Skip to content

Commit

Permalink
Too many changes to count
Browse files Browse the repository at this point in the history
  • Loading branch information
@vineet1992
vineet1992 committed Jun 12, 2019
1 parent 4687529 commit e798e3e
Show file tree
Hide file tree
Showing 13 changed files with 650 additions and 61 deletions.
34 changes: 23 additions & 11 deletions tetrad-lib/src/main/java/edu/pitt/csb/Olja_Cancer_Analysis/PriorPrediction.java
View file
Original file line number Diff line number Diff line change
Expand Up @@ -228,7 +228,7 @@ else if(args[index].equals("-rv"))

DataSet predictionData = data;

if(steps)
if(steps && !prefDiv)
{
System.out.print("Running STEPS...");
STEPS s = new STEPS(temp,initLambdas,g,samps);
Expand All @@ -249,7 +249,7 @@ else if(prefDiv)
printer.println(test);
printer.flush();
printer.close();
String cmd = "java -jar PrefDiv.jar -data temp_train.txt -dataTest temp_test.txt -loocv " + " -t " + target + " -name temp -numSelect " + numSelect + " -keep ";
String cmd = "java -jar PrefDiv.jar -data temp_train.txt -dataTest temp_test.txt -loocv " + "-t " + target + " -name temp_" + i + " -numSelect " + numSelect + " -keep ";

for(String s:toRemove)
{
Expand Down Expand Up @@ -299,12 +299,20 @@ else if(ctype== RunPrefDiv.ClusterType.MEDIAN)
System.err.println(s);
}

out = GraphUtils.loadGraphTxt(new File("temp/graph.txt"));
if(steps)
{
out = GraphUtils.loadGraphTxt(new File("temp_" + i + "/GRAPH_temp_train.txt"));
}
else
{
out = GraphUtils.loadGraphTxt(new File("temp_" + i + "/piMGM/Graph.txt"));
}
/***Construct training and testing set by getting list of genes and clusters from files***/

DataSet train = MixedUtils.loadDataSet2("temp/summarized.txt");
test = MixedUtils.loadDataSet2("temp/summarized_test.txt");
DataSet full = DataUtils.concatenate(train,test);
DataSet train = MixedUtils.loadDataSet2("temp_" + i + "/OUT_temp_train.txt");
test = MixedUtils.loadTestData("temp_" + i + "/summarized_test.txt",3,train);
DataSet full = train.copy();
MixedUtils.concatenateTo(test,full);

predictionData = full;
temp = train;
Expand All @@ -330,22 +338,26 @@ else if(ctype== RunPrefDiv.ClusterType.MEDIAN)
Graph cpcOut = cpc.search();


int testIndex = i;

if(prefDiv)
testIndex = predictionData.getNumRows()-1;

/***piMGM Predictions***/

List<Node> neighbors = out.getAdjacentNodes(out.getNode(target));
System.out.println("Adjacent To Target: " + neighbors);
double pred = getRegressionResult(neighbors,temp,predictionData,i,target);
double pred = getRegressionResult(neighbors,temp,predictionData,testIndex,target);
double real = -1;



if(predictionData.getVariable(target)instanceof DiscreteVariable)
{
real = predictionData.getInt(i,predictionData.getColumn(predictionData.getVariable(target)));
real = predictionData.getInt(testIndex,predictionData.getColumn(predictionData.getVariable(target)));
}else
{
real = predictionData.getDouble(i,predictionData.getColumn(predictionData.getVariable(target)));
real = predictionData.getDouble(testIndex,predictionData.getColumn(predictionData.getVariable(target)));
}
System.out.println("piMGM Prediction: " + pred);
pi1.print(i + "\t" + pred +"\t");
Expand All @@ -356,7 +368,7 @@ else if(ctype== RunPrefDiv.ClusterType.MEDIAN)
if(neighbors.size()==0) {
neighbors = out.getAdjacentNodes(out.getNode(target));
}
pred = getRegressionResult(neighbors,temp,predictionData,i,target);
pred = getRegressionResult(neighbors,temp,predictionData,testIndex,target);
System.out.println("CPC-MB Neighbors: " + neighbors);
System.out.println("CPC-MB Prediction: " + pred);

Expand All @@ -371,7 +383,7 @@ else if(ctype== RunPrefDiv.ClusterType.MEDIAN)
neighbors = cpcOut.getAdjacentNodes(cpcOut.getNode(target));
if(neighbors.size()==0)
neighbors = out.getAdjacentNodes(out.getNode(target));
pred = getRegressionResult(neighbors,temp,predictionData,i,target);
pred = getRegressionResult(neighbors,temp,predictionData,testIndex,target);
System.out.println("CPC Neighbors: " + neighbors);
System.out.println("CPC Prediction: " + pred);

Expand Down
142 changes: 138 additions & 4 deletions tetrad-lib/src/main/java/edu/pitt/csb/Pref_Div/Functions.java
View file
Original file line number Diff line number Diff line change
Expand Up @@ -403,6 +403,106 @@ protected void compute(){



/***
* THIS METHOD RUNS IN PARALLEL
* @param items List of Genes, only the symbol needs to be filled in
* @param d DataSet d, a dataset on which to compute correlations
* return A float [] with all of the gene-gene correlations
***/
public static float [] computeAllAssociationsPar(final ArrayList<Gene> items, final DataSet d)
{

/***Construct mapping between indices and nodes***/
final HashMap<Integer,Integer> mapping = new HashMap<Integer,Integer>();
final HashMap<Integer,Integer> categories = new HashMap<Integer,Integer>();

/***Ensure correct mapping from genes to nodes in data***/
for(int i = 0; i < items.size();i++)
{
Node temp = d.getVariable(items.get(i).symbol);
mapping.put(i,d.getColumn(temp));
if(temp instanceof DiscreteVariable)
{
DiscreteVariable dv = (DiscreteVariable)temp;
categories.put(i,dv.getNumCategories());
}
}

/***Rows are variable, columns are samples***/
final double [][] datArray = d.getDoubleData().transpose().toArray();


int total = (items.size()*(items.size()-1))/2;
final float [] corrs = new float[total];

final ForkJoinPool pool = ForkJoinPoolInstance.getInstance().getPool();

class StabilityAction extends RecursiveAction {
private int chunk;
private int from;
private int to;

public StabilityAction(int chunk, int from, int to){
this.chunk = chunk;
this.from = from;
this.to = to;
}

@Override
protected void compute(){
if (to - from <= chunk) {
for (int s = from; s < to; s++) {
//RadiiWP is the columns, RadiiNP is the rows
double[] one = datArray[mapping.get(s)];

int index = Functions.getIndex(s,s+1,items.size());
for(int i = s+1; i < items.size();i++)
{
double [] two = datArray[mapping.get(i)];

if(categories.get(s)!=null || categories.get(i)!=null)
{
//corrs[index] = (float)mixedMI(two,one,categories.get(s));
IndTestMultinomialAJ ind = new IndTestMultinomialAJ(d,0.05,true);
ind.isIndependent(d.getVariable(mapping.get(i)),d.getVariable(mapping.get(s)));
corrs[index] = (float)(1-ind.getPValue());
}
else
{
corrs[index] = (float)Math.abs(StatUtils.correlation(one,two));
}


index++;
}
}

return;
} else {
List<StabilityAction> tasks = new ArrayList<>();

final int mid = (to + from) / 2;

tasks.add(new StabilityAction(chunk, from, mid));
tasks.add(new StabilityAction(chunk, mid, to));

invokeAll(tasks);

return;
}
}
}

final int chunk = d.getNumColumns()/Runtime.getRuntime().availableProcessors();
StabilityAction sa = new StabilityAction(chunk,0, items.size());
pool.invoke(sa);


return corrs;
}




/***
*
Expand Down Expand Up @@ -452,6 +552,7 @@ private static double[] getPartialCorr(TetradMatrix c, DataSet d, int x, int y)
int total = (items.size()*(items.size()-1))/2;
float [] corrs = new float[total];
IndTestCorrelationT ind = new IndTestCorrelationT(d,0.05);
IndTestMultinomialAJ indDisc = new IndTestMultinomialAJ(d,0.05);
for(int i = 0; i < items.size();i++)
{
Node one = nodes[i];
Expand All @@ -465,6 +566,16 @@ private static double[] getPartialCorr(TetradMatrix c, DataSet d, int x, int y)
int y = mapping.get(j);
corrs[index] = (float)(-1*c.get(x,y)/(c.get(x,x)*c.get(y,y)));
}
else if(d.isMixed())
{
indDisc.isIndependent(one,two);
if(ind.getPValue()>threshold && !withPrior[index])
corrs[index] = 0;
else if(ind.getPValue()>thresholdWP && withPrior[index])
corrs[index] = 0;
else
corrs[index] = (float) (1-ind.getPValue());
}
else {
ind.isIndependent(one,two);
if(ind.getPValue()>threshold && !withPrior[index])
Expand Down Expand Up @@ -562,7 +673,15 @@ public static ArrayList<Gene> computeAllIntensities(ArrayList<Gene> g1, double a
c = new CovarianceMatrix(data).getMatrix().ginverse();
}

IndTestCorrelationT ind = new IndTestCorrelationT(data,0.05);
IndependenceTest ind;
if(data.isMixed())
{
ind = new IndTestMultinomialAJ(data,0.05,true);
}
else
{
ind = new IndTestCorrelationT(data,0.05);
}
float [] corrs = new float[g1.size()];

/***For each gene in the gene list, compute correlation***/
Expand All @@ -586,7 +705,10 @@ else if(cont) //Use Correlation
corrs[i] = (float) Math.abs(StatUtils.correlation(temp[y], temp[data.getColumn(data.getVariable(g1.get(i).symbol))]));
}
else //Use Mutual Information
corrs[i] = (float) mixedMI(temp[data.getColumn(data.getVariable(i))],temp[y],numCats);
{
ind.isIndependent(data.getVariable(i),one);
corrs[i] = (float)(1-ind.getPValue());
}
}

/****Optional: Normalize the correlations to mean 0 Gaussian Dist***/
Expand Down Expand Up @@ -694,7 +816,16 @@ public static ArrayList<Gene> computeAllIntensities(ArrayList<Gene> g1, double a
c = new CovarianceMatrix(data).getMatrix();
c = c.ginverse();
}
IndTestCorrelationT ind = new IndTestCorrelationT(data,0.05);

IndependenceTest ind;
if(data.isMixed())
{
ind = new IndTestMultinomialAJ(data,0.05);
}
else
{
ind = new IndTestCorrelationT(data,0.05);
}
float [] corrs = new float[g1.size()];
for(int i = 0; i < g1.size();i++) {
Node two = data.getVariable(g1.get(i).symbol);
Expand All @@ -714,7 +845,10 @@ else if(ind.getPValue()>thresholdWP && withPriors[i])
corrs[i] = (float) Math.abs(StatUtils.correlation(temp[y], temp[data.getColumn(data.getVariable(g1.get(i).symbol))]));
}
else //Use Mutual Information
corrs[i] = (float) mixedMI(temp[data.getColumn(data.getVariable(i))],temp[y],numCats);
{
ind.isIndependent(one,two);
corrs[i] = (float) (1-ind.getPValue());
}
}
for (int i = 0; i < g1.size(); i++) {

Expand Down
46 changes: 41 additions & 5 deletions tetrad-lib/src/main/java/edu/pitt/csb/Pref_Div/PiPrefDiv4.java
View file
Original file line number Diff line number Diff line change
Expand Up @@ -149,6 +149,8 @@ public Map<Gene,List<Gene>> getLastCluster()
public double [] getAdjustP(){return adjustP;}
public void computePValue(boolean pvals){computeP = pvals;}
public void profile(){profiling = true;}
public DataSet getData(){return data;}

public double [] getNormTao()
{
if(normTao==null)
Expand Down Expand Up @@ -212,7 +214,11 @@ private double[] constrictRange(double[]init, DataSet data,boolean threshold)


float [] corrs;
if(parallel)
if(curr.isMixed())
{
corrs = Functions.computeAllAssociationsPar(temp,curr);
}
else if(parallel)
{
corrs = Functions.computeAllCorrelationsPar(temp,curr);
}else
Expand Down Expand Up @@ -326,12 +332,21 @@ public ArrayList<Gene> selectGenes(double radiiNP)


long time = System.nanoTime();


/***Add correlation to the target variable in the intensity value and fold change fields of each gene***/
ArrayList<Gene> meanGenes = Functions.computeAllIntensities(temp,1,data,target,partialCorrs,false,false,1);


/***Order of correlation is based upon the order of the genes in the array list (needs to be the same as the data order)***/
float [] meanDis = Functions.computeAllCorrelations(meanGenes,data,partialCorrs,false,false,1);
float [] meanDis;
if(data.isMixed())
{
meanDis = Functions.computeAllAssociationsPar(meanGenes,data);
}else
{
meanDis = Functions.computeAllCorrelations(meanGenes,data,partialCorrs,false,false,1);
}

if(profiling)
{
Expand Down Expand Up @@ -413,7 +428,15 @@ public ArrayList<Gene> selectGenes(double radiiNP, double radiiWP, String [] dFi


/***Compute correlation for each pair of genes***/
float [] meanDis = Functions.computeAllCorrelations(meanGenes,data,false,1,1,dPriorsWT);
float [] meanDis;
if(data.isMixed())
{
meanDis = Functions.computeAllAssociationsPar(meanGenes,data);
}else
{
meanDis = Functions.computeAllCorrelations(meanGenes,data,false,1,1,dPriorsWT);

}


/***Shuffle and sort the list of genes based upon intensity value***/
Expand Down Expand Up @@ -1009,7 +1032,15 @@ private int[] getCounts(int i, boolean needCorrs)
{
ArrayList<Gene> temp = createGenes(data,target,true);
DataSet currData = data.subsetRows(subsamples[k]);
float [] corrs = Functions.computeAllCorrelations(temp,currData,partialCorrs,false,false,1);
float [] corrs;
if(currData.isMixed())
{
corrs = Functions.computeAllAssociationsPar(temp,currData);
}
else
{
corrs = Functions.computeAllCorrelations(temp,currData,partialCorrs,false,false,1);
}
addGeneConnections(corrs,curr,initRadii[i]);

}
Expand Down Expand Up @@ -1494,7 +1525,12 @@ private void getPhi(float[] in, int numSubs)
float [] corrs;
long time = System.nanoTime();

if(parallel)

if(currData.isMixed())
{
corrs = Functions.computeAllAssociationsPar(temp,currData);
}
else if(parallel)
{
corrs = Functions.computeAllCorrelationsPar(temp,currData);
}
Expand Down
Loading

0 comments on commit e798e3e

Please sign in to comment.