/// <summary>
/// Developer debug method: ROC curve data generation
/// - Report all intervals, associated QScores and QScore predictor values to an extended report output file
/// - Report called (i.e. TP+FP) intervals grouped by QScore
/// - Generate 2 ROC outputs
/// - ROC_intervals: FP vs TP rate, unit=1 interval
/// - ROC_bases: FP vs TP rate, unit=1 base
/// (Note: In both cases, we ignore intervals shorter than 1kb as most of them are due to imprecise ends of segments, which we don't want to give any weight to)
/// </summary>
private void GenerateReportAndRocDataForQscoreMethod(CanvasSegment.QScoreMethod qscoreMethod, Dictionary<string, List<CNInterval>> resegmentedKnownCN)
{
// Create map interval->{segment+qscore}, ignoring intervals shorter than 1kb
Dictionary<CNInterval, Tuple<CanvasSegment, int>> Interval2Segment = new Dictionary<CNInterval, Tuple<CanvasSegment, int>>();
foreach (string chr in resegmentedKnownCN.Keys)
{
foreach (CNInterval interval in resegmentedKnownCN[chr])
{
foreach (CanvasSegment segment in this.Segments)
{
if (segment.Chr == chr && (segment.Begin == interval.Start || segment.End == interval.End))
{
if (interval.End - interval.Start >= 1000)
Interval2Segment[interval] = new Tuple<CanvasSegment, int>(segment, segment.ComputeQScore(qscoreMethod));
}
}
}
}
// Classify intervals by QScore
List<List<CNInterval>> intervalsByQScore = new List<List<CNInterval>>();
foreach (CNInterval interval in Interval2Segment.Keys)
{
int qscore = Interval2Segment[interval].Item2;
// Resize list to hold this qscore's entries
while (qscore >= intervalsByQScore.Count())
{
intervalsByQScore.Add(new List<CNInterval>());
}
intervalsByQScore[qscore].Add(interval);
}
// Output data as ExtendedCallsVersusKnownCN.txt
string debugPath = Path.Combine(this.OutputFolder, "qscore_" + qscoreMethod.ToString() + "_ExtendedCallsVersusKnownCN.txt");
using (StreamWriter writer = new StreamWriter(debugPath))
{
writer.Write("#Chr\tBegin\tEnd\tTruthSetCN\tCalledCN\tMajorChromCount\tQScore\tInfo");
foreach (CanvasSegment.QScorePredictor predictorId in CanvasSegment.QScorePredictor.GetValues(typeof(CanvasSegment.QScorePredictor)))
{
writer.Write("\tPredictor_{0}", predictorId.ToString());
}
writer.WriteLine("");
foreach (string chr in resegmentedKnownCN.Keys)
{
foreach (CNInterval interval in resegmentedKnownCN[chr])
{
if (Interval2Segment.ContainsKey(interval))
{
CanvasSegment segment = Interval2Segment[interval].Item1;
int qscore = Interval2Segment[interval].Item2;
string info = (interval.CN == segment.CopyNumber) ? "Correct" : "Incorrect";
writer.Write("{0}\t{1}\t{2}\t{3}\t{4}\t{5}\t{6}\t{7}", chr, interval.Start, interval.End, interval.CN, segment.CopyNumber, segment.MajorChromosomeCount, qscore, info);
foreach (CanvasSegment.QScorePredictor predictorId in CanvasSegment.QScorePredictor.GetValues(typeof(CanvasSegment.QScorePredictor)))
{
writer.Write("\t{0}", segment.GetQScorePredictor(predictorId));
}
writer.WriteLine("");
}
else
{
string info = "Missing";
int CN = -1;
int majorChromosomeCount = -1;
writer.WriteLine("{0}\t{1}\t{2}\t{3}\t{4}\t{5}\t{6}", chr, interval.Start, interval.End, interval.CN, CN, majorChromosomeCount, info);
}
}
}
}
// Output data by QScore
debugPath = Path.Combine(this.OutputFolder, "qscore_" + qscoreMethod + "_cnaPerQscore.txt");
using (StreamWriter writer = new StreamWriter(debugPath))
{
writer.WriteLine("#Chr\tBegin\tEnd\tTruthSetCN\tCalledCN\tMajorChromCount\tMedianMAF\tMedianCoverage\tQScore\tInfo");
for (int qscore = 0; qscore < intervalsByQScore.Count(); qscore++)
{
foreach (CNInterval interval in intervalsByQScore[qscore])
{
CanvasSegment segment = Interval2Segment[interval].Item1;
string info = (interval.CN == segment.CopyNumber) ? "Correct" : "Incorrect";
writer.WriteLine("{0}\t{1}\t{2}\t{3}\t{4}\t{5}\t{6}\t{7}", segment.Chr, interval.Start, interval.End, interval.CN, segment.CopyNumber, segment.MajorChromosomeCount, qscore, info);
}
}
}
// ROC output per interval
debugPath = Path.Combine(this.OutputFolder, "qscore_" + qscoreMethod + "_ROC_intervals.txt");
GenerateRocOutput(debugPath, intervalsByQScore, Interval2Segment, false, false);
// ROC output per base
debugPath = Path.Combine(this.OutputFolder, "qscore_" + qscoreMethod + "_ROC_bases.txt");
GenerateRocOutput(debugPath, intervalsByQScore, Interval2Segment, true, false);
}
