public static DataRow extractFeatureVector(String accession, List <PeptideScore> peptideScores)
{
IdentificationFeatures idf = extractFeatures(accession, peptideScores);
int cardinality = idf.getCardinality();
Double highestConfidenceScore = idf.getHighestConfidenceScore();
Double meanConfidenceScore = idf.getMeanConfidenceScore();
Double medianConfidenceScore = idf.getMedianConfidenceScore();
Double highestDCN = idf.getHighestDCN();
Double meanDCN = idf.getMeanDCN();
Double medianDCN = idf.getMedianDCN();
#if STDEVINCLUDED
Double stdevConfidenceScore = idf.getStdevConfidenceScore();
if (cardinality == 0 || cardinality == 1)
{
stdevConfidenceScore = DEFAULT_STDEV_MAX;
//stdevConfidenceScore = max_stdev;
//if ((int)Math.Round(stdevConfidenceScore) == 0)
//{
// stdevConfidenceScore = DEFAULT_STDEV_MAX;
//}
}
DataRow r = CreateRow(accession, cardinality, highestConfidenceScore, meanConfidenceScore,
medianConfidenceScore, highestDCN, meanDCN, medianDCN,
stdevConfidenceScore);
#else
DataRow r = CreateRow(accession, cardinality, highestConfidenceScore, meanConfidenceScore,
medianConfidenceScore, highestDCN, meanDCN, medianDCN);
#endif
return(r);
}
/*
* Write the identification features used for training the logistic regression
* classifier
*/
public static void WriteIdentificationFeaturesFile(String file_path,
List <IdentificationFeatures> positiveTrainingSet,
List <IdentificationFeatures> negativeTrainingSet)
{
log.Debug("Writing Identification Features to a file...");
try
{
StreamWriter writer = new StreamWriter(file_path);
log.Debug("File name: " + file_path);
// Write header TODO remove
String header = "label," + IdentificationFeatures.getHeader();
writer.Write(header);
// in the first column, 1 indicates positive training set
foreach (IdentificationFeatures i in positiveTrainingSet)
{
writer.Write("\n" + "1," + i.writeToFile());
writer.Flush();
}
// in the first column, 0 indicates negative training set
foreach (IdentificationFeatures i in negativeTrainingSet)
{
writer.Write("\n" + "0," + i.writeToFile());
writer.Flush();
}
writer.Flush();
writer.Close();
}
catch (Exception e)
{
Console.WriteLine(e.ToString());
log.Error("Writing file unsuccessful!!!");
Console.ReadKey();
Environment.Exit(0);
}
log.Debug("Writing file successful.");
}
public static IdentificationFeatures extractFeatures(String accession, List <PeptideScore> peptideScores)
{
int cardinality = peptideScores.Count;
if (cardinality < 1)
{
return(new IdentificationFeatures(accession, cardinality, 0, 0, 0, 0, 0, 0, 0));
}
//TODO should this be number of unique peptides or number of peptide scores??
Double highestConfidenceScore = Double.MinValue;
Double meanConfidenceScore;
Double medianConfidenceScore;
Double stdevConfidenceScore;
Double highestDCN = Double.MinValue;
Double meanDCN;
Double medianDCN;
double[] scores = new double[cardinality];
double[] dCNList = new double[cardinality];
for (int i = 0; i < peptideScores.Count; i++)
{
PeptideScore s = peptideScores[i];
Double confidenceScore = s.getXCorr();
Double dCN = s.getdCN();
scores[i] = confidenceScore;
dCNList[i] = dCN;
// calculate highest confidence score
if (confidenceScore > highestConfidenceScore)
{
highestConfidenceScore = confidenceScore;
}
if (dCN > highestDCN)
{
highestDCN = dCN;
}
}
// calculate mean
meanConfidenceScore = scores.Average();
meanDCN = dCNList.Average();
// calculate median
medianConfidenceScore = CalculateMedian(scores);
medianDCN = CalculateMedian(dCNList);
// calculate stdev
// bias correction set to true
// that means the stdev formula uses "N-1" as the denominator, I believe.
// this helps to estimate the variance more accurately for a small N
stdevConfidenceScore = CalculateStdDev(scores);
IdentificationFeatures f = new IdentificationFeatures(accession, cardinality, highestConfidenceScore, meanConfidenceScore, medianConfidenceScore,
highestDCN, meanDCN, medianDCN, stdevConfidenceScore);
return(f);
}
