[gcBinIndex].Count / median);
}
}
// DebugPrintCountsByGC(bins, "CountsByGC-After.txt");
}
/// <summary>
/// Perform variance stabilization by GC bins.
/// </summary>
/// <param name="bins">Bins whose counts are to be normalized.</param>
static bool NormalizeVarianceByGC(List <SampleGenomicBin> bins, NexteraManifest manifest = null)
{
// DebugPrintCountsByGC(bins, "CountsByGCVariance-Before.txt");
// An array of lists. Each array element (0-100) will hold a list of counts whose bins have the same GC content.
List <float>[] countsByGC;
// Will hold all of the autosomal counts present in 'bins'
List <float> counts;
EnrichmentUtilities.GetCountsByGC(bins, manifest, out countsByGC, out counts);
// Estimate quartiles of all bins genomewide
var globalQuartiles = Utilities.Quartiles(counts);
// Will hold interquartile range (IQR) separately for each GC bin
List <float> localIQR = new List <float>(countsByGC.Length);
// Will hold quartiles separately for each GC bin
List <Tuple <float, float, float> > localQuartiles = new List <Tuple <float, float, float> >(countsByGC.Length);
// calculate interquartile range (IQR) for GC bins and populate localQuartiles list
for (int i = 0; i < countsByGC.Length; i++)
{
if (countsByGC[i].Count == 0)
{
localIQR.Add(-1f);
localQuartiles.Add(new Tuple <float, float, float>(-1f, -1f, -1f));
}
else if (countsByGC[i].Count >= defaultMinNumberOfBinsPerGC)
{
localQuartiles.Add(Utilities.Quartiles(countsByGC[i]));
localIQR.Add(localQuartiles[i].Item3 - localQuartiles[i].Item1);
}
else
{
List <Tuple <float, float> > weightedCounts = GetWeightedCounts(countsByGC, i);
double[] quartiles = Utilities.WeightedQuantiles(weightedCounts, new List <float>()
{
0.25f, 0.5f, 0.75f
});
localQuartiles.Add(new Tuple <float, float, float>((float)quartiles[0], (float)quartiles[1], (float)quartiles[2]));
localIQR.Add((float)(quartiles[2] - quartiles[0]));
}
}
// Identify if particular GC bins have IQR twice as large as IQR genomewide
float globalIQR = globalQuartiles.Item3 - globalQuartiles.Item1;
// Holder for GC bins with large IQR (compared to genomewide IQR)
int significantIQRcounter = 0;
for (int i = 10; i < 90; i++)
{
if (globalIQR < localIQR[i] * 2f)
{
significantIQRcounter++;
}
}
if (significantIQRcounter <= 0)
{
return(false);
}
// Divide each count by the median count of bins with the same GC content
foreach (SampleGenomicBin bin in bins)
{
var scaledLocalIqr = localIQR[bin.GenomicBin.GC] * 0.8f;
if (globalIQR >= scaledLocalIqr)
{
continue;
}
// ratio of GC bins and global IQRs
float iqrRatio = scaledLocalIqr / globalIQR;
var medianGCCount = localQuartiles[bin.GenomicBin.GC].Item2;
bin.Count = medianGCCount + (bin.Count - medianGCCount) / iqrRatio;
}
// DebugPrintCountsByGC(bins, "CountsByGCVariance-After.txt");
return(true);
}
