Class Cluster

java.lang.Object
  |
  +--Cluster

public class Cluster

extends java.lang.Object

Class Cluster.java - Cluster the current ProtPlot data. This work was produced by Peter Lemkin of the National Cancer Institute, an agency of the United States Government and Djamel Medjahed (SAIC-Frederick). As a work of the United States Government there is no associated copyright. It is offered as open source software under the Mozilla Public License (version 1.1) subject to the limitations noted in the accompanying LEGAL file.

Version:

$Date: 2002/10/14 20:17:17 $ $Revision: $

Author:

P. Lemkin (NCI), Djamel Medjahed (SAIC), NCI-Frederick, Frederick, MD

See Also:

ProtPlot Home

This work was derived from MAExplorer under the Mozilla 1.1 Open Source Public License by Peter Lemkin of the National Cancer Institute, an agency of the United States Government subject to the limitations noted in the accompanying LEGAL file. See licence info on http://maexplorer.sourceforge.org/

Field Summary

static boolean	allowClusterReportUpdatesFlag flag set if clustering is enabled for the current cluster method
static int	CLUST_KMEANS_PROTEINS_METHOD K-means proteins clustering method
static int	CLUST_SIMILAR_PROTEINS_METHOD Similar proteins clustering method
boolean	clusterDataReadyFlag flag that indicates that the cluster data exists
static int	clusterMethod Cluster method computation method using the CLUST_xxxx values
static int	DIST_AVG_ALL_DATA Jain method 4 for computing distance using missing values using average of all data in the database.
static int	DIST_FISHER_LOW_VALUES method for computing distance using the Fisher clustering low values distance metric that maps low or missing values counts to a low value (i.e.
static int	DIST_WEIGHT Jain method 3 for computing distance using missing values by weighting
(package private) float	distIJ current distance between mPidI and mPidJ
static int	distMode distance computation method using the DIST_xxxx values
(package private) float[]	distSimilarCluster distances [0:mMasterPid-1] sorted by prp.mPidSimilarCluster[] list
static float	distThreshold Distance threshold for similarity clustering
(package private) float[]	dMean mean distances for all samples [0:nPRPs-1]
static float	MAX_DISTANCE value to use if no samples exist
(package private) float[][]	mExprVector Master Protein List [0:nMasterPids-1][0:nSamples-1] of protein expression values from all PRP DBs
(package private) int	mPidI current mPid i
(package private) int	mPidJ current mPid J
(package private) int	nBoth # of samples in both mPidI and mPidJ in current expr profile sample list
(package private) int	nMissing # of missing samples in either mPidI or mPidJ in current expr profile sample list
(package private) int	nPRPs # of samples in entire DB
(package private) int	nSamples # of samples in current expr profile sample list
static boolean	popupClusterWindowFlag flag set it popup cluster control window GUI
private static ProtPlot	prp instance of ProtPlot
java.lang.String	simClustTitle title for current similar cluster string
ShowStringPopup	similarRptPopup the instance of the cluster report.
private static UtilPRP	util Utility instance

Constructor Summary

Cluster(ProtPlot prp)
Cluster() - constructor to setup globals for clustering

Method Summary

float	calcDistIJ(int mPidI, int mPidJ) calcDistIJ() - compute the distance between midI and midJ using distMode method.
void	calcMeanDistIJ() calcMeanDistIJ() - compute mean distances for all samples
boolean	cluster() cluster() - cluster if allowClusterReportUpdatesFlag enabled using the current clusterMethod.
boolean	clusterKMeansProteins() clusterKMeansProteins() - cluster proteins passing the K-means filter.
boolean	clusterKMeansProteins(boolean popupClusterWindowFlag) clusterKMeansProteins() - cluster proteins passing the filter similar to the current protein.
boolean	clusterSimilarProteins() clusterSimilarProteins() - cluster proteins passing the filter similar to the current protein.
boolean	clusterSimilarProteins(boolean popupClusterWindowFlag) clusterSimilarProteins() - cluster proteins passing the filter similar to the current protein.
float	distFisherLowValues(int mPidI, int mPidJ) distFisherLowValues() - compute the Fisher clustering low values distance metric by mapping low or missing values counts to a low value (i.e.
float	distMissing(int mPidI, int mPidJ) distMissing() - return average missing distance between mPidI and mPidJ.
float	distMissingIJ(int s, int mPidI, int mPidJ) distMissingIJ() - compute sample s distance between Esi and Esj for mPidI and mPidJ.
float	distMissingMean(int mPidI, int mPidJ) distMissingMean() - return average missing distance between mPidI and mPidJ.
float	distMissingMeanIJ(int s, int mPidI, int mPidJ) distMissingMeanIJ() - compute sample s distance between Esi and Esj for mPidI and mPidJ.
boolean	isSampleInExprProfileList(int s) isSampleInExprProfileList() - test if sample s is in current Expr Profile list of samples
void	saveCluster() saveCluster() - save current clustered proteins into prp.savedClusterMPIDs[]
static float	setClusterMethod(int clustMethod) setClusterMethod() - set method to use for clustering
static int	setDistMode(int distanceMode) setDistMode() - set the method to use for clustering
static float	setDistThreshold(float distThr) setDistThreshold() - set the distance threshold to use for clustering
java.lang.String	similarProteinsReportStr() similarProteinsReportStr() - compute report string of similar-cluster of filtered proteins
static boolean	updateSeedProtein(int newSeedMPID) updateSeedProtein() - update the seed protein mPid used for clustering and recluster if enabled
void	updateSimilarProteinsReport() updateSimilarProteinsReport() - compute report string of similar-cluster of filtered proteins [TODO] This will update the Cluster Report window text area.

Methods inherited from class java.lang.Object

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

Field Detail

MAX_DISTANCE

public static final float MAX_DISTANCE

value to use if no samples exist

DIST_WEIGHT

public static final int DIST_WEIGHT

Jain method 3 for computing distance using missing values by weighting

DIST_AVG_ALL_DATA

public static final int DIST_AVG_ALL_DATA

Jain method 4 for computing distance using missing values using average of all data in the database.

DIST_FISHER_LOW_VALUES

public static final int DIST_FISHER_LOW_VALUES

method for computing distance using the Fisher clustering low values distance metric that maps low or missing values counts to a low value (i.e. a count of 1). NOTE: this requires the actual counts data.

CLUST_SIMILAR_PROTEINS_METHOD

public static final int CLUST_SIMILAR_PROTEINS_METHOD

Similar proteins clustering method

CLUST_KMEANS_PROTEINS_METHOD

public static final int CLUST_KMEANS_PROTEINS_METHOD

K-means proteins clustering method

prp

private static ProtPlot prp

instance of ProtPlot

util

private static UtilPRP util

Utility instance

mExprVector

float[][] mExprVector

Master Protein List [0:nMasterPids-1][0:nSamples-1] of protein expression values from all PRP DBs

nPRPs

int nPRPs

# of samples in entire DB

nSamples

int nSamples

# of samples in current expr profile sample list

nMissing

int nMissing

# of missing samples in either mPidI or mPidJ in current expr profile sample list

nBoth

int nBoth

# of samples in both mPidI and mPidJ in current expr profile sample list

mPidI

int mPidI

current mPid i

mPidJ

int mPidJ

current mPid J

distIJ

float distIJ

current distance between mPidI and mPidJ

dMean

float[] dMean

mean distances for all samples [0:nPRPs-1]

distSimilarCluster

float[] distSimilarCluster

distances [0:mMasterPid-1] sorted by prp.mPidSimilarCluster[] list

simClustTitle

public java.lang.String simClustTitle

title for current similar cluster string

clusterDataReadyFlag

public boolean clusterDataReadyFlag

flag that indicates that the cluster data exists

allowClusterReportUpdatesFlag

public static boolean allowClusterReportUpdatesFlag

flag set if clustering is enabled for the current cluster method

popupClusterWindowFlag

public static boolean popupClusterWindowFlag

flag set it popup cluster control window GUI

distThreshold

public static float distThreshold

Distance threshold for similarity clustering

distMode

public static int distMode

distance computation method using the DIST_xxxx values

clusterMethod

public static int clusterMethod

Cluster method computation method using the CLUST_xxxx values

similarRptPopup

public ShowStringPopup similarRptPopup

the instance of the cluster report. We make this public since we may update it from elsewhere

Constructor Detail

Cluster

public Cluster(ProtPlot prp)

Cluster() - constructor to setup globals for clustering

Parameters:

prp - is instance of ProtPlot

Method Detail

setDistMode

public static int setDistMode(int distanceMode)

setDistMode() - set the method to use for clustering

Parameters:

distanceMode - is method to use

Returns:

the mode if succeed, else -1

setDistThreshold

public static float setDistThreshold(float distThr)

setDistThreshold() - set the distance threshold to use for clustering

Parameters:

distThr - is the distance threshold to use if > 0.0

Returns:

the threshold if succeed, else set threshold to MAX_DISTANCE_SQ

updateSeedProtein

public static boolean updateSeedProtein(int newSeedMPID)

updateSeedProtein() - update the seed protein mPid used for clustering and recluster if enabled

Parameters:

newSeedMPID - is the new seed mPid protein

Returns:

true if suceed

setClusterMethod

public static float setClusterMethod(int clustMethod)

setClusterMethod() - set method to use for clustering

Parameters:

clustMethod - is the cluster method to use including:

Returns:

the clusterMethod if succeed, else -1

cluster

public boolean cluster()

cluster() - cluster if allowClusterReportUpdatesFlag enabled using the current clusterMethod.

calcDistIJ

public float calcDistIJ(int mPidI,
                        int mPidJ)

calcDistIJ() - compute the distance between midI and midJ using distMode method.

Parameters:

mPidI - is protein mPid # i

mPidJ - is protein mPid # j

Returns:

distance computed as above.

calcMeanDistIJ

public void calcMeanDistIJ()

calcMeanDistIJ() - compute mean distances for all samples

                         n   i-1
 dMean[s]= (2/n(n-1)) * Sum  Sum | esi-esj | 
                        i=1  j=0

 See A. K. Jain, R.C. Dubes, Algorithms for Clustering Data, Prentice Hall,
 NJ, 1988. Section 2.2.3 Missing Data, page 20 for discussion of this 
 algorithm.

distMissingIJ

public float distMissingIJ(int s,
                           int mPidI,
                           int mPidJ)

distMissingIJ() - compute sample s distance between Esi and Esj for mPidI and mPidJ.

 If Esi or Esj is missing (i.e. 0.0) , then return  -1.0
 else return |Esi - Esj|.

 See A. K. Jain, R.C. Dubes, Algorithms for Clustering Data, Prentice Hall,
 NJ, 1988. Section 2.2.3 Missing Data, page 19 for (method 3) discussion of
 this algorithm.

Parameters:

s - is ith sample

mPidI - is protein mPid # i

mPidJ - is protein mPid # j

Returns:

distance computed as above, -1.0F if bad data.

distMissing

public float distMissing(int mPidI,
                         int mPidJ)

distMissing() - return average missing distance between mPidI and mPidJ.

 dist(i,j) = (nSamples/nBoth) SUM dist(s,i,j)
                              s in EP list

 See A. K. Jain, R.C. Dubes, Algorithms for Clustering Data, Prentice Hall,
 NJ, 1988. Section 2.2.3 Missing Data, page 20 (method 3) for discussion of
 this algorithm.

Parameters:

mPidI - is protein mPid # i

mPidJ - is protein mPid # j

Returns:

distance computed as above.

distMissingMeanIJ

public float distMissingMeanIJ(int s,
                               int mPidI,
                               int mPidJ)

distMissingMeanIJ() - compute sample s distance between Esi and Esj for mPidI and mPidJ.

 If Esi or Esj is missing (i.e. 0.0) , then return dMean
 else return (Esi - Esj).

 See A. K. Jain, R.C. Dubes, Algorithms for Clustering Data, Prentice Hall,
 NJ, 1988. Section 2.2.3 Missing Data, page 20 (method 4) for discussion 
 of this algorithm.

Parameters:

s - is ith sample

mPidI - is protein mPid # i

mPidJ - is protein mPid # j

Returns:

distance computed as above.

distMissingMean

public float distMissingMean(int mPidI,
                             int mPidJ)

distMissingMean() - return average missing distance between mPidI and mPidJ.

 dist(i,j) = (nSamples/nBoth) SUM distMean(s,dMean,i,j)
                              s in EP list

 See A. K. Jain, R.C. Dubes, Algorithms for Clustering Data, Prentice Hall,
 NJ, 1988. Section 2.2.3 Missing Data, page 20 (method 4) for discussion of
 this algorithm.

Parameters:

mPidI - is protein mPid # i

mPidJ - is protein mPid # j

Returns:

distance computed as above.

distFisherLowValues

public float distFisherLowValues(int mPidI,
                                 int mPidJ)

distFisherLowValues() - compute the Fisher clustering low values distance metric by mapping low or missing values counts to a low value (i.e. a count of 1). [TODO] Note this requires the actual counts data (not available as of V0.29.2).

Parameters:

mPidI - is protein mPid # i

mPidJ - is protein mPid # j

Returns:

distance computed as above.

isSampleInExprProfileList

public boolean isSampleInExprProfileList(int s)

isSampleInExprProfileList() - test if sample s is in current Expr Profile list of samples

Parameters:

s - sample to test

Returns:

true if s is in the current expr profile sample list.

clusterKMeansProteins

public boolean clusterKMeansProteins(boolean popupClusterWindowFlag)

clusterKMeansProteins() - cluster proteins passing the filter similar to the current protein. Cluster all filtered genes, prp.mPidFiltered[0:prp.nMpidFiltered-1], by distance from curMPID. If successful, it builds list prp.mPidSimilarCluster[0:prp.nMpidSimilarCluster-1].

Returns:

true if succeed
[TODO] see clusterKMeansProteins() for notes

clusterKMeansProteins

public boolean clusterKMeansProteins()

clusterKMeansProteins() - cluster proteins passing the K-means filter. Cluster all filtered genes, prp.mPidFiltered[0:prp.nMpidFiltered-1], by K-means for the number of clusters specified by a slider. If successful, it builds list prp.mPidSimilarCluster[0:prp.nMpidSimilarCluster-1].

Returns:

true if succeed
[TODO] 1) implement # of clusters slider, 2) implement K-means, etc. 3) need to add data structure to support cluster identification

clusterSimilarProteins

public boolean clusterSimilarProteins(boolean popupClusterWindowFlag)

clusterSimilarProteins() - cluster proteins passing the filter similar to the current protein. Cluster all filtered genes, prp.mPidFiltered[0:prp.nMpidFiltered-1], by distance from curMPID. If successful, it builds list prp.mPidSimilarCluster[0:prp.nMpidSimilarCluster-1].

Returns:

true if succeed

clusterSimilarProteins

public boolean clusterSimilarProteins()

clusterSimilarProteins() - cluster proteins passing the filter similar to the current protein. Cluster all filtered genes, prp.mPidFiltered[0:prp.nMpidFiltered-1], by distance from curMPID. If successful, it builds list prp.mPidSimilarCluster[0:prp.nMpidSimilarCluster-1].

Returns:

true if succeed

updateSimilarProteinsReport

public void updateSimilarProteinsReport()

updateSimilarProteinsReport() - compute report string of similar-cluster of filtered proteins [TODO] This will update the Cluster Report window text area. Until that is finished, it will generate a new similar cluster report window.

similarProteinsReportStr

public java.lang.String similarProteinsReportStr()

similarProteinsReportStr() - compute report string of similar-cluster of filtered proteins

Returns:

report string if succeed, else null

saveCluster

public void saveCluster()

saveCluster() - save current clustered proteins into prp.savedClusterMPIDs[]