public FalseDiscoveryRateComputationTask(string sInputFile, string sOutputFile)
{
InitializeComponent();
m_fdrTask = null;
m_bCancel = false;
m_sInputFile = sInputFile;
m_sOutputFile = sOutputFile;
}
private void btnCompute_Click(object sender, EventArgs e)
{
bool bHasColumnHeaders = chkColumnHeaders.Checked;
bool bFullOutput = chkFullOutput.Checked;
bool bReportProgress = chkReportProgress.Checked;
bool bHuge = chkHuge.Checked;
FalseDiscoveryRate.PiMethod mPi = FalseDiscoveryRate.PiMethod.One;
if (cmbPIMethod.SelectedIndex == 0)
{
mPi = FalseDiscoveryRate.PiMethod.One;
}
else if (cmbPIMethod.SelectedIndex == 1)
{
mPi = FalseDiscoveryRate.PiMethod.WeightedSum;
}
if (cmbPIMethod.SelectedIndex == 2)
{
mPi = FalseDiscoveryRate.PiMethod.DoubleAverage;
}
if (chkFiltering.Checked)
{
mPi = FalseDiscoveryRate.PiMethod.Filtering;
}
double dCutoff = (double)udFDRCutoff.Value;
int cTableNamesColumns = (int)udNameColumns.Value - 1;
bool bPositiveFDR = chkPFDR.Checked;
bool bSampling = chkSampling.Checked;
bool bAutomatedSampling = false;
int iSampleSize = 0;
double dConvergenceEpsilon = 0.0;
if (bSampling)
{
iSampleSize = (int)udSampleSize.Value;
bAutomatedSampling = chkAutomatedSampling.Checked;
if (bAutomatedSampling)
{
dConvergenceEpsilon = 0.001;
}
}
string sInputFile = txtInputFile.Text;
string sOutputFile = txtOutputFile.Text;
DateTime dtBefore = DateTime.Now;
FalseDiscoveryRateComputationTask task = new FalseDiscoveryRateComputationTask(sInputFile, sOutputFile);
FalseDiscoveryRate t = new FalseDiscoveryRate(cTableNamesColumns, bReportProgress, dCutoff, bHuge, iSampleSize, dConvergenceEpsilon, bHasColumnHeaders, mPi, bPositiveFDR, bFullOutput, task);
task.setTask(t);
task.run();
DateTime dtAfter = DateTime.Now;
}
public void LoadFromFile(string fileName)
{
if (!File.Exists(fileName))
{
throw new FileNotFoundException("Parameter file not found", fileName);
}
XElement docRoot = XElement.Load(fileName);
ApplicationTitle = docRoot.Element("Version").Value;
MergeResult = Convert.ToBoolean(docRoot.Element("MergeResult").Value);
ConflictType = ResolveSearchEngineConflictTypeFactory.Find(docRoot.GetChildValue("ConflictType", ResolveSearchEngineConflictTypeFactory.DiscardAll.Name));
if (docRoot.Element("MinimumEngineAgreeCount") != null)
{
MinimumEngineAgreeCount = int.Parse(docRoot.Element("MinimumEngineAgreeCount").Value);
}
if (docRoot.Element("MergeResultFromSameEngineButDifferentSearchParameters") != null)
{
KeepTopPeptideFromSameEngineButDifferentSearchParameters = bool.Parse(docRoot.Element("MergeResultFromSameEngineButDifferentSearchParameters").Value);
}
if (docRoot.Element("PeptideRetrieval") != null)
{
PeptideRetrieval = bool.Parse(docRoot.Element("PeptideRetrieval").Value);
}
Database.Load(docRoot);
FalseDiscoveryRate.Load(docRoot);
Classification.Load(docRoot);
PeptideFilter.Load(docRoot);
try
{
DatasetList.Load(docRoot);
DatasetList.ForEach(m => m.Parent = this);
}
catch (Exception ex)
{
MessageBox.Show("Load dataset error :" + ex.Message);
}
}
private void CalculateBasicFDR(IList <InputFile> csvFiles, bool isBatched = false)
{
string msg = "Calculating first order FDR threshold";
if (isBatched)
{
msg += " in batch...";
Log(msg);
// Calculate global threshold
double threshold = FalseDiscoveryRate <Peptide, double> .CalculateThreshold(_allPeptides, _maximumFalseDiscoveryRate);
foreach (InputFile csvFile in csvFiles)
{
csvFile.ScoreThreshold = threshold;
}
}
else
{
msg += " separately...";
Log(msg);
// Calculate each file separately
foreach (InputFile csvFile in csvFiles)
{
csvFile.ScoreThreshold = FalseDiscoveryRate <Peptide, double> .CalculateThreshold(csvFile.Peptides, _maximumFalseDiscoveryRate);
}
}
foreach (InputFile csvFile in csvFiles)
{
double threshold = csvFile.ScoreThreshold;
List <Peptide> passingPeptides = csvFile.Peptides.Where(peptide => peptide.FdrScoreMetric <= threshold).ToList();
csvFile.FdrFilteredPeptides = passingPeptides;
List <PSM> passingPsms = csvFile.PeptideSpectralMatches.Where(psm => psm.FdrScoreMetric <= threshold).ToList();
csvFile.FdrFilteredPSMs = passingPsms;
int total = csvFile.FdrFilteredPeptides.Count;
int decoys = csvFile.FdrFilteredPeptides.Count(peptide => peptide.IsDecoy);
int targets = total - decoys;
Log(
string.Format(
"{0:N0} peptides ({1:N0} decoys FDR = {2:F4}) pass the e-value threshold of {3:G4} for {4}",
targets, decoys, 100.0 * decoys / (double)targets, csvFile.ScoreThreshold, csvFile.Name));
}
}
private Tuple <double, double> CalculateBestPPMError(IEnumerable <Peptide> inputPeptides, double maximumFalseDisoveryRate = 0.01, int steps = 10, double minimumIncrement = 0.05)
{
List <Peptide> peptides = inputPeptides.OrderBy(pep => pep.CorrectedPrecursorErrorPPM).ToList();
double[] precursorPPMs = peptides.Select(pep => pep.CorrectedPrecursorErrorPPM).ToArray();
double bestppmError = 0;
double max = peptides[peptides.Count - 1].CorrectedPrecursorErrorPPM;
double maxPrecursorError = Math.Min(max, _maximumPPMError);
double minPrecursorError = 0;
double increment = (maxPrecursorError - minPrecursorError) / steps;
increment = Math.Max(increment, minimumIncrement);
double bestCount = 0;
for (double ppmError = minPrecursorError; ppmError <= maxPrecursorError; ppmError += increment)
{
int index = Array.BinarySearch(precursorPPMs, ppmError);
if (index < 0)
{
index = ~index;
}
int count = FalseDiscoveryRate <Peptide, double> .Count(peptides.Take(index).ToList(), maximumFalseDisoveryRate);
if (count <= bestCount)
{
continue;
}
bestCount = count;
bestppmError = ppmError;
}
List <Peptide> filteredPeptides = new List <Peptide>(peptides.Where(pep => pep.CorrectedPrecursorErrorPPM <= bestppmError));
// Calculate the e-value threshold for those filtered peptides
double threshold = FalseDiscoveryRate <Peptide, double> .CalculateThreshold(filteredPeptides, maximumFalseDisoveryRate);
return(new Tuple <double, double>(bestppmError, threshold));
}
public XElement ToXml()
{
XElement docRoot = new XElement("BuildSummaryOption",
new XElement("Version", ApplicationTitle),
new XElement("MergeResult", MergeResult),
new XElement("ConflictType", ConflictType),
new XElement("MinimumEngineAgreeCount", MinimumEngineAgreeCount),
new XElement("MergeResultFromSameEngineButDifferentSearchParameters", KeepTopPeptideFromSameEngineButDifferentSearchParameters),
new XElement("PeptideRetrieval", PeptideRetrieval));
Database.Save(docRoot);
FalseDiscoveryRate.Save(docRoot);
Classification.Save(docRoot);
PeptideFilter.Save(docRoot);
DatasetList.Save(docRoot);
return(docRoot);
}
public IIdentifiedSpectrumBuilder GetSpectrumBuilder()
{
return(FalseDiscoveryRate.GetSpectrumBuilder());
}
internal void setTask(FalseDiscoveryRate t)
{
m_fdrTask = t;
}
/*
* static void Main(string[] args)
* {
* DirectoryInfo dir = new DirectoryInfo( "D:\\fdr\\SyntheticMultiple" );
* FileInfo[] aFiles = dir.GetFiles();
* int iFile = 0, cFiles = aFiles.Length;
* foreach (FileInfo f in aFiles)
* {
* if (f.Name.StartsWith("synth") && f.Name.EndsWith(".txt"))
* {
* Console.WriteLine("Processing file " + iFile + " out of " + cFiles);
* FalseDiscoveryRate t = new FalseDiscoveryRate(1, true, double.PositiveInfinity, false, -1, 0.0, true, FalseDiscoveryRate.PiMethod.WeightedSum);
* t.computeFDR(f.FullName, f.FullName + ".FDR");
* iFile++;
* }
* }
* }
*/
static void Main(string[] args)
{
if (args.Length < 2)
{
Console.WriteLine("FisherExactTest utility for computing Fisher Exact Test and False Discovery Rate (FDR) for contingency tables.");
Console.WriteLine("Written by Guy Shani, Microsoft Research, October 2008.");
Console.WriteLine("Use the format: FisherExactTest inputfile outputfile <options>");
Console.WriteLine("inputfile - name of the text file containing the contingency table data. Each contingency table resides in a single row with the format a b c d, separateed by tabs.");
Console.WriteLine("outputfile - name of the text file where the results should be written.");
Console.WriteLine("Additional available options:");
Console.WriteLine("[-UseTableNames:<column count>] - each contingency table can be perceeded by table names columns. Use this option to set the number of columns perceeding each table. The default is 0 - no table names.");
Console.WriteLine("[-SilentMode] - execute the application without any progress messages.");
Console.WriteLine("[-FDRCutoff:x] - Allows to avoid outputing tables with an FDR exciding the threshold x.");
Console.WriteLine("[-Huge] - When the input data is huge (more than 100K tables) this option avoids several caching procedures to allow handling more data.");
Console.WriteLine("[-UseSampling[:n]] - Computes FDRs using only n sampled tables. Default value for n is 100K.");
Console.WriteLine("[-AutomatedSampling[:d]] - Activates the automated sampling algorithm, computing FDRs repeatedly with increasing sample sizes until the largest change in FDR drops below d. Default value for d is 0.1.");
Console.WriteLine("[-ColumnHeaders] - use when the input files has column headers.");
Console.WriteLine("[-EvaluatePi] - by default pi0=1, set this parameter to evaluate pi0=sum(observed p-values)/sum(null p-values).");
Console.WriteLine("[-Filtering] - Computes pi0 using only relevant tables. This flag supersedes the EvaluatePi flag.");
Console.WriteLine("[-FullOutput] - outputs all the statistics that were computed. By default only p-values and q-values are written to the output file.");
Console.WriteLine("[-pFDR] - Compute positive FDR. The default is to compute FDR rather than pFDR.");
}
else
{
string sInputFileName = args[0];
string sOutputFileName = args[1];
int cTableNamesColumns = 0;
bool bReportProgress = (findArgument(args, "-SilentMode") == -1);
bool bHuge = (findArgument(args, "-Huge") != -1);
double dCutoff = double.PositiveInfinity;
int iSampleSize = -1;
double dMinimalChangeBetweenSamples = -1.0;
bool bHasColumnHeaders = false;
FalseDiscoveryRate.PiMethod mPi = FalseDiscoveryRate.PiMethod.One;
bool bFullOutput = false;
bool bPositiveFDR = false;
int iColumnHeaders = findArgument(args, "-ColumnHeaders");
if (iColumnHeaders != -1)
{
bHasColumnHeaders = true;
}
int iFullOutput = findArgument(args, "-FullOutput");
if (iFullOutput != -1)
{
bFullOutput = true;
}
int iPFDR = findArgument(args, "-pFDR");
if (iPFDR != -1)
{
bPositiveFDR = true;
}
int iEvaluatePi = findArgument(args, "-EvaluatePi");
if (iEvaluatePi != -1)
{
mPi = FalseDiscoveryRate.PiMethod.WeightedSum;
}
iEvaluatePi = findArgument(args, "-Filtering");
if (iEvaluatePi != -1)
{
mPi = FalseDiscoveryRate.PiMethod.Filtering;
}
int iCutoff = findArgument(args, "-FDRCutoff");
if (iCutoff != -1)
{
int idx = args[iCutoff].IndexOf(':');
dCutoff = double.Parse(args[iCutoff].Substring(idx + 1));
}
int iTableNames = findArgument(args, "-UseTableNames");
if (iTableNames != -1)
{
int idx = args[iTableNames].IndexOf(':');
cTableNamesColumns = int.Parse(args[iTableNames].Substring(idx + 1));
}
int iSampling = findArgument(args, "-UseSampling");
if (iSampling != -1)
{
int idx = args[iSampling].IndexOf(':');
if (idx > 0)
{
iSampleSize = int.Parse(args[iSampling].Substring(idx + 1));
}
else
{
iSampleSize = 100000;
}
}
int iAutomatedSampling = findArgument(args, "-AutomatedSampling");
if (iAutomatedSampling != -1)
{
int idx = args[iAutomatedSampling].IndexOf(':');
if (iSampleSize == -1)
{
iSampleSize = 100000;
}
if (idx < 0)
{
dMinimalChangeBetweenSamples = 0.01;
}
else
{
dMinimalChangeBetweenSamples = double.Parse(args[iAutomatedSampling].Substring(idx + 1));
}
}
DateTime dtBefore = DateTime.Now;
ProgressReport pr = new ConsoleProgressReport();
FalseDiscoveryRate t = new FalseDiscoveryRate(cTableNamesColumns, bReportProgress, dCutoff, bHuge, iSampleSize, dMinimalChangeBetweenSamples, bHasColumnHeaders, mPi, bPositiveFDR, bFullOutput, pr);
t.computeFDR(sInputFileName, sOutputFileName);
DateTime dtAfter = DateTime.Now;
if (bReportProgress)
{
Console.WriteLine("Total execution time " + dtAfter.Subtract(dtBefore));
}
}
}