long AssemblyJobSsDeBruijn::combineTreesMidHelp(SeqNode* topNodeA, int kmerSize,
SeqNode* branchA, SeqNode* branchB, int remainingKmer){
char nucs[] = {'A','T','C','G'};
long newNodeSum = 0;
if (remainingKmer == 1){
for (int n = 0; n < 4; ++n){
DeBruijnNode* newNodeB = dynamic_cast<DeBruijnNode*>(branchB->getBranch(nucs[n]));
if (newNodeB != 0){
DeBruijnNode* newNodeA = dynamic_cast<DeBruijnNode*>(branchA->getBranch(nucs[n]));
if (newNodeA == 0){
newNodeA = new DeBruijnNode(branchA,nucs[n]);
branchA->addBranch( newNodeA );
newNodeSum += 1;
}
newNodeA->addKmerScore( newNodeB->getKmerScore() );
newNodeSum += combineTreesBottomHelp(topNodeA, kmerSize, newNodeA, newNodeB);
}
}
} else {
for (int n = 0; n < 4; ++n){
SeqNode* newNodeB = branchB->getBranch(nucs[n]);
if (newNodeB != 0){
SeqNode* newNodeA = branchA->getBranch(nucs[n]);
if (newNodeA == 0){
newNodeA = new NucNode(branchA,nucs[n]);
branchA->addBranch( newNodeA );
}
newNodeSum += combineTreesMidHelp(topNodeA, kmerSize, newNodeA, newNodeB, remainingKmer-1);
}
}
}
return newNodeSum;
}
pair<AssemblyJobSsDeBruijn::NucNode*,long> AssemblyJobSsDeBruijn::makeGraphTree(set<ScoredSeq*>* seqs, int kmerSize){
NucNode* topNode = new NucNode(0, '\0');
long numBaseNodes = 0;
if (kmerSize > 0){
for (set<ScoredSeq*>::iterator seqIt = seqs->begin(); seqIt != seqs->end(); ++seqIt){
ScoredSeq* seq = *seqIt;
// these are the nodes for which the current nuc is at the Nth position in the kmer;
// the elements earlier in the array represent kmers that come later in the sequence
SeqNode* currentNodes[kmerSize-1];
float worstScores[kmerSize-1];
// for making the de Bruijn graph at the tips
DeBruijnNode* finishedNode = 0;
DeBruijnNode* priorNode = 0;
// used to determine how far along the kmer construction has made it
long maxCurrentNodeIndex = 0;
bool createFinishedNode = false;
char* nucs = seq->getSeq('+');
float* scores = seq->getScores('+');
float* links = seq->getLinks('+');
long seqSize = seq->size();
long seqSizeM1 = seqSize - 1;
for (long pos = 0; pos < seqSize; ++pos){
char nuc = nucs[pos];
float nucScore = scores[pos];
// will only be used up to the second-to-last position
float linkScore;
if (pos < seqSizeM1){ linkScore = links[pos]; }
else { linkScore = 0; }
// re-set the derivation of kmers if the nucleotide is an N
if (nuc == 'N'){
maxCurrentNodeIndex = 0;
createFinishedNode = false;
finishedNode = 0;
priorNode = 0;
} else {
priorNode = finishedNode;
if (createFinishedNode){
SeqNode* oldNode = currentNodes[maxCurrentNodeIndex];
float worstScore = worstScores[maxCurrentNodeIndex];
finishedNode = dynamic_cast<DeBruijnNode*>( oldNode->getBranch(nuc) );
if (finishedNode == 0){
finishedNode = new DeBruijnNode(oldNode, nuc);
oldNode->addBranch(finishedNode);
numBaseNodes++;
}
// add the kmer score to the node; the worst score possible!
if ( nucScore < worstScore ){ worstScore = nucScore; }
finishedNode->addKmerScore(worstScore);
// add the 3p-directed link score
if (priorNode != 0){
float link3pScore = links[pos-1];
float link5pScore = links[pos-kmerSize];
float worseLinkScore = link3pScore;
if (link5pScore < worseLinkScore){ worseLinkScore = link5pScore; }
priorNode->add3pLink(finishedNode,worseLinkScore);
finishedNode->add5pLink(priorNode,worseLinkScore);
}
}
for (int n = maxCurrentNodeIndex; n >= 0; --n){
SeqNode* oldNode;
float worstScore;
if (n==0){
oldNode = topNode;
worstScore = nucScore;
}
else {
oldNode = currentNodes[n-1];
worstScore = worstScores[n-1];
if ( nucScore < worstScore ){ worstScore = nucScore; }
}
if ( linkScore < worstScore ){ worstScore = linkScore; }
SeqNode* newNode = oldNode->getBranch(nuc);
if (newNode == 0){
newNode = new NucNode(oldNode,nuc);
oldNode->addBranch( newNode );
}
currentNodes[n] = newNode;
worstScores[n] = worstScore;
}
if (! createFinishedNode){
if (maxCurrentNodeIndex < kmerSize-2){ maxCurrentNodeIndex++; }
else { createFinishedNode = true; }
}
}
}
delete [] nucs;
delete [] scores;
delete [] links;
}
}
return pair<NucNode*,long>(topNode,numBaseNodes);
}
