public List<DoubleList<Split, Integer>> findConflictingSplitCounts(
int threshold, boolean thresholdIsLength) {
findGreedySplits();
List<DoubleList<Split, Integer>> results =
new Vector<DoubleList<Split, Integer>>(greedySplits.size());
// could be more efficient by storing where the greedy splits appear in the sorted list.
for (int i : greedySplitIndex) {
String splitStr = sortedSplits.elementAt(i);
Split split = splits.get(splitStr);
if (!thresholdIsLength && counts.get(splitStr) < threshold)
break; // ignore splits with frequency below threshold
DoubleList<Split, Integer> splitList = new DoubleList<Split, Integer>();
splitList.add(split, counts.get(splitStr));
// Only check splits after this one in sorted list: ones before this one are guaranteed to be
// compatible, else this split would not be in the greedy list.
for (int j = i + 1; j < sortedSplits.size(); j++) {
String otherSplitString = sortedSplits.elementAt(j);
if (!thresholdIsLength && counts.get(otherSplitString) < threshold)
break; // ignore conflicting splits with frequency below threshold
Split otherSplit = splits.get(otherSplitString);
if (!split.compatible(otherSplit)) {
splitList.add(otherSplit, counts.get(otherSplitString));
} // if !compatible
if (thresholdIsLength && splitList.size() == threshold)
break; // have enough secondary splits now
} // for otherSplit (j)
results.add(splitList);
} // for i over sortedSplits
return results;
}
/*
* Return a twin list of the greedy consensus tree splits and their internode certainties
*/
public DoubleList<Split, Double> getICs() {
DoubleList<Split, Double> ic = new DoubleList<Split, Double>();
List<DoubleList<Split, Integer>> greedyTreeConflicts = findConflictingSplitCounts(2, true);
for (DoubleList<Split, Integer> splitList : greedyTreeConflicts) {
ic.add(splitList.getA(0), internodeCertainty(splitList));
}
return ic;
}
/**
* Returns the sum of Internode Certainty All over a greedy consensus tree.
*
* @return
*/
public double treeCertaintyAll(int threshold) {
double tca = 0;
DoubleList<Split, Double> ic = getICAs(threshold);
for (int i = 0; i < ic.size(); i++) {
tca += ic.getB(i);
}
return tca;
}
/**
* Returns the sum of Internode Certainties over a greedy consensus tree.
*
* @return
*/
public double treeCertainty() {
double tc = 0;
DoubleList<Split, Double> ic = getICs();
for (int i = 0; i < ic.size(); i++) {
tc += ic.getB(i);
}
return tc;
}
/*
* Return a twin list of the greedy consensus tree splits and their ICA (internode certainty all)
* scores, with a threshold for which incompatible splits are included in the ICA calculation
*/
public DoubleList<Split, Double> getICAs(int threshold) {
DoubleList<Split, Double> ica = new DoubleList<Split, Double>();
List<DoubleList<Split, Integer>> greedyTreeConflicts =
findConflictingSplitCounts(threshold, false);
for (DoubleList<Split, Integer> splitList : greedyTreeConflicts) {
ica.add(splitList.getA(0), internodeCertainty(splitList));
}
return ica;
}
/*
* Salichos Stamatakis and Rokas, MBE v31 p1261 (2014)
* If only two splits are listed, returns IC, the Internode Certainty.
* If more splits are listed, returns ICA, (IC All).
*/
private double internodeCertainty(DoubleList<Split, Integer> splits) {
int sum = 0;
int n = splits.size(); // number of splits under consideration
for (int i = 0; i < n; i++) sum += splits.getB(i);
double ic = 1;
if (n > 1) {
ic = Math.log(n);
for (int i = 0; i < n; i++) {
double p = ((double) splits.getB(i)) / sum;
ic += p * Math.log(p);
}
ic /= Math.log(n); // convert from natural log to log base n
}
return ic;
}
public void printInternodeCertainties(PrintWriter out) {
DoubleList<Split, Double> ic = getICs();
for (int i = 0; i < ic.size(); i++) {
out.printf("IC = %f for split %s\n", ic.getB(i), ic.getA(i).toString());
}
}
