void ClusterByHeight(const Tree &tree, double dMaxHeight, unsigned Subtrees[],
unsigned *ptruSubtreeCount)
{
if (!tree.IsRooted())
Quit("ClusterByHeight: requires rooted tree");
#if TRACE
Log("ClusterByHeight, max height=%g\n", dMaxHeight);
#endif
unsigned uSubtreeCount = 0;
const unsigned uNodeCount = tree.GetNodeCount();
for (unsigned uNodeIndex = 0; uNodeIndex < uNodeCount; ++uNodeIndex)
{
if (tree.IsRoot(uNodeIndex))
continue;
unsigned uParent = tree.GetParent(uNodeIndex);
double dHeight = tree.GetNodeHeight(uNodeIndex);
double dParentHeight = tree.GetNodeHeight(uParent);
#if TRACE
Log("Node %3u Height %5.2f ParentHeight %5.2f\n",
uNodeIndex, dHeight, dParentHeight);
#endif
if (dParentHeight > dMaxHeight && dHeight <= dMaxHeight)
{
Subtrees[uSubtreeCount] = uNodeIndex;
#if TRACE
Log("Subtree[%u]=%u\n", uSubtreeCount, uNodeIndex);
#endif
++uSubtreeCount;
}
}
*ptruSubtreeCount = uSubtreeCount;
}
static unsigned SubFamRecurse(const Tree &tree, unsigned uNodeIndex, unsigned uMaxLeafCount,
unsigned SubFams[], unsigned &uSubFamCount)
{
if (tree.IsLeaf(uNodeIndex))
return 1;
unsigned uLeft = tree.GetLeft(uNodeIndex);
unsigned uRight = tree.GetRight(uNodeIndex);
unsigned uLeftCount = SubFamRecurse(tree, uLeft, uMaxLeafCount, SubFams, uSubFamCount);
unsigned uRightCount = SubFamRecurse(tree, uRight, uMaxLeafCount, SubFams, uSubFamCount);
unsigned uLeafCount = uLeftCount + uRightCount;
if (uLeftCount + uRightCount > uMaxLeafCount)
{
if (uLeftCount <= uMaxLeafCount)
SubFams[uSubFamCount++] = uLeft;
if (uRightCount <= uMaxLeafCount)
SubFams[uSubFamCount++] = uRight;
}
else if (tree.IsRoot(uNodeIndex))
{
if (uSubFamCount != 0)
Quit("Error in SubFamRecurse");
SubFams[uSubFamCount++] = uNodeIndex;
}
return uLeafCount;
}
// Return false when done
bool PhyEnumEdges(const Tree &tree, PhyEnumEdgeState &ES)
{
unsigned uNode1 = uInsane;
if (!ES.m_bInit)
{
if (tree.GetNodeCount() <= 1)
{
ES.m_uNodeIndex1 = NULL_NEIGHBOR;
ES.m_uNodeIndex2 = NULL_NEIGHBOR;
return false;
}
uNode1 = tree.FirstDepthFirstNode();
ES.m_bInit = true;
}
else
{
uNode1 = tree.NextDepthFirstNode(ES.m_uNodeIndex1);
if (NULL_NEIGHBOR == uNode1)
return false;
if (tree.IsRooted() && tree.IsRoot(uNode1))
{
uNode1 = tree.NextDepthFirstNode(uNode1);
if (NULL_NEIGHBOR == uNode1)
return false;
}
}
unsigned uNode2 = tree.GetParent(uNode1);
ES.m_uNodeIndex1 = uNode1;
ES.m_uNodeIndex2 = uNode2;
return true;
}
bool PhyEnumBiParts(const Tree &tree, PhyEnumEdgeState &ES,
unsigned Leaves1[], unsigned *ptruCount1,
unsigned Leaves2[], unsigned *ptruCount2)
{
bool bOk = PhyEnumEdges(tree, ES);
if (!bOk)
{
*ptruCount1 = 0;
*ptruCount2 = 0;
return false;
}
// Special case: in a rooted tree, both edges from the root
// give the same bipartition, so skip one of them.
if (tree.IsRooted() && tree.IsRoot(ES.m_uNodeIndex2)
&& tree.GetRight(ES.m_uNodeIndex2) == ES.m_uNodeIndex1)
{
bOk = PhyEnumEdges(tree, ES);
if (!bOk)
return false;
}
PhyGetLeaves(tree, ES.m_uNodeIndex1, ES.m_uNodeIndex2, Leaves1, ptruCount1);
PhyGetLeaves(tree, ES.m_uNodeIndex2, ES.m_uNodeIndex1, Leaves2, ptruCount2);
if (*ptruCount1 + *ptruCount2 != tree.GetLeafCount())
Quit("PhyEnumBiParts %u + %u != %u",
*ptruCount1, *ptruCount2, tree.GetLeafCount());
#if DEBUG
{
for (unsigned i = 0; i < *ptruCount1; ++i)
{
if (!tree.IsLeaf(Leaves1[i]))
Quit("PhyEnumByParts: not leaf");
for (unsigned j = 0; j < *ptruCount2; ++j)
{
if (!tree.IsLeaf(Leaves2[j]))
Quit("PhyEnumByParts: not leaf");
if (Leaves1[i] == Leaves2[j])
Quit("PhyEnumByParts: dupe");
}
}
}
#endif
return true;
}
void CalcClustalWWeights(const Tree &tree, WEIGHT Weights[])
{
#if TRACE
Log("CalcClustalWWeights\n");
tree.LogMe();
#endif
const unsigned uLeafCount = tree.GetLeafCount();
if (0 == uLeafCount)
return;
else if (1 == uLeafCount)
{
Weights[0] = (WEIGHT) 1.0;
return;
}
else if (2 == uLeafCount)
{
Weights[0] = (WEIGHT) 0.5;
Weights[1] = (WEIGHT) 0.5;
return;
}
if (!tree.IsRooted())
Quit("CalcClustalWWeights requires rooted tree");
const unsigned uNodeCount = tree.GetNodeCount();
unsigned *LeavesUnderNode = new unsigned[uNodeCount];
memset(LeavesUnderNode, 0, uNodeCount*sizeof(unsigned));
const unsigned uRootNodeIndex = tree.GetRootNodeIndex();
unsigned uLeavesUnderRoot = CountLeaves(tree, uRootNodeIndex, LeavesUnderNode);
if (uLeavesUnderRoot != uLeafCount)
Quit("WeightsFromTreee: Internal error, root count %u %u",
uLeavesUnderRoot, uLeafCount);
#if TRACE
Log("Node Leaves Length Strength\n");
Log("---- ------ -------- --------\n");
// 1234 123456 12345678 12345678
#endif
double *Strengths = new double[uNodeCount];
for (unsigned uNodeIndex = 0; uNodeIndex < uNodeCount; ++uNodeIndex)
{
if (tree.IsRoot(uNodeIndex))
{
Strengths[uNodeIndex] = 0.0;
continue;
}
const unsigned uParent = tree.GetParent(uNodeIndex);
const double dLength = tree.GetEdgeLength(uNodeIndex, uParent);
const unsigned uLeaves = LeavesUnderNode[uNodeIndex];
const double dStrength = dLength / (double) uLeaves;
Strengths[uNodeIndex] = dStrength;
#if TRACE
Log("%4u %6u %8g %8g\n", uNodeIndex, uLeaves, dLength, dStrength);
#endif
}
#if TRACE
Log("\n");
Log(" Seq Path..Weight\n");
Log("-------------------- ------------\n");
#endif
for (unsigned n = 0; n < uLeafCount; ++n)
{
const unsigned uLeafNodeIndex = tree.LeafIndexToNodeIndex(n);
#if TRACE
Log("%20.20s %4u ", tree.GetLeafName(uLeafNodeIndex), uLeafNodeIndex);
#endif
if (!tree.IsLeaf(uLeafNodeIndex))
Quit("CalcClustalWWeights: leaf");
double dWeight = 0;
unsigned uNode = uLeafNodeIndex;
while (!tree.IsRoot(uNode))
{
dWeight += Strengths[uNode];
uNode = tree.GetParent(uNode);
#if TRACE
Log("->%u(%g)", uNode, Strengths[uNode]);
#endif
}
if (dWeight < 0.0001)
{
#if TRACE
Log("zero->one");
#endif
dWeight = 1.0;
}
Weights[n] = (WEIGHT) dWeight;
#if TRACE
Log(" = %g\n", dWeight);
#endif
}
delete[] Strengths;
delete[] LeavesUnderNode;
Normalize(Weights, uLeafCount);
}
static void ProgressiveAlignSubfams(const Tree &tree, const unsigned Subfams[],
unsigned uSubfamCount, const MSA SubfamMSAs[], MSA &msa)
{
const unsigned uNodeCount = tree.GetNodeCount();
bool *Ready = new bool[uNodeCount];
MSA **MSAs = new MSA *[uNodeCount];
for (unsigned uNodeIndex = 0; uNodeIndex < uNodeCount; ++uNodeIndex)
{
Ready[uNodeIndex] = false;
MSAs[uNodeIndex] = 0;
}
for (unsigned uSubfamIndex = 0; uSubfamIndex < uSubfamCount; ++uSubfamIndex)
{
unsigned uNodeIndex = Subfams[uSubfamIndex];
Ready[uNodeIndex] = true;
MSA *ptrMSA = new MSA;
// TODO: Wasteful copy, needs re-design
ptrMSA->Copy(SubfamMSAs[uSubfamIndex]);
MSAs[uNodeIndex] = ptrMSA;
}
for (unsigned uNodeIndex = tree.FirstDepthFirstNode();
NULL_NEIGHBOR != uNodeIndex;
uNodeIndex = tree.NextDepthFirstNode(uNodeIndex))
{
if (tree.IsLeaf(uNodeIndex))
continue;
unsigned uRight = tree.GetRight(uNodeIndex);
unsigned uLeft = tree.GetLeft(uNodeIndex);
if (!Ready[uRight] || !Ready[uLeft])
continue;
MSA *ptrLeft = MSAs[uLeft];
MSA *ptrRight = MSAs[uRight];
assert(ptrLeft != 0 && ptrRight != 0);
MSA *ptrParent = new MSA;
PWPath Path;
AlignTwoMSAs(*ptrLeft, *ptrRight, *ptrParent, Path);
MSAs[uNodeIndex] = ptrParent;
Ready[uNodeIndex] = true;
Ready[uLeft] = false;
Ready[uRight] = false;
delete MSAs[uLeft];
delete MSAs[uRight];
MSAs[uLeft] = 0;
MSAs[uRight] = 0;
}
#if DEBUG
{
unsigned uReadyCount = 0;
for (unsigned uNodeIndex = 0; uNodeIndex < uNodeCount; ++uNodeIndex)
{
if (Ready[uNodeIndex])
{
assert(tree.IsRoot(uNodeIndex));
++uReadyCount;
assert(0 != MSAs[uNodeIndex]);
}
else
assert(0 == MSAs[uNodeIndex]);
}
assert(1 == uReadyCount);
}
#endif
const unsigned uRoot = tree.GetRootNodeIndex();
MSA *ptrRootAlignment = MSAs[uRoot];
msa.Copy(*ptrRootAlignment);
delete ptrRootAlignment;
delete[] Ready;
#if TRACE
Log("After refine subfamilies, root alignment=\n");
msa.LogMe();
#endif
}
void Tree::PruneTree(const Tree &tree, unsigned Subfams[],
unsigned uSubfamCount)
{
if (!tree.IsRooted())
Quit("Tree::PruneTree: requires rooted tree");
Clear();
m_uNodeCount = 2*uSubfamCount - 1;
InitCache(m_uNodeCount);
const unsigned uUnprunedNodeCount = tree.GetNodeCount();
unsigned *uUnprunedToPrunedIndex = new unsigned[uUnprunedNodeCount];
unsigned *uPrunedToUnprunedIndex = new unsigned[m_uNodeCount];
for (unsigned n = 0; n < uUnprunedNodeCount; ++n)
uUnprunedToPrunedIndex[n] = NULL_NEIGHBOR;
for (unsigned n = 0; n < m_uNodeCount; ++n)
uPrunedToUnprunedIndex[n] = NULL_NEIGHBOR;
// Create mapping between unpruned and pruned node indexes
unsigned uInternalNodeIndex = uSubfamCount;
for (unsigned uSubfamIndex = 0; uSubfamIndex < uSubfamCount; ++uSubfamIndex)
{
unsigned uUnprunedNodeIndex = Subfams[uSubfamIndex];
uUnprunedToPrunedIndex[uUnprunedNodeIndex] = uSubfamIndex;
uPrunedToUnprunedIndex[uSubfamIndex] = uUnprunedNodeIndex;
for (;;)
{
uUnprunedNodeIndex = tree.GetParent(uUnprunedNodeIndex);
if (tree.IsRoot(uUnprunedNodeIndex))
break;
// Already visited this node?
if (NULL_NEIGHBOR != uUnprunedToPrunedIndex[uUnprunedNodeIndex])
break;
uUnprunedToPrunedIndex[uUnprunedNodeIndex] = uInternalNodeIndex;
uPrunedToUnprunedIndex[uInternalNodeIndex] = uUnprunedNodeIndex;
++uInternalNodeIndex;
}
}
const unsigned uUnprunedRootIndex = tree.GetRootNodeIndex();
uUnprunedToPrunedIndex[uUnprunedRootIndex] = uInternalNodeIndex;
uPrunedToUnprunedIndex[uInternalNodeIndex] = uUnprunedRootIndex;
#if TRACE
{
Log("Pruned to unpruned:\n");
for (unsigned i = 0; i < m_uNodeCount; ++i)
Log(" [%u]=%u", i, uPrunedToUnprunedIndex[i]);
Log("\n");
Log("Unpruned to pruned:\n");
for (unsigned i = 0; i < uUnprunedNodeCount; ++i)
{
unsigned n = uUnprunedToPrunedIndex[i];
if (n != NULL_NEIGHBOR)
Log(" [%u]=%u", i, n);
}
Log("\n");
}
#endif
if (uInternalNodeIndex != m_uNodeCount - 1)
Quit("Tree::PruneTree, Internal error");
// Nodes 0, 1 ... are the leaves
for (unsigned uSubfamIndex = 0; uSubfamIndex < uSubfamCount; ++uSubfamIndex)
{
char szName[32];
sprintf(szName, "Subfam_%u", uSubfamIndex + 1);
m_ptrName[uSubfamIndex] = strsave(szName);
}
for (unsigned uPrunedNodeIndex = uSubfamCount; uPrunedNodeIndex < m_uNodeCount;
++uPrunedNodeIndex)
{
unsigned uUnprunedNodeIndex = uPrunedToUnprunedIndex[uPrunedNodeIndex];
const unsigned uUnprunedLeft = tree.GetLeft(uUnprunedNodeIndex);
const unsigned uUnprunedRight = tree.GetRight(uUnprunedNodeIndex);
const unsigned uPrunedLeft = uUnprunedToPrunedIndex[uUnprunedLeft];
const unsigned uPrunedRight = uUnprunedToPrunedIndex[uUnprunedRight];
const double dLeftLength =
tree.GetEdgeLength(uUnprunedNodeIndex, uUnprunedLeft);
const double dRightLength =
tree.GetEdgeLength(uUnprunedNodeIndex, uUnprunedRight);
m_uNeighbor2[uPrunedNodeIndex] = uPrunedLeft;
m_uNeighbor3[uPrunedNodeIndex] = uPrunedRight;
m_dEdgeLength1[uPrunedLeft] = dLeftLength;
m_dEdgeLength1[uPrunedRight] = dRightLength;
m_uNeighbor1[uPrunedLeft] = uPrunedNodeIndex;
m_uNeighbor1[uPrunedRight] = uPrunedNodeIndex;
m_bHasEdgeLength1[uPrunedLeft] = true;
m_bHasEdgeLength1[uPrunedRight] = true;
m_dEdgeLength2[uPrunedNodeIndex] = dLeftLength;
m_dEdgeLength3[uPrunedNodeIndex] = dRightLength;
m_bHasEdgeLength2[uPrunedNodeIndex] = true;
m_bHasEdgeLength3[uPrunedNodeIndex] = true;
}
m_uRootNodeIndex = uUnprunedToPrunedIndex[uUnprunedRootIndex];
m_bRooted = true;
Validate();
delete[] uUnprunedToPrunedIndex;
}
void DiffTrees(const Tree &Tree1, const Tree &Tree2, Tree &Diffs,
unsigned IdToDiffsLeafNodeIndex[])
{
#if TRACE
Log("Tree1:\n");
Tree1.LogMe();
Log("\n");
Log("Tree2:\n");
Tree2.LogMe();
#endif
if (!Tree1.IsRooted() || !Tree2.IsRooted())
Quit("DiffTrees: requires rooted trees");
const unsigned uNodeCount = Tree1.GetNodeCount();
const unsigned uNodeCount2 = Tree2.GetNodeCount();
const unsigned uLeafCount = Tree1.GetLeafCount();
const unsigned uLeafCount2 = Tree2.GetLeafCount();
assert(uLeafCount == uLeafCount2);
if (uNodeCount != uNodeCount2)
Quit("DiffTrees: different node counts");
// Allocate tables so we can convert tree node index to
// and from the unique id with a O(1) lookup.
unsigned *NodeIndexToId1 = new unsigned[uNodeCount];
unsigned *IdToNodeIndex2 = new unsigned[uNodeCount];
bool *bIsBachelor1 = new bool[uNodeCount];
bool *bIsDiff1 = new bool[uNodeCount];
for (unsigned uNodeIndex = 0; uNodeIndex < uNodeCount; ++uNodeIndex)
{
NodeIndexToId1[uNodeIndex] = uNodeCount;
bIsBachelor1[uNodeIndex] = false;
bIsDiff1[uNodeIndex] = false;
// Use uNodeCount as value meaning "not set".
IdToNodeIndex2[uNodeIndex] = uNodeCount;
}
// Initialize node index <-> id lookup tables
for (unsigned uNodeIndex = 0; uNodeIndex < uNodeCount; ++uNodeIndex)
{
if (Tree1.IsLeaf(uNodeIndex))
{
const unsigned uId = Tree1.GetLeafId(uNodeIndex);
if (uId >= uNodeCount)
Quit("Diff trees requires existing leaf ids in range 0 .. (N-1)");
NodeIndexToId1[uNodeIndex] = uId;
}
if (Tree2.IsLeaf(uNodeIndex))
{
const unsigned uId = Tree2.GetLeafId(uNodeIndex);
if (uId >= uNodeCount)
Quit("Diff trees requires existing leaf ids in range 0 .. (N-1)");
IdToNodeIndex2[uId] = uNodeIndex;
}
}
// Validity check. This verifies that the ids
// pre-assigned to the leaves in Tree1 are unique
// (note that the id<N check above does not rule
// out two leaves having duplicate ids).
for (unsigned uId = 0; uId < uLeafCount; ++uId)
{
unsigned uNodeIndex2 = IdToNodeIndex2[uId];
if (uNodeCount == uNodeIndex2)
Quit("DiffTrees, check 2");
}
// Ids assigned to internal nodes are N, N+1 ...
// An internal node id uniquely identifies a set
// of two or more leaves.
unsigned uInternalNodeId = uLeafCount;
// Depth-first traversal of tree.
// The order guarantees that a node is visited before
// its parent is visited.
for (unsigned uNodeIndex1 = Tree1.FirstDepthFirstNode();
NULL_NEIGHBOR != uNodeIndex1;
uNodeIndex1 = Tree1.NextDepthFirstNode(uNodeIndex1))
{
#if TRACE
Log("Main loop: Node1=%u IsLeaf=%d IsBachelor=%d\n",
uNodeIndex1,
Tree1.IsLeaf(uNodeIndex1),
bIsBachelor1[uNodeIndex1]);
#endif
// Leaves are trivial; nothing to do.
if (Tree1.IsLeaf(uNodeIndex1) || bIsBachelor1[uNodeIndex1])
continue;
// If either child is a bachelor, flag
// this node as a bachelor and continue.
unsigned uLeft1 = Tree1.GetLeft(uNodeIndex1);
if (bIsBachelor1[uLeft1])
{
bIsBachelor1[uNodeIndex1] = true;
continue;
}
unsigned uRight1 = Tree1.GetRight(uNodeIndex1);
if (bIsBachelor1[uRight1])
{
bIsBachelor1[uNodeIndex1] = true;
continue;
}
// Both children are married.
// Married nodes are guaranteed to have an id.
unsigned uIdLeft = NodeIndexToId1[uLeft1];
unsigned uIdRight = NodeIndexToId1[uRight1];
if (uIdLeft == uNodeCount || uIdRight == uNodeCount)
Quit("DiffTrees, check 5");
// uLeft2 is the spouse of uLeft1, and similarly for uRight2.
unsigned uLeft2 = IdToNodeIndex2[uIdLeft];
unsigned uRight2 = IdToNodeIndex2[uIdRight];
if (uLeft2 == uNodeCount || uRight2 == uNodeCount)
Quit("DiffTrees, check 6");
// If the spouses of uLeft1 and uRight1 have the same
// parent, then this parent is the spouse of uNodeIndex1.
// Otherwise, uNodeIndex1 is a diff.
unsigned uParentLeft2 = Tree2.GetParent(uLeft2);
unsigned uParentRight2 = Tree2.GetParent(uRight2);
#if TRACE
Log("L1=%u R1=%u L2=%u R2=%u PL2=%u PR2=%u\n",
uLeft1,
uRight1,
uLeft2,
uRight2,
uParentLeft2,
uParentRight2);
#endif
if (uParentLeft2 == uParentRight2)
{
NodeIndexToId1[uNodeIndex1] = uInternalNodeId;
IdToNodeIndex2[uInternalNodeId] = uParentLeft2;
++uInternalNodeId;
}
else
bIsBachelor1[uNodeIndex1] = true;
}
unsigned uDiffCount = 0;
for (unsigned uNodeIndex = 0; uNodeIndex < uNodeCount; ++uNodeIndex)
{
if (bIsBachelor1[uNodeIndex])
continue;
if (Tree1.IsRoot(uNodeIndex))
{
// Special case: if no bachelors, consider the
// root a diff.
if (!bIsBachelor1[uNodeIndex])
bIsDiff1[uNodeIndex] = true;
continue;
}
const unsigned uParent = Tree1.GetParent(uNodeIndex);
if (bIsBachelor1[uParent])
{
bIsDiff1[uNodeIndex] = true;
++uDiffCount;
}
}
#if TRACE
Log("Tree1:\n");
Log("Node Id Bach Diff Name\n");
Log("---- ---- ---- ---- ----\n");
for (unsigned n = 0; n < uNodeCount; ++n)
{
Log("%4u %4u %d %d",
n,
NodeIndexToId1[n],
bIsBachelor1[n],
bIsDiff1[n]);
if (Tree1.IsLeaf(n))
Log(" %s", Tree1.GetLeafName(n));
Log("\n");
}
Log("\n");
Log("Tree2:\n");
Log("Node Id Name\n");
Log("---- ---- ----\n");
for (unsigned n = 0; n < uNodeCount; ++n)
{
Log("%4u ", n);
if (Tree2.IsLeaf(n))
Log(" %s", Tree2.GetLeafName(n));
Log("\n");
}
#endif
Diffs.CreateRooted();
const unsigned uDiffsRootIndex = Diffs.GetRootNodeIndex();
const unsigned uRootIndex1 = Tree1.GetRootNodeIndex();
for (unsigned n = 0; n < uLeafCount; ++n)
IdToDiffsLeafNodeIndex[n] = uNodeCount;
BuildDiffs(Tree1, uRootIndex1, bIsDiff1, Diffs, uDiffsRootIndex,
IdToDiffsLeafNodeIndex);
#if TRACE
Log("\n");
Log("Diffs:\n");
Diffs.LogMe();
Log("\n");
Log("IdToDiffsLeafNodeIndex:");
for (unsigned n = 0; n < uLeafCount; ++n)
{
if (n%16 == 0)
Log("\n");
else
Log(" ");
Log("%u=%u", n, IdToDiffsLeafNodeIndex[n]);
}
Log("\n");
#endif
for (unsigned n = 0; n < uLeafCount; ++n)
if (IdToDiffsLeafNodeIndex[n] == uNodeCount)
Quit("TreeDiffs check 7");
delete[] NodeIndexToId1;
delete[] IdToNodeIndex2;
delete[] bIsBachelor1;
delete[] bIsDiff1;
}
