org.fhcrc.cpl.viewer.amt
Class AmtMatchProbabilityAssigner

java.lang.Object
  org.fhcrc.cpl.viewer.amt.AmtMatchProbabilityAssigner

public class AmtMatchProbabilityAssigner
extends java.lang.Object

This class assigns a probability to every AMT match, based on the distribution of target and decoy mass and H errors. First, all matches (including multiple matches to the same feature) are mined for their mass and H error values. Those values are fed to an Expectation-Maximization algorithm, implemented in R, which estimates parameters for a mixed distribution. We assume the true distribution to be a uniform distribution of false matches mixed with a bivariate normal distribution of true matches. We also feed the EM algorithm an initial estimate of the proportion of matches in the bivariate normal distribution, which is based on a match to a decoy database. The EM algorithm returns distribution parameters, as well as a probability for each match indicating how likely it is that the match is in the bivariate normal distribution of true matches. Lots of charts are provided in order to let the user judge the success of the algorithm in modeling the distribution.

Field Summary

protected boolean	converged
protected java.util.List<java.lang.Double>	decoyAreas
protected java.util.List<java.lang.Integer>	decoyWindowFeatureNumbers
static float	DEFAULT_EM_MAX_DELTA_P_FOR_STABLE
static int	DEFAULT_EM_MAX_ITERATIONS_STABLE_FOR_CONVERGENCE
static int	DEFAULT_MAX_EM_ITERATIONS
static float	DEFAULT_MAX_SECONDBEST_PROBABILITY
static int	DEFAULT_MIN_EM_ITERATIONS
static float	DEFAULT_MIN_MATCH_PROBABILITY
static float	DEFAULT_MIN_SECONDBEST_PROBABILITY_DIFFERENCE
protected float	expectedTrue
protected float	initialProportionTrue
protected boolean	keepStatistics
static float	KS_CUTOFF_FOR_WARN
protected float	ks_score_x
protected float	ks_score_y
static int	MAX_R_PROB_ASSIGNMENT_MILLIS
static int	MAX_TREE_DEPTH
protected float	maxDeltaElution
protected float	maxDeltaMass
protected int	maxEMIterations
protected int	maxRProbAssignmentMillis
protected float	maxSecondBestProbability
protected static float	maxXScaff
protected static float	maxYScaff
protected float	meanProbability
protected float	minDeltaElution
protected float	minDeltaMass
protected int	minEMIterations
protected float	minMatchProbability
protected float	minSecondBestProbabilityDifference
protected static float	minXScaff
protected static float	minYScaff
protected float	mu_x
protected float	mu_y
static int	NUM_DECOY_POINTS
protected int	num_iterations
static int	NUM_TARGET_POINTS
protected float	proportion
protected float	quantileBetaX
protected float	quantileBetaY
protected float	quantileCorrX
protected float	quantileCorrY
static int	SCALING_FACTOR
protected float	sigma_x
protected float	sigma_y
protected java.util.List<java.lang.Double>	targetAreas
protected java.util.List<java.lang.Integer>	targetWindowFeatureNumbers
protected float	totalElutionRange
protected float	totalMassRange

Constructor Summary

AmtMatchProbabilityAssigner(float minDeltaMass, float maxDeltaMass, float minDeltaElution, float maxDeltaElution, float minMatchProbability)
Set the loose matching parameters.

Method Summary

java.util.List<Feature>	assignMatchesAndProbabilities(FeatureSetMatcher.FeatureMatchingResult targetMatchingResult, FeatureSetMatcher.FeatureMatchingResult decoyMatchingResult, boolean showCharts) Use the initial (loose-tolerance) matching data, to both the target and decoy databases, to determine a probability for each loose match.
protected double	calculateLocalNormalizedDensity(double xpos, double ypos, double expectedNumTrue, double binSizeX, double binSizeY) Use the normal CDF to calculate the density of a small area of the distribution
float[]	calculateProbabilitiesEM(java.util.List<java.lang.Float> targetMassErrorDataList, java.util.List<java.lang.Float> targetHErrorDataList, float proportionTrue, boolean showCharts) Use the Expectation Maximization algorithm to determine match probabilities by modeling the true hits as a normal distribution, and the false hits as a uniform distribution, superimposed in the target distribution.
float	getExpectedTrue()
float	getInitialProportionTrue()
float	getKsScoreX()
float	getKsScoreY()
int	getMaxEMIterations()
int	getMaxRProbAssignmentMillis()
float	getMaxSecondBestProbability()
int	getMinEMIterations()
float	getMinSecondBestProbabilityDifference()
float	getMuX()
float	getMuY()
int	getNumIterations()
float	getQuantileBetaX()
float	getQuantileBetaY()
float	getQuantileCorrX()
float	getQuantileCorrY()
float	getSigmaX()
float	getSigmaY()
boolean	isConverged()
protected boolean	isInScaffRange(double x, double y)
void	setInitialProportionTrue(float initialProportionTrue)
void	setMaxEMIterations(int maxEMIterations)
void	setMaxRProbAssignmentMillis(int maxRProbAssignmentMillis)
void	setMaxSecondBestProbability(float maxSecondBestProbability)
void	setMinEMIterations(int minEMIterations)
void	setMinSecondBestProbabilityDifference(float minSecondBestProbabilityDifference)

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

minXScaff

protected static float minXScaff

maxXScaff

protected static float maxXScaff

minYScaff

protected static float minYScaff

maxYScaff

protected static float maxYScaff

MAX_R_PROB_ASSIGNMENT_MILLIS

public static final int MAX_R_PROB_ASSIGNMENT_MILLIS

See Also:

Constant Field Values

maxRProbAssignmentMillis

protected int maxRProbAssignmentMillis

minDeltaMass

protected float minDeltaMass

maxDeltaMass

protected float maxDeltaMass

minDeltaElution

protected float minDeltaElution

maxDeltaElution

protected float maxDeltaElution

totalMassRange

protected float totalMassRange

totalElutionRange

protected float totalElutionRange

DEFAULT_MIN_EM_ITERATIONS

public static final int DEFAULT_MIN_EM_ITERATIONS

See Also:

Constant Field Values

DEFAULT_MAX_EM_ITERATIONS

public static final int DEFAULT_MAX_EM_ITERATIONS

See Also:

Constant Field Values

DEFAULT_EM_MAX_DELTA_P_FOR_STABLE

public static final float DEFAULT_EM_MAX_DELTA_P_FOR_STABLE

See Also:

Constant Field Values

DEFAULT_EM_MAX_ITERATIONS_STABLE_FOR_CONVERGENCE

public static final int DEFAULT_EM_MAX_ITERATIONS_STABLE_FOR_CONVERGENCE

See Also:

Constant Field Values

minEMIterations

protected int minEMIterations

maxEMIterations

protected int maxEMIterations

minMatchProbability

protected float minMatchProbability

KS_CUTOFF_FOR_WARN

public static final float KS_CUTOFF_FOR_WARN

See Also:

Constant Field Values

DEFAULT_MIN_MATCH_PROBABILITY

public static final float DEFAULT_MIN_MATCH_PROBABILITY

See Also:

Constant Field Values

maxSecondBestProbability

protected float maxSecondBestProbability

minSecondBestProbabilityDifference

protected float minSecondBestProbabilityDifference

MAX_TREE_DEPTH

public static final int MAX_TREE_DEPTH

See Also:

Constant Field Values

SCALING_FACTOR

public static final int SCALING_FACTOR

See Also:

Constant Field Values

NUM_TARGET_POINTS

public static final int NUM_TARGET_POINTS

See Also:

Constant Field Values

NUM_DECOY_POINTS

public static final int NUM_DECOY_POINTS

See Also:

Constant Field Values

DEFAULT_MAX_SECONDBEST_PROBABILITY

public static final float DEFAULT_MAX_SECONDBEST_PROBABILITY

See Also:

Constant Field Values

DEFAULT_MIN_SECONDBEST_PROBABILITY_DIFFERENCE

public static final float DEFAULT_MIN_SECONDBEST_PROBABILITY_DIFFERENCE

See Also:

Constant Field Values

keepStatistics

protected boolean keepStatistics

targetAreas

protected java.util.List<java.lang.Double> targetAreas

decoyAreas

protected java.util.List<java.lang.Double> decoyAreas

targetWindowFeatureNumbers

protected java.util.List<java.lang.Integer> targetWindowFeatureNumbers

decoyWindowFeatureNumbers

protected java.util.List<java.lang.Integer> decoyWindowFeatureNumbers

initialProportionTrue

protected float initialProportionTrue

meanProbability

protected float meanProbability

expectedTrue

protected float expectedTrue

ks_score_x

protected float ks_score_x

ks_score_y

protected float ks_score_y

quantileCorrX

protected float quantileCorrX

quantileCorrY

protected float quantileCorrY

quantileBetaX

protected float quantileBetaX

quantileBetaY

protected float quantileBetaY

mu_x

protected float mu_x

mu_y

protected float mu_y

sigma_x

protected float sigma_x

sigma_y

protected float sigma_y

num_iterations

protected int num_iterations

converged

protected boolean converged

proportion

protected float proportion

Constructor Detail

AmtMatchProbabilityAssigner

public AmtMatchProbabilityAssigner(float minDeltaMass,
                                   float maxDeltaMass,
                                   float minDeltaElution,
                                   float maxDeltaElution,
                                   float minMatchProbability)

Set the loose matching parameters. Possibly should just specify the target and decoy matching results here, too.

Parameters:

minDeltaMass -

maxDeltaMass -

minDeltaElution -

maxDeltaElution -

minMatchProbability -

Method Detail

isInScaffRange

protected boolean isInScaffRange(double x,
                                 double y)

assignMatchesAndProbabilities

public java.util.List<Feature> assignMatchesAndProbabilities(FeatureSetMatcher.FeatureMatchingResult targetMatchingResult,
                                                             FeatureSetMatcher.FeatureMatchingResult decoyMatchingResult,
                                                             boolean showCharts)
                                                      throws java.io.IOException

Use the initial (loose-tolerance) matching data, to both the target and decoy databases, to determine a probability for each loose match. Probabilities get assigned directly to the MS1 features. This method also makes the peptide assignments: one peptide per MS1 feature, break ties using probability of match. Probability for each match is (estimated target density - estimated decoy density) / estimated target density

Parameters:

targetMatchingResult -

decoyMatchingResult -

showCharts -

Throws:

java.io.IOException

calculateProbabilitiesEM

public float[] calculateProbabilitiesEM(java.util.List<java.lang.Float> targetMassErrorDataList,
                                        java.util.List<java.lang.Float> targetHErrorDataList,
                                        float proportionTrue,
                                        boolean showCharts)
                                 throws java.io.IOException

Use the Expectation Maximization algorithm to determine match probabilities by modeling the true hits as a normal distribution, and the false hits as a uniform distribution, superimposed in the target distribution.

Parameters:

targetMassErrorDataList -

targetHErrorDataList -

proportionTrue -

showCharts -

Returns:

Throws:

java.io.IOException

calculateLocalNormalizedDensity

protected double calculateLocalNormalizedDensity(double xpos,
                                                 double ypos,
                                                 double expectedNumTrue,
                                                 double binSizeX,
                                                 double binSizeY)

Use the normal CDF to calculate the density of a small area of the distribution

Parameters:

xpos -

ypos -

expectedNumTrue -

binSizeX -

binSizeY -

Returns:

getInitialProportionTrue

public float getInitialProportionTrue()

setInitialProportionTrue

public void setInitialProportionTrue(float initialProportionTrue)

getExpectedTrue

public float getExpectedTrue()

getKsScoreX

public float getKsScoreX()

getKsScoreY

public float getKsScoreY()

getQuantileCorrX

public float getQuantileCorrX()

getQuantileCorrY

public float getQuantileCorrY()

getQuantileBetaX

public float getQuantileBetaX()

getQuantileBetaY

public float getQuantileBetaY()

getMaxRProbAssignmentMillis

public int getMaxRProbAssignmentMillis()

setMaxRProbAssignmentMillis

public void setMaxRProbAssignmentMillis(int maxRProbAssignmentMillis)

getMuX

public float getMuX()

getMuY

public float getMuY()

getSigmaX

public float getSigmaX()

getSigmaY

public float getSigmaY()

getNumIterations

public int getNumIterations()

getMaxSecondBestProbability

public float getMaxSecondBestProbability()

setMaxSecondBestProbability

public void setMaxSecondBestProbability(float maxSecondBestProbability)

getMinSecondBestProbabilityDifference

public float getMinSecondBestProbabilityDifference()

setMinSecondBestProbabilityDifference

public void setMinSecondBestProbabilityDifference(float minSecondBestProbabilityDifference)

getMaxEMIterations

public int getMaxEMIterations()

setMaxEMIterations

public void setMaxEMIterations(int maxEMIterations)

getMinEMIterations

public int getMinEMIterations()

setMinEMIterations

public void setMinEMIterations(int minEMIterations)

isConverged

public boolean isConverged()