Package org.fhcrc.cpl.viewer.feature.extraction

Interface Summary
FeatureScorer Assign a score to a group of peaks, considered together as a Feature.
PeakCombiner Combine peaks into peptide features
PeakExtractor Extract peaks from spectra.
 

Class Summary
AccurateMassAdjuster For adjusting the masses of features found in a resampled space, to take advantage of the higher mass accuracy to be found in the unresampled space
BackgroundRemover Removes background noise from spectra
BasePeakCombiner superclass for peak combiners
DefaultFeatureScorer Default feature scorer.
DefaultPeakCombiner Default class for creating Features from Peaks 07/29/20008 dhmay enhanced to handle negatively-charged ions
FeatureFinder  
FeatureFindingBroker This class exists to bridge the gap between old-school and new-school feature finders.
SmallMoleculePeakCombiner class for creating Features from Peaks, specific to small-molecule analysis, e.g., metabolites.
SmootherCreator // Utility methods for creating different kinds of smoothers // So what's up with all these smoothers??? // // Really this is a problem, the sample rate varies drastically // across experimental setups.
SpectrumResampler Resample spectra onto a grid with the specified frequency.
WaveletPeakExtractor smooth, ExtractMaxima2D.analyze() Sort by decreasing intensity Throw out peaks with < 5 scans Break up what look like double-humped elution profiles by breaking peaks at valleys in the profile For each peak: -Compute the extents -Walking down scans, make sure that the peak's intensity is greater than the intensities of the two m/z's around it -Make sure that intensity is above threshold -If you start going back up in intensity, stop -Same for end scan -If too short (<5 scans), toss it out -"Integrate" to determine total intensity: for each scan, add a block of intensity corresponding to the intensity on that scan multiplied by half the amount of time between the adjoining scans -Create a feature to represent the individual peak Exclude all features initially Filter features: for each feature: find all neighbors within 5 scans and 1.1 m/z units if the m/z's are too close (within 1.5 / _resamplingFrequency, for some reason), no dice make sure they're lined up, scan-wise If not enough scans combined in the two features, no dice If intensities not high enough, no dice If we got here, unexclude both
 



Fred Hutchinson Cancer Research Center