public PositionWeightMatrix permuteColumns(boolean preserveGCDinucleotides) {
PositionWeightMatrix original = copy();
List<Integer> dinucleotidesToPreserveFirstIdx =
preserveGCDinucleotides ? original.gcDinucleotidesFistIdxs() : new ArrayList<Integer>();
PositionWeightMatrix permutted = new PositionWeightMatrix(getName() + "_perm");
Random r = new Random();
List<Integer> idxList = new ArrayList<Integer>(original.size());
for (int i = 0; i < original.size(); i++) {
idxList.add(i);
}
while (idxList.size() > 0) {
int idxOfIdx = r.nextInt(idxList.size());
int idx = idxList.remove(idxOfIdx);
if (dinucleotidesToPreserveFirstIdx.contains(idx - 1)) {
PositionWeightColumn col = original.get(idx - 1);
permutted.add(col);
PositionWeightColumn nextCol = original.get(idx);
idxList.remove(idxOfIdx - 1);
permutted.add(nextCol);
} else {
PositionWeightColumn col = original.get(idx);
permutted.add(col);
if (dinucleotidesToPreserveFirstIdx.contains(idx)) {
PositionWeightColumn nextCol = original.get(idx + 1);
idxList.remove(idxOfIdx);
permutted.add(nextCol);
}
}
}
return permutted;
}
public double kullbackLeiber(PositionWeightMatrix other) {
double kl = 0;
for (int i = 0; i < other.getNumCol(); i++) {
kl += get(i).kullbackLeiber(other.get(i));
}
return kl;
}
/**
* Computes the euclidean centroid, it may not be the formal centroid for different metrics but
* intuitively the average counts should provide a good cluster representative which is what this
* method intends to return
*
* @param pwmSet - Collection of pwms from which to compute the centroid0
* @return The euclidean centroid
* @throws IllegalArgumentException - When not all PWMs have the same dimension.
*/
public PositionWeightMatrix centroidOf(Collection<PositionWeightMatrix> pwmSet)
throws IllegalArgumentException {
Matrix centroidMatrix = null;
Iterator<PositionWeightMatrix> pwmIt = pwmSet.iterator();
if (pwmIt.hasNext()) {
PositionWeightMatrix first = pwmIt.next();
PositionWeightColumn firstCol = first.get(0);
centroidMatrix = new Matrix(firstCol.getAlphabetSize(), first.getNumCol());
addToCentroid(centroidMatrix, first);
}
while (pwmIt.hasNext()) {
PositionWeightMatrix pwm = pwmIt.next();
if (pwm.getNumCol() != centroidMatrix.getColumnDimension()) {
throw new IllegalArgumentException(
"Error computing centroid. All PWMs in set should have the same dimension");
}
addToCentroid(centroidMatrix, pwm);
}
centroidMatrix.times(1 / (double) pwmSet.size());
PositionWeightMatrix centroid = new PositionWeightMatrix("centroid");
for (int j = 0; j < centroidMatrix.getColumnDimension(); j++) {
centroid.addColumn(centroidMatrix.getColumn(j));
}
return centroid;
}
protected void addToCentroid(Matrix m, PositionWeightMatrix pwm) {
for (int i = 0; i < m.getColumnDimension(); i++) {
PositionWeightColumn c = pwm.get(i);
for (int j = 0; j < m.getRowDimension(); j++) {
m.set(j, i, m.get(j, i) + c.getWeight(j));
}
}
}