private static RegressionLine CalcConversion(IList <TimeScorePair> listPepCorr, int minCount)
{
var listTime = listPepCorr.Select(p => p.Time).ToList();
var listScore = listPepCorr.Select(p => p.Score).ToList();
return(IrtRegression.TryGet <RegressionLine>(listScore, listTime, minCount, out var line)
? (RegressionLine)line
: null);
}
private bool TryGetCirtRegression(int count, out RegressionLine regression, out List <ScoredPeptide> peptides)
{
peptides = new List <ScoredPeptide>(_cirtPeptides);
var rts = _cirtPeptides.Select(pep => pep.Peptide.RetentionTime).ToList();
var irts = _cirtPeptides.Select(pep => _cirtAll[pep.Peptide.Target]).ToList();
var removedValues = new List <Tuple <double, double> >();
var success = IrtRegression.TryGet <RegressionLine>(rts, irts, count, out var line, removedValues);
regression = (RegressionLine)line;
if (!success)
{
return(false);
}
for (var i = peptides.Count - 1; i >= 0; i--)
{
if (removedValues.Contains(Tuple.Create(rts[i], irts[i])))
{
peptides.RemoveAt(i);
}
}
return(peptides.Count >= count);
}
public RetentionTimeProviderData(IrtRegressionType regressionType, IRetentionTimeProvider retentionTimes,
IReadOnlyList <DbIrtPeptide> standardPeptides, IReadOnlyList <DbIrtPeptide> heavyStandardPeptides)
{
RetentionTimeProvider = retentionTimes;
Peptides = new List <Peptide>(standardPeptides.Count);
for (var i = 0; i < standardPeptides.Count; i++)
{
var heavy = heavyStandardPeptides[i] != null;
var standard = heavy ? heavyStandardPeptides[i] : standardPeptides[i];
var rt = retentionTimes.GetRetentionTime(standard.ModifiedTarget);
if (!rt.HasValue && heavy)
{
standard = standardPeptides[i];
rt = retentionTimes.GetRetentionTime(standard.ModifiedTarget);
}
Peptides.Add(new Peptide(standard.ModifiedTarget, rt, standard.Irt));
}
Peptides.Sort((x, y) => x.Irt.CompareTo(y.Irt));
if (!FilteredPeptides.Any())
{
Regression = null;
RegressionRefined = null;
RegressionSuccess = false;
}
var filteredRt = FilteredPeptides.Select(pep => pep.RetentionTime.Value).ToList();
var filteredIrt = FilteredPeptides.Select(pep => pep.Irt).ToList();
var removed = new List <Tuple <double, double> >();
if (ReferenceEquals(regressionType, IrtRegressionType.LINEAR))
{
Regression = new RegressionLine(filteredRt.ToArray(), filteredIrt.ToArray());
}
else if (ReferenceEquals(regressionType, IrtRegressionType.LOGARITHMIC))
{
Regression = new LogRegression(filteredRt, filteredIrt);
}
else if (ReferenceEquals(regressionType, IrtRegressionType.LOWESS))
{
Regression = new LoessRegression(filteredRt.ToArray(), filteredIrt.ToArray());
}
else
{
throw new ArgumentException();
}
IIrtRegression regressionRefined;
if (IrtRegression.Accept(Regression, MinPoints))
{
regressionRefined = Regression;
Regression = null;
RegressionSuccess = true;
}
else
{
RegressionSuccess = IrtRegression.TryGet(Regression, filteredRt, filteredIrt, MinPoints, out regressionRefined, removed);
}
RegressionRefined = regressionRefined;
foreach (var remove in removed)
{
for (var i = 0; i < Peptides.Count; i++)
{
var peptide = Peptides[i];
if (peptide.RetentionTime.Equals(remove.Item1) && peptide.Irt.Equals(remove.Item2))
{
Peptides[i] = new Peptide(peptide, true);
}
}
}
}
public static ProcessedIrtAverages ProcessRetentionTimesCirt(IProgressMonitor monitor,
IRetentionTimeProvider[] providers, DbIrtPeptide[] cirtPeptides, int numCirt, IrtRegressionType regressionType, out DbIrtPeptide[] chosenCirtPeptides)
{
chosenCirtPeptides = new DbIrtPeptide[0];
var irts = new TargetMap <DbIrtPeptide>(IrtStandard.CIRT.Peptides.Select(pep => new KeyValuePair <Target, DbIrtPeptide>(pep.ModifiedTarget, pep)));
var targetRts = new Dictionary <Target, List <double> >();
var targetCounts = new Dictionary <Target, int>(); // count how many successful regressions each peptide participated in
foreach (var provider in providers)
{
if (monitor.IsCanceled)
{
return(null);
}
var times = (
from pep in cirtPeptides
let rt = provider.GetRetentionTime(pep.ModifiedTarget)
where rt.HasValue
select Tuple.Create(pep.ModifiedTarget, rt.Value, irts[pep.ModifiedTarget].Irt)).ToList();
foreach (var(target, rt, _) in times)
{
if (targetRts.TryGetValue(target, out var list))
{
list.Add(rt);
}
else
{
targetRts[target] = new List <double> {
rt
}
};
}
var removed = new List <Tuple <double, double> >();
if (!IrtRegression.TryGet <RegressionLine>(times.Select(t => t.Item2).ToList(), times.Select(t => t.Item3).ToList(),
MIN_PEPTIDES_COUNT, out _, removed))
{
continue;
}
foreach (var(removeRt, removeIrt) in removed)
{
times.Remove(times.First(time => time.Item2.Equals(removeRt) && time.Item3.Equals(removeIrt)));
}
foreach (var(target, _, _) in times)
{
targetCounts[target] = targetCounts.TryGetValue(target, out var existing) ? existing + 1 : 1;
}
}
// for each target, only keep median retention time
var dupTargets = targetRts.Where(kvp => kvp.Value.Count > 1).Select(kvp => kvp.Key).ToArray();
foreach (var target in dupTargets)
{
targetRts[target] = new List <double> {
new Statistics(targetRts[target]).Median()
}
}
;
// separate targets into equal retention time bins
var candidateBins = new List <Tuple <Target, double> > [numCirt];
for (var i = 0; i < candidateBins.Length; i++)
{
candidateBins[i] = new List <Tuple <Target, double> >();
}
var minRt = double.MaxValue;
var maxRt = -1d;
foreach (var rt in targetRts.Values.Select(list => list.First()))
{
if (rt < minRt)
{
minRt = rt;
}
if (rt > maxRt)
{
maxRt = rt;
}
}
var binSize = (maxRt - minRt) / numCirt;
var lastBinIdx = candidateBins.Length - 1;
foreach (var target in targetRts)
{
foreach (var rt in target.Value)
{
candidateBins[Math.Min((int)((rt - minRt) / binSize), lastBinIdx)].Add(Tuple.Create(target.Key, rt));
}
}
Tuple <Target, double> GetBest(List <Tuple <Target, double> > bin, int binIdx, out int bestCount, out double bestRtDelta)
{
if (bin.Count == 0)
{
bestCount = 0;
bestRtDelta = 0;
return(null);
}
bestCount = 0;
var filtered = new List <Tuple <Target, double> >();
foreach (var t in bin)
{
if (!targetCounts.TryGetValue(t.Item1, out var count))
{
continue;
}
if (count > bestCount)
{
bestCount = count;
filtered.Clear();
filtered.Add(t);
}
else if (count == bestCount)
{
filtered.Add(t);
}
}
if (filtered.Count == 0)
{
filtered = bin;
}
var targetRt = ((minRt + binSize * binIdx) + (minRt + binSize * (binIdx + 1))) / 2;
var closest = filtered.Aggregate((x, y) => Math.Abs(x.Item2 - targetRt) < Math.Abs(y.Item2 - targetRt) ? x : y);
bestRtDelta = Math.Abs(closest.Item2 - targetRt);
return(closest);
}
var chosenList = new List <DbIrtPeptide>();
var emptyBins = new HashSet <int>();
for (var i = 0; i < candidateBins.Length; i++)
{
var bin = candidateBins[i];
if (bin.Count > 0)
{
// choose the best from this bin
var best = GetBest(bin, i, out _, out _);
chosenList.Add(irts[best.Item1]);
bin.Remove(best);
}
else
{
emptyBins.Add(i);
}
}
foreach (var emptyIdx in emptyBins)
{
var bins = new List <int>();
var left = emptyIdx - 1;
var right = emptyIdx + 1;
while (bins.Count == 0)
{
if (left >= 0)
{
bins.Add(left);
}
if (right < candidateBins.Length)
{
bins.Add(right);
}
left--;
right++;
}
Tuple <Target, double> best = null;
var bestBinIdx = -1;
var bestCount = 0;
var bestRtDelta = 0d;
foreach (var i in bins)
{
var current = GetBest(candidateBins[i], i, out var count, out var rtDelta);
if (count > bestCount || (count == bestCount && rtDelta < bestRtDelta))
{
best = current;
bestBinIdx = i;
bestCount = count;
bestRtDelta = rtDelta;
}
}
if (best != null)
{
chosenList.Add(irts[best.Item1]);
candidateBins[bestBinIdx].Remove(best);
}
}
// Process retention times using the chosen peptides
chosenCirtPeptides = chosenList.ToArray();
return(ProcessRetentionTimes(monitor, providers, chosenCirtPeptides, new DbIrtPeptide[0], regressionType));
}
public void GetCurve(RetentionTimeStatistics statistics, out double[] hydroScores, out double[] predictions)
{
IrtRegression.GetCurve(this, statistics, out hydroScores, out predictions);
}
