void AlignSubFam(SeqVect &vAll, const Tree &GuideTree, unsigned uNodeIndex,
MSA &msaOut)
{
const unsigned uSeqCount = vAll.GetSeqCount();
const char *InTmp = "asf_in.tmp";
const char *OutTmp = "asf_out.tmp";
unsigned *Leaves = new unsigned[uSeqCount];
unsigned uLeafCount;
GetLeaves(GuideTree, uNodeIndex, Leaves, &uLeafCount);
SeqVect v;
for (unsigned i = 0; i < uLeafCount; ++i)
{
unsigned uLeafNodeIndex = Leaves[i];
unsigned uId = GuideTree.GetLeafId(uLeafNodeIndex);
Seq &s = vAll.GetSeqById(uId);
v.AppendSeq(s);
}
#if TRACE
{
Log("Align subfam[node=%d, size=%d] ", uNodeIndex, uLeafCount);
for (unsigned i = 0; i < uLeafCount; ++i)
Log(" %s", v.GetSeqName(i));
Log("\n");
}
#endif
TextFile fIn(InTmp, true);
v.ToFASTAFile(fIn);
fIn.Close();
char CmdLine[4096];
sprintf(CmdLine, "probcons %s > %s 2> /dev/null", InTmp, OutTmp);
// sprintf(CmdLine, "muscle -in %s -out %s -maxiters 1", InTmp, OutTmp);
system(CmdLine);
TextFile fOut(OutTmp);
msaOut.FromFile(fOut);
for (unsigned uSeqIndex = 0; uSeqIndex < uLeafCount; ++uSeqIndex)
{
const char *Name = msaOut.GetSeqName(uSeqIndex);
unsigned uId = vAll.GetSeqIdFromName(Name);
msaOut.SetSeqId(uSeqIndex, uId);
}
unlink(InTmp);
unlink(OutTmp);
delete[] Leaves;
}
static void LogLeafNames(const Tree &tree, unsigned uNodeIndex)
{
const unsigned uNodeCount = tree.GetNodeCount();
unsigned *Leaves = new unsigned[uNodeCount];
unsigned uLeafCount;
GetLeaves(tree, uNodeIndex, Leaves, &uLeafCount);
for (unsigned i = 0; i < uLeafCount; ++i)
{
if (i > 0)
Log(",");
Log("%s", tree.GetLeafName(Leaves[i]));
}
delete[] Leaves;
}
static void LogSubfams(const Tree &tree, const unsigned Subfams[],
unsigned uSubfamCount)
{
const unsigned uNodeCount = tree.GetNodeCount();
Log("%u subfamilies found\n", uSubfamCount);
Log("Subfam Sequence\n");
Log("------ --------\n");
unsigned *Leaves = new unsigned[uNodeCount];
for (unsigned uSubfamIndex = 0; uSubfamIndex < uSubfamCount; ++uSubfamIndex)
{
unsigned uSubfamNodeIndex = Subfams[uSubfamIndex];
unsigned uLeafCount;
GetLeaves(tree, uSubfamNodeIndex, Leaves, &uLeafCount);
for (unsigned uLeafIndex = 0; uLeafIndex < uLeafCount; ++uLeafIndex)
Log("%6u %s\n", uSubfamIndex + 1, tree.GetLeafName(Leaves[uLeafIndex]));
Log("\n");
}
delete[] Leaves;
}
static void BuildDiffs(const Tree &tree, unsigned uTreeNodeIndex,
const bool bIsDiff[], Tree &Diffs, unsigned uDiffsNodeIndex,
unsigned IdToDiffsLeafNodeIndex[])
{
#if TRACE
Log("BuildDiffs(TreeNode=%u IsDiff=%d IsLeaf=%d)\n",
uTreeNodeIndex, bIsDiff[uTreeNodeIndex], tree.IsLeaf(uTreeNodeIndex));
#endif
if (bIsDiff[uTreeNodeIndex])
{
unsigned uLeafCount = tree.GetLeafCount();
unsigned *Leaves = new unsigned[uLeafCount];
GetLeaves(tree, uTreeNodeIndex, Leaves, &uLeafCount);
for (unsigned n = 0; n < uLeafCount; ++n)
{
const unsigned uLeafNodeIndex = Leaves[n];
const unsigned uId = tree.GetLeafId(uLeafNodeIndex);
if (uId >= tree.GetLeafCount())
Quit("BuildDiffs, id out of range");
IdToDiffsLeafNodeIndex[uId] = uDiffsNodeIndex;
#if TRACE
Log(" Leaf id=%u DiffsNode=%u\n", uId, uDiffsNodeIndex);
#endif
}
delete[] Leaves;
return;
}
if (tree.IsLeaf(uTreeNodeIndex))
Quit("BuildDiffs: should never reach leaf");
const unsigned uTreeLeft = tree.GetLeft(uTreeNodeIndex);
const unsigned uTreeRight = tree.GetRight(uTreeNodeIndex);
const unsigned uDiffsLeft = Diffs.AppendBranch(uDiffsNodeIndex);
const unsigned uDiffsRight = uDiffsLeft + 1;
BuildDiffs(tree, uTreeLeft, bIsDiff, Diffs, uDiffsLeft, IdToDiffsLeafNodeIndex);
BuildDiffs(tree, uTreeRight, bIsDiff, Diffs, uDiffsRight, IdToDiffsLeafNodeIndex);
}
bool RefineSubfams(MSA &msa, const Tree &tree, unsigned uIters)
{
MuscleContext *ctx = getMuscleContext();
CLUSTER &g_Cluster2 = ctx->params.g_Cluster2;
DISTANCE &g_Distance2 = ctx->params.g_Distance2;
ROOT &g_Root2 = ctx->params.g_Root2;
bool &g_bAnchors = ctx->params.g_bAnchors;
const unsigned uSeqCount = msa.GetSeqCount();
if (uSeqCount < 3)
return false;
const double dMaxHeight = 0.6;
const unsigned uMaxSubfamCount = 16;
//const unsigned uNodeCount = tree.GetNodeCount();
unsigned *Subfams;
unsigned uSubfamCount;
GetSubfams(tree, dMaxHeight, uMaxSubfamCount, &Subfams, &uSubfamCount);
assert(uSubfamCount <= uSeqCount);
if (ctx->params.g_bVerbose)
LogSubfams(tree, Subfams, uSubfamCount);
MSA *SubfamMSAs = new MSA[uSubfamCount];
unsigned *Leaves = new unsigned[uSeqCount];
unsigned *Ids = new unsigned[uSeqCount];
bool bAnyChanges = false;
for (unsigned uSubfamIndex = 0; uSubfamIndex < uSubfamCount; ++uSubfamIndex)
{
unsigned uSubfam = Subfams[uSubfamIndex];
unsigned uLeafCount;
GetLeaves(tree, uSubfam, Leaves, &uLeafCount);
assert(uLeafCount <= uSeqCount);
LeafIndexesToIds(tree, Leaves, uLeafCount, Ids);
MSA &msaSubfam = SubfamMSAs[uSubfamIndex];
MSASubsetByIds(msa, Ids, uLeafCount, msaSubfam);
DeleteGappedCols(msaSubfam);
#if TRACE
Log("Subfam %u MSA=\n", uSubfamIndex);
msaSubfam.LogMe();
#endif
if (msaSubfam.GetSeqCount() <= 2)
continue;
// TODO /////////////////////////////////////////
// Try using existing tree, may actually hurt to
// re-estimate, may also be a waste of CPU & mem.
/////////////////////////////////////////////////
Tree SubfamTree;
TreeFromMSA(msaSubfam, SubfamTree, g_Cluster2, g_Distance2, g_Root2);
bool bAnyChangesThisSubfam;
if (g_bAnchors)
bAnyChangesThisSubfam = RefineVert(msaSubfam, SubfamTree, uIters);
else
bAnyChangesThisSubfam = RefineHoriz(msaSubfam, SubfamTree, uIters, false, false);
#if TRACE
Log("Subfam %u Changed %d\n", uSubfamIndex, bAnyChangesThisSubfam);
#endif
if (bAnyChangesThisSubfam)
bAnyChanges = true;
}
if (bAnyChanges)
ProgressiveAlignSubfams(tree, Subfams, uSubfamCount, SubfamMSAs, msa);
delete[] Leaves;
delete[] Subfams;
delete[] SubfamMSAs;
return bAnyChanges;
}
void RefineWorker::_run() {
unsigned i = 0;
#if TRACE
algoLog.trace(QString("Worker %1 start. Wait...").arg(QString::number(workerID)));
#endif
workpool->mainSem.acquire();
#if TRACE
algoLog.trace(QString("Worker %1: Stop wait. Start (mainSem %2, childSem %3)").arg(QString::number(workerID)).
arg(QString::number(workpool->mainSem.available())).arg(QString::number(workpool->mainSem.available())));
#endif
while(!workpool->isRefineDone())
{
MSA msaIn;
i = workpool->refineGetJob(&msaIn, workerID);
MuscleContext *ctx = workpool->ctx;
// unsigned &g_uTreeSplitNode1 = ctx->muscle.g_uTreeSplitNode1;
// unsigned &g_uTreeSplitNode2 = ctx->muscle.g_uTreeSplitNode2;
// unsigned &g_uRefineHeightSubtree = ctx->refinehoriz.g_uRefineHeightSubtree;
// unsigned &g_uRefineHeightSubtreeTotal = ctx->refinehoriz.g_uRefineHeightSubtreeTotal;
Tree &tree = workpool->GuideTree;
const unsigned uSeqCount = msaIn.GetSeqCount();
// const unsigned uInternalNodeCount = uSeqCount - 1;
unsigned *Leaves1 = new unsigned[uSeqCount];
unsigned *Leaves2 = new unsigned[uSeqCount];
const unsigned uRootNodeIndex = tree.GetRootNodeIndex();
while (i != NULL_NEIGHBOR) {
const unsigned uInternalNodeIndex = workpool->InternalNodeIndexes[i];
unsigned uNeighborNodeIndex;
if (tree.IsRoot(uInternalNodeIndex) && !workpool->bRight) {
i = workpool->refineGetNextJob(&msaIn, false, -1, i, workerID);
continue;
}
else if (workpool->bRight)
uNeighborNodeIndex = tree.GetRight(uInternalNodeIndex);
else
uNeighborNodeIndex = tree.GetLeft(uInternalNodeIndex);
// g_uTreeSplitNode1 = uInternalNodeIndex;
// g_uTreeSplitNode2 = uNeighborNodeIndex;
unsigned uCount1;
unsigned uCount2;
GetLeaves(tree, uNeighborNodeIndex, Leaves1, &uCount1);
GetLeavesExcluding(tree, uRootNodeIndex, uNeighborNodeIndex,
Leaves2, &uCount2);
SCORE scoreBefore;
SCORE scoreAfter;
bool bAccepted = TryRealign(msaIn, tree, Leaves1, uCount1, Leaves2, uCount2,
&scoreBefore, &scoreAfter, workpool->bLockLeft, workpool->bLockRight);
SCORE scoreMax = scoreAfter > scoreBefore? scoreAfter : scoreBefore;
//bool bRepeated = workpool->History->SetScore(workpool->uIter, uInternalNodeIndex, workpool->bRight, scoreMax);
i = workpool->refineGetNextJob(&msaIn, bAccepted, scoreMax, i, workerID);
}
delete[] Leaves1;
delete[] Leaves2;
#if TRACE
algoLog.trace(QString("Worker %1: no job available. Wait... (mainSem %2, childSem %3)").arg(QString::number(workerID)).
arg(QString::number(workpool->mainSem.available())).arg(QString::number(workpool->mainSem.available())));
#endif
workpool->childSem.release();
workpool->mainSem.acquire();
#if TRACE
algoLog.trace(QString("Worker %1: Stop wait. Start (mainSem %2, childSem %3)").arg(QString::number(workerID)).
arg(QString::number(workpool->mainSem.available())).arg(QString::number(workpool->mainSem.available())));
#endif
}
#if TRACE
algoLog.trace(QString("Worker %1: Refine done. Exit").arg(QString::number(workerID)));
#endif
}
void CvGBTrees::change_values(CvDTree* tree, const int _k)
{
CvDTreeNode** predictions = new pCvDTreeNode[get_len(subsample_train)];
int* sample_data = sample_idx->data.i;
int* subsample_data = subsample_train->data.i;
int s_step = (sample_idx->cols > sample_idx->rows) ? 1
: sample_idx->step/CV_ELEM_SIZE(sample_idx->type);
CvMat x;
CvMat miss_x;
for (int i=0; i<get_len(subsample_train); ++i)
{
int idx = *(sample_data + subsample_data[i]*s_step);
if (data->tflag == CV_ROW_SAMPLE)
cvGetRow( data->train_data, &x, idx);
else
cvGetCol( data->train_data, &x, idx);
if (missing)
{
if (data->tflag == CV_ROW_SAMPLE)
cvGetRow( missing, &miss_x, idx);
else
cvGetCol( missing, &miss_x, idx);
predictions[i] = tree->predict(&x, &miss_x);
}
else
predictions[i] = tree->predict(&x);
}
CvDTreeNode** leaves;
int leaves_count = 0;
leaves = GetLeaves( tree, leaves_count);
for (int i=0; i<leaves_count; ++i)
{
int samples_in_leaf = 0;
for (int j=0; j<get_len(subsample_train); ++j)
{
if (leaves[i] == predictions[j]) samples_in_leaf++;
}
if (!samples_in_leaf) // It should not be done anyways! but...
{
leaves[i]->value = 0.0;
continue;
}
CvMat* leaf_idx = cvCreateMat(1, samples_in_leaf, CV_32S);
int* leaf_idx_data = leaf_idx->data.i;
for (int j=0; j<get_len(subsample_train); ++j)
{
int idx = *(sample_data + subsample_data[j]*s_step);
if (leaves[i] == predictions[j])
*leaf_idx_data++ = idx;
}
float value = find_optimal_value(leaf_idx);
leaves[i]->value = value;
leaf_idx_data = leaf_idx->data.i;
int len = sum_response_tmp->cols;
for (int j=0; j<get_len(leaf_idx); ++j)
{
int idx = leaf_idx_data[j];
sum_response_tmp->data.fl[idx + _k*len] =
sum_response->data.fl[idx + _k*len] +
params.shrinkage * value;
}
leaf_idx_data = 0;
cvReleaseMat(&leaf_idx);
}
// releasing the memory
for (int i=0; i<get_len(subsample_train); ++i)
{
predictions[i] = 0;
}
delete[] predictions;
for (int i=0; i<leaves_count; ++i)
{
leaves[i] = 0;
}
delete[] leaves;
}
