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;
}
// 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;
}
// Divide a tree containing N leaves into k families by
// cutting the tree at a horizontal line at some height.
// Each internal node defines a height for the cut,
// considering all internal nodes enumerates all distinct
// cuts. Visit internal nodes in decreasing order of height.
// Visiting the node corresponds to moving the horizontal
// line down to cut the tree at the height of that node.
// We consider the cut to be "infinitestimally below"
// the node, so the effect is to remove the current node
// from the list of subfamilies and add its two children.
// We must visit a parent before its children (so care may
// be needed to handle zero edge lengths properly).
// We assume that N is small, and write dumb O(N^2) code.
// More efficient strategies are possible for large N
// by maintaining a list of nodes sorted by height.
void ClusterBySubfamCount(const Tree &tree, unsigned uSubfamCount,
unsigned Subfams[], unsigned *ptruSubfamCount)
{
const unsigned uNodeCount = tree.GetNodeCount();
const unsigned uLeafCount = (uNodeCount + 1)/2;
// Special case: empty tree
if (0 == uNodeCount)
{
*ptruSubfamCount = 0;
return;
}
// Special case: more subfamilies than leaves
if (uSubfamCount >= uLeafCount)
{
for (unsigned n = 0; n < uLeafCount; ++n)
Subfams[n] = n;
*ptruSubfamCount = uLeafCount;
return;
}
// Initialize list of subfamilies to be root
Subfams[0] = tree.GetRootNodeIndex();
// Iterate
for (unsigned i = 1; i < uSubfamCount; ++i)
ClusterBySubfamCount_Iteration(tree, Subfams, i);
*ptruSubfamCount = uSubfamCount;
}
void RefineTreeE(MSA &msa, const SeqVect &v, Tree &tree, ProgNode *ProgNodes)
{
MuscleContext *ctx = getMuscleContext();
const unsigned uSeqCount = msa.GetSeqCount();
if (tree.GetLeafCount() != uSeqCount)
Quit("Refine tree, tree has different number of nodes");
if (uSeqCount < 3)
return;
#if DEBUG
ValidateMuscleIds(msa);
ValidateMuscleIds(tree);
#endif
const unsigned uNodeCount = tree.GetNodeCount();
unsigned *uNewNodeIndexToOldNodeIndex= new unsigned[uNodeCount];
Tree Tree2;
TreeFromMSA(msa, Tree2, ctx->params.g_Cluster2, ctx->params.g_Distance2, ctx->params.g_Root2, ctx->params.g_pstrDistMxFileName2);
#if DEBUG
ValidateMuscleIds(Tree2);
#endif
DiffTreesE(Tree2, tree, uNewNodeIndexToOldNodeIndex);
unsigned uRoot = Tree2.GetRootNodeIndex();
if (NODE_CHANGED == uNewNodeIndexToOldNodeIndex[uRoot])
{
MSA msa2;
RealignDiffsE(msa, v, Tree2, tree, uNewNodeIndexToOldNodeIndex, msa2, ProgNodes);
if (!ctx->isCanceled()) {
tree.Copy(Tree2);
msa.Copy(msa2);
#if DEBUG
ValidateMuscleIds(msa2);
#endif
}
}
delete[] uNewNodeIndexToOldNodeIndex;
if (ctx->isCanceled()) {
throw MuscleException("Canceled");
}
SetCurrentAlignment(msa);
ProgressStepsDone();
}
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;
}
void GetInternalNodesInHeightOrder(const Tree &tree, unsigned NodeIndexes[])
{
const unsigned uNodeCount = tree.GetNodeCount();
if (uNodeCount < 3)
Quit("GetInternalNodesInHeightOrder: %u nodes, none are internal",
uNodeCount);
const unsigned uInternalNodeCount = (uNodeCount - 1)/2;
double *Heights = new double[uInternalNodeCount];
unsigned uIndex = 0;
for (unsigned uNodeIndex = 0; uNodeIndex < uNodeCount; ++uNodeIndex)
{
if (tree.IsLeaf(uNodeIndex))
continue;
NodeIndexes[uIndex] = uNodeIndex;
Heights[uIndex] = tree.GetNodeHeight(uNodeIndex);
++uIndex;
}
if (uIndex != uInternalNodeCount)
Quit("Internal error: GetInternalNodesInHeightOrder");
// Simple but slow bubble sort (probably don't care about speed here)
bool bDone = false;
while (!bDone)
{
bDone = true;
for (unsigned i = 0; i < uInternalNodeCount - 1; ++i)
{
if (Heights[i] > Heights[i+1])
{
double dTmp = Heights[i];
Heights[i] = Heights[i+1];
Heights[i+1] = dTmp;
unsigned uTmp = NodeIndexes[i];
NodeIndexes[i] = NodeIndexes[i+1];
NodeIndexes[i+1] = uTmp;
bDone = false;
}
}
}
#if TRACE
Log("Internal node index Height\n");
Log("------------------- --------\n");
// 1234567890123456789 123456789
for (unsigned n = 0; n < uInternalNodeCount; ++n)
Log("%19u %9.3f\n", NodeIndexes[n], Heights[n]);
#endif
delete[] Heights;
}
// Identify subfamilies in a tree.
// Returns array of internal node indexes, one for each subfamily.
// First try is to select groups by height (which should approximate
// minimum percent identity), if this gives too many subfamilies then
// we cut at a point that gives the maximum allowed number of subfams.
static void GetSubfams(const Tree &tree, double dMaxHeight,
unsigned uMaxSubfamCount, unsigned **ptrptrSubfams, unsigned *ptruSubfamCount)
{
const unsigned uNodeCount = tree.GetNodeCount();
unsigned *Subfams = new unsigned[uNodeCount];
unsigned uSubfamCount;
ClusterByHeight(tree, dMaxHeight, Subfams, &uSubfamCount);
if (uSubfamCount > uMaxSubfamCount)
ClusterBySubfamCount(tree, uMaxSubfamCount, Subfams, &uSubfamCount);
*ptrptrSubfams = Subfams;
*ptruSubfamCount = uSubfamCount;
}
void ApplyMinEdgeLength(Tree &tree, double dMinEdgeLength)
{
const unsigned uNodeCount = tree.GetNodeCount();
for (unsigned uNodeIndex = 0; uNodeIndex < uNodeCount; ++uNodeIndex)
{
const unsigned uNeighborCount = tree.GetNeighborCount(uNodeIndex);
for (unsigned n = 0; n < uNeighborCount; ++n)
{
const unsigned uNeighborNodeIndex = tree.GetNeighbor(uNodeIndex, n);
if (!tree.HasEdgeLength(uNodeIndex, uNeighborNodeIndex))
continue;
if (tree.GetEdgeLength(uNodeIndex, uNeighborNodeIndex) < dMinEdgeLength)
tree.SetEdgeLength(uNodeIndex, uNeighborNodeIndex, dMinEdgeLength);
}
}
}
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 unsigned GetNextNodeIndex(const Tree &tree, unsigned uPrevNodeIndex)
{
if (g_bStable)
{
const unsigned uNodeCount = tree.GetNodeCount();
unsigned uNodeIndex = uPrevNodeIndex;
for (;;)
{
++uNodeIndex;
if (uNodeIndex >= uNodeCount)
return NULL_NEIGHBOR;
if (tree.IsLeaf(uNodeIndex))
return uNodeIndex;
}
}
unsigned uNodeIndex = uPrevNodeIndex;
for (;;)
{
uNodeIndex = tree.NextDepthFirstNode(uNodeIndex);
if (NULL_NEIGHBOR == uNodeIndex || tree.IsLeaf(uNodeIndex))
return uNodeIndex;
}
}
void TestBiPart()
{
SetListFileName("c:\\tmp\\lobster.log", false);
Tree tree;
TextFile fileIn("c:\\tmp\\test.phy");
tree.FromFile(fileIn);
tree.LogMe();
const unsigned uNodeCount = tree.GetNodeCount();
unsigned *Leaves1 = new unsigned[uNodeCount];
unsigned *Leaves2 = new unsigned[uNodeCount];
PhyEnumEdgeState ES;
bool bDone = false;
for (;;)
{
unsigned uCount1 = uInsane;
unsigned uCount2 = uInsane;
bool bOk = PhyEnumBiParts(tree, ES, Leaves1, &uCount1, Leaves2, &uCount2);
Log("PEBP=%d ES.Init=%d ES.ni1=%d ES.ni2=%d\n",
bOk,
ES.m_bInit,
ES.m_uNodeIndex1,
ES.m_uNodeIndex2);
if (!bOk)
break;
Log("\n");
Log("Part1: ");
for (unsigned n = 0; n < uCount1; ++n)
Log(" %d(%s)", Leaves1[n], tree.GetLeafName(Leaves1[n]));
Log("\n");
Log("Part2: ");
for (unsigned n = 0; n < uCount2; ++n)
Log(" %d(%s)", Leaves2[n], tree.GetLeafName(Leaves2[n]));
Log("\n");
}
}
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);
}
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;
}
void DoMuscle()
{
SetOutputFileName(g_pstrOutFileName.get());
SetInputFileName(g_pstrInFileName.get());
SetMaxIters(g_uMaxIters.get());
SetSeqWeightMethod(g_SeqWeight1.get());
TextFile fileIn(g_pstrInFileName.get());
SeqVect v;
v.FromFASTAFile(fileIn);
const unsigned uSeqCount = v.Length();
if (0 == uSeqCount)
Quit("No sequences in input file");
ALPHA Alpha = ALPHA_Undefined;
switch (g_SeqType.get())
{
case SEQTYPE_Auto:
Alpha = v.GuessAlpha();
break;
case SEQTYPE_Protein:
Alpha = ALPHA_Amino;
break;
case SEQTYPE_DNA:
Alpha = ALPHA_DNA;
break;
case SEQTYPE_RNA:
Alpha = ALPHA_RNA;
break;
default:
Quit("Invalid seq type");
}
SetAlpha(Alpha);
v.FixAlpha();
//
// AED 21/12/06: Moved matrix loading code inside the PP param function so it gets called for all alignment types
//
SetPPScore();
unsigned uMaxL = 0;
unsigned uTotL = 0;
for (unsigned uSeqIndex = 0; uSeqIndex < uSeqCount; ++uSeqIndex)
{
unsigned L = v.GetSeq(uSeqIndex).Length();
uTotL += L;
if (L > uMaxL)
uMaxL = L;
}
SetIter(1);
g_bDiags.get() = g_bDiags1.get();
SetSeqStats(uSeqCount, uMaxL, uTotL/uSeqCount);
SetMuscleSeqVect(v);
MSA::SetIdCount(uSeqCount);
// Initialize sequence ids.
// From this point on, ids must somehow propogate from here.
for (unsigned uSeqIndex = 0; uSeqIndex < uSeqCount; ++uSeqIndex)
v.SetSeqId(uSeqIndex, uSeqIndex);
if (0 == uSeqCount)
Quit("Input file '%s' has no sequences", g_pstrInFileName.get());
if (1 == uSeqCount)
{
TextFile fileOut(g_pstrOutFileName.get(), true);
v.ToFile(fileOut);
return;
}
if (uSeqCount > 1)
MHackStart(v);
// First iteration
Tree GuideTree;
if (0 != g_pstrUseTreeFileName.get())
{
// Discourage users...
if (!g_bUseTreeNoWarn.get())
fprintf(stderr, g_strUseTreeWarning);
// Read tree from file
TextFile TreeFile(g_pstrUseTreeFileName.get());
GuideTree.FromFile(TreeFile);
// Make sure tree is rooted
if (!GuideTree.IsRooted())
Quit("User tree must be rooted");
if (GuideTree.GetLeafCount() != uSeqCount)
Quit("User tree does not match input sequences");
const unsigned uNodeCount = GuideTree.GetNodeCount();
for (unsigned uNodeIndex = 0; uNodeIndex < uNodeCount; ++uNodeIndex)
{
if (!GuideTree.IsLeaf(uNodeIndex))
continue;
const char *LeafName = GuideTree.GetLeafName(uNodeIndex);
unsigned uSeqIndex;
bool SeqFound = v.FindName(LeafName, &uSeqIndex);
if (!SeqFound)
Quit("Label %s in tree does not match sequences", LeafName);
unsigned uId = v.GetSeqIdFromName(LeafName);
GuideTree.SetLeafId(uNodeIndex, uId);
}
}
else
TreeFromSeqVect(v, GuideTree, g_Cluster1.get(), g_Distance1.get(), g_Root1.get(),
g_pstrDistMxFileName1.get());
const char *Tree1 = ValueOpt("Tree1");
if (0 != Tree1)
{
TextFile f(Tree1, true);
GuideTree.ToFile(f);
if (g_bClusterOnly.get())
return;
}
SetMuscleTree(GuideTree);
ValidateMuscleIds(GuideTree);
MSA msa;
ProgNode *ProgNodes = 0;
if (g_bLow.get())
ProgNodes = ProgressiveAlignE(v, GuideTree, msa);
else
ProgressiveAlign(v, GuideTree, msa);
SetCurrentAlignment(msa);
if (0 != g_pstrComputeWeightsFileName.get())
{
extern void OutWeights(const char *FileName, const MSA &msa);
SetMSAWeightsMuscle(msa);
OutWeights(g_pstrComputeWeightsFileName.get(), msa);
return;
}
ValidateMuscleIds(msa);
if (1 == g_uMaxIters.get() || 2 == uSeqCount)
{
//TextFile fileOut(g_pstrOutFileName.get(), true);
//MHackEnd(msa);
//msa.ToFile(fileOut);
MuscleOutput(msa);
return;
}
if (0 == g_pstrUseTreeFileName.get())
{
g_bDiags.get() = g_bDiags2.get();
SetIter(2);
if (g_bLow.get())
{
if (0 != g_uMaxTreeRefineIters.get())
RefineTreeE(msa, v, GuideTree, ProgNodes);
}
else
RefineTree(msa, GuideTree);
const char *Tree2 = ValueOpt("Tree2");
if (0 != Tree2)
{
TextFile f(Tree2, true);
GuideTree.ToFile(f);
}
}
SetSeqWeightMethod(g_SeqWeight2.get());
SetMuscleTree(GuideTree);
if (g_bAnchors.get())
RefineVert(msa, GuideTree, g_uMaxIters.get() - 2);
else
RefineHoriz(msa, GuideTree, g_uMaxIters.get() - 2, false, false);
#if 0
// Refining by subfamilies is disabled as it didn't give better
// results. I tried doing this before and after RefineHoriz.
// Should get back to this as it seems like this should work.
RefineSubfams(msa, GuideTree, g_uMaxIters.get() - 2);
#endif
ValidateMuscleIds(msa);
ValidateMuscleIds(GuideTree);
//TextFile fileOut(g_pstrOutFileName.get(), true);
//MHackEnd(msa);
//msa.ToFile(fileOut);
MuscleOutput(msa);
}
void DiffTreesE(const Tree &NewTree, const Tree &OldTree,
unsigned NewNodeIndexToOldNodeIndex[])
{
#if TRACE
Log("DiffTreesE NewTree:\n");
NewTree.LogMe();
Log("\n");
Log("OldTree:\n");
OldTree.LogMe();
#endif
if (!NewTree.IsRooted() || !OldTree.IsRooted())
Quit("DiffTrees: requires rooted trees");
const unsigned uNodeCount = NewTree.GetNodeCount();
const unsigned uOldNodeCount = OldTree.GetNodeCount();
const unsigned uLeafCount = NewTree.GetLeafCount();
const unsigned uOldLeafCount = OldTree.GetLeafCount();
if (uNodeCount != uOldNodeCount || uLeafCount != uOldLeafCount)
Quit("DiffTreesE: different node counts");
{
unsigned *IdToOldNodeIndex = new unsigned[uNodeCount];
for (unsigned uOldNodeIndex = 0; uOldNodeIndex < uNodeCount; ++uOldNodeIndex)
{
if (OldTree.IsLeaf(uOldNodeIndex))
{
unsigned Id = OldTree.GetLeafId(uOldNodeIndex);
IdToOldNodeIndex[Id] = uOldNodeIndex;
}
}
// Initialize NewNodeIndexToOldNodeIndex[]
// All internal nodes are marked as changed, but may be updated later.
for (unsigned uNewNodeIndex = 0; uNewNodeIndex < uNodeCount; ++uNewNodeIndex)
{
if (NewTree.IsLeaf(uNewNodeIndex))
{
unsigned uId = NewTree.GetLeafId(uNewNodeIndex);
assert(uId < uLeafCount);
unsigned uOldNodeIndex = IdToOldNodeIndex[uId];
assert(uOldNodeIndex < uNodeCount);
NewNodeIndexToOldNodeIndex[uNewNodeIndex] = uOldNodeIndex;
}
else
NewNodeIndexToOldNodeIndex[uNewNodeIndex] = NODE_CHANGED;
}
delete[] IdToOldNodeIndex;
}
// Depth-first traversal of tree.
// The order guarantees that a node is visited before
// its parent is visited.
for (unsigned uNewNodeIndex = NewTree.FirstDepthFirstNode();
NULL_NEIGHBOR != uNewNodeIndex;
uNewNodeIndex = NewTree.NextDepthFirstNode(uNewNodeIndex))
{
if (NewTree.IsLeaf(uNewNodeIndex))
continue;
// If either child is changed, flag this node as changed and continue.
unsigned uNewLeft = NewTree.GetLeft(uNewNodeIndex);
unsigned uOldLeft = NewNodeIndexToOldNodeIndex[uNewLeft];
if (NODE_CHANGED == uOldLeft)
{
NewNodeIndexToOldNodeIndex[uNewLeft] = NODE_CHANGED;
continue;
}
unsigned uNewRight = NewTree.GetRight(uNewNodeIndex);
unsigned uOldRight = NewNodeIndexToOldNodeIndex[uNewRight];
if (NODE_CHANGED == NewNodeIndexToOldNodeIndex[uNewRight])
{
NewNodeIndexToOldNodeIndex[uNewRight] = NODE_CHANGED;
continue;
}
unsigned uOldParentLeft = OldTree.GetParent(uOldLeft);
unsigned uOldParentRight = OldTree.GetParent(uOldRight);
if (uOldParentLeft == uOldParentRight)
NewNodeIndexToOldNodeIndex[uNewNodeIndex] = uOldParentLeft;
else
NewNodeIndexToOldNodeIndex[uNewNodeIndex] = NODE_CHANGED;
}
#if TRACE
{
Log("NewToOld ");
for (unsigned uNewNodeIndex = 0; uNewNodeIndex < uNodeCount; ++uNewNodeIndex)
{
Log(" [%3u]=", uNewNodeIndex);
if (NODE_CHANGED == NewNodeIndexToOldNodeIndex[uNewNodeIndex])
Log(" X");
else
Log("%3u", NewNodeIndexToOldNodeIndex[uNewNodeIndex]);
if ((uNewNodeIndex+1)%8 == 0)
Log("\n ");
}
Log("\n");
}
#endif
#if DEBUG
{
for (unsigned uNewNodeIndex = 0; uNewNodeIndex < uNodeCount; ++uNewNodeIndex)
{
unsigned uOld = NewNodeIndexToOldNodeIndex[uNewNodeIndex];
if (NewTree.IsLeaf(uNewNodeIndex))
{
if (uOld >= uNodeCount)
{
Log("NewNode=%u uOld=%u > uNodeCount=%u\n",
uNewNodeIndex, uOld, uNodeCount);
Quit("Diff check failed");
}
unsigned uIdNew = NewTree.GetLeafId(uNewNodeIndex);
unsigned uIdOld = OldTree.GetLeafId(uOld);
if (uIdNew != uIdOld)
{
Log("NewNode=%u uOld=%u IdNew=%u IdOld=%u\n",
uNewNodeIndex, uOld, uIdNew, uIdOld);
Quit("Diff check failed");
}
continue;
}
if (NODE_CHANGED == uOld)
continue;
unsigned uNewLeft = NewTree.GetLeft(uNewNodeIndex);
unsigned uNewRight = NewTree.GetRight(uNewNodeIndex);
unsigned uOldLeft = OldTree.GetLeft(uOld);
unsigned uOldRight = OldTree.GetRight(uOld);
unsigned uNewLeftPartner = NewNodeIndexToOldNodeIndex[uNewLeft];
unsigned uNewRightPartner = NewNodeIndexToOldNodeIndex[uNewRight];
bool bSameNotRotated = (uNewLeftPartner == uOldLeft && uNewRightPartner == uOldRight);
bool bSameRotated = (uNewLeftPartner == uOldRight && uNewRightPartner == uOldLeft);
if (!bSameNotRotated && !bSameRotated)
{
Log("NewNode=%u NewL=%u NewR=%u\n", uNewNodeIndex, uNewLeft, uNewRight);
Log("OldNode=%u OldL=%u OldR=%u\n", uOld, uOldLeft, uOldRight);
Log("NewLPartner=%u NewRPartner=%u\n", uNewLeftPartner, uNewRightPartner);
Quit("Diff check failed");
}
}
}
#endif
}
void DoMuscle(CompositeVect*CVLocation)
{
SetOutputFileName(g_pstrOutFileName);
SetInputFileName(g_pstrInFileName);
SetMaxIters(g_uMaxIters);
SetSeqWeightMethod(g_SeqWeight1);
TextFile fileIn(g_pstrInFileName);
SeqVect v;
v.FromFASTAFile(fileIn);
const unsigned uSeqCount = v.Length();
if (0 == uSeqCount)
Quit("No sequences in input file");
ALPHA Alpha = ALPHA_Undefined;
switch (g_SeqType)
{
case SEQTYPE_Auto:
Alpha = v.GuessAlpha();
break;
case SEQTYPE_Protein:
Alpha = ALPHA_Amino;
break;
case SEQTYPE_DNA:
Alpha = ALPHA_DNA;
break;
case SEQTYPE_RNA:
Alpha = ALPHA_RNA;
break;
default:
Quit("Invalid seq type");
}
SetAlpha(Alpha);
v.FixAlpha();
PTR_SCOREMATRIX UserMatrix = 0;
if (0 != g_pstrMatrixFileName)
{
const char *FileName = g_pstrMatrixFileName;
const char *Path = getenv("MUSCLE_MXPATH");
if (Path != 0)
{
size_t n = strlen(Path) + 1 + strlen(FileName) + 1;
char *NewFileName = new char[n];
sprintf(NewFileName, "%s/%s", Path, FileName);
FileName = NewFileName;
}
TextFile File(FileName);
UserMatrix = ReadMx(File);
g_Alpha = ALPHA_Amino;
g_PPScore = PPSCORE_SP;
}
SetPPScore();
if (0 != UserMatrix)
g_ptrScoreMatrix = UserMatrix;
unsigned uMaxL = 0;
unsigned uTotL = 0;
for (unsigned uSeqIndex = 0; uSeqIndex < uSeqCount; ++uSeqIndex)
{
unsigned L = v.GetSeq(uSeqIndex).Length();
uTotL += L;
if (L > uMaxL)
uMaxL = L;
}
SetIter(1);
g_bDiags = g_bDiags1;
SetSeqStats(uSeqCount, uMaxL, uTotL/uSeqCount);
SetMuscleSeqVect(v);
MSA::SetIdCount(uSeqCount);
// Initialize sequence ids.
// From this point on, ids must somehow propogate from here.
for (unsigned uSeqIndex = 0; uSeqIndex < uSeqCount; ++uSeqIndex)
v.SetSeqId(uSeqIndex, uSeqIndex);
if (0 == uSeqCount)
Quit("Input file '%s' has no sequences", g_pstrInFileName);
if (1 == uSeqCount)
{
TextFile fileOut(g_pstrOutFileName, true);
v.ToFile(fileOut);
return;
}
if (uSeqCount > 1)
MHackStart(v);
// First iteration
Tree GuideTree;
if (0 != g_pstrUseTreeFileName)
{
// Discourage users...
if (!g_bUseTreeNoWarn)
fprintf(stderr, "%s", g_strUseTreeWarning);
// Read tree from file
TextFile TreeFile(g_pstrUseTreeFileName);
GuideTree.FromFile(TreeFile);
// Make sure tree is rooted
if (!GuideTree.IsRooted())
Quit("User tree must be rooted");
if (GuideTree.GetLeafCount() != uSeqCount)
Quit("User tree does not match input sequences");
const unsigned uNodeCount = GuideTree.GetNodeCount();
for (unsigned uNodeIndex = 0; uNodeIndex < uNodeCount; ++uNodeIndex)
{
if (!GuideTree.IsLeaf(uNodeIndex))
continue;
const char *LeafName = GuideTree.GetLeafName(uNodeIndex);
unsigned uSeqIndex;
bool SeqFound = v.FindName(LeafName, &uSeqIndex);
if (!SeqFound)
Quit("Label %s in tree does not match sequences", LeafName);
unsigned uId = v.GetSeqIdFromName(LeafName);
GuideTree.SetLeafId(uNodeIndex, uId);
}
}
else
TreeFromSeqVect(v, GuideTree, g_Cluster1, g_Distance1, g_Root1,
g_pstrDistMxFileName1);
const char *Tree1 = ValueOpt("Tree1");
if (0 != Tree1)
{
TextFile f(Tree1, true);
GuideTree.ToFile(f);
if (g_bClusterOnly)
return;
}
SetMuscleTree(GuideTree);
ValidateMuscleIds(GuideTree);
MSA msa;
msa.SetCompositeVector(CVLocation);
ProgNode *ProgNodes = 0;
if (g_bLow)
ProgNodes = ProgressiveAlignE(v, GuideTree, msa);
else
ProgressiveAlign(v, GuideTree, msa);
SetCurrentAlignment(msa);
if (0 != g_pstrComputeWeightsFileName)
{
extern void OutWeights(const char *FileName, const MSA &msa);
SetMSAWeightsMuscle(msa);
OutWeights(g_pstrComputeWeightsFileName, msa);
return;
}
ValidateMuscleIds(msa);
if (1 == g_uMaxIters || 2 == uSeqCount)
{
//TextFile fileOut(g_pstrOutFileName, true);
//MHackEnd(msa);
//msa.ToFile(fileOut);
MuscleOutput(msa);
return;
}
if (0 == g_pstrUseTreeFileName)
{
g_bDiags = g_bDiags2;
SetIter(2);
if (g_bLow)
{
if (0 != g_uMaxTreeRefineIters)
RefineTreeE(msa, v, GuideTree, ProgNodes);
}
else
RefineTree(msa, GuideTree);
const char *Tree2 = ValueOpt("Tree2");
if (0 != Tree2)
{
TextFile f(Tree2, true);
GuideTree.ToFile(f);
}
}
SetSeqWeightMethod(g_SeqWeight2);
SetMuscleTree(GuideTree);
if (g_bAnchors)
RefineVert(msa, GuideTree, g_uMaxIters - 2);
else
RefineHoriz(msa, GuideTree, g_uMaxIters - 2, false, false);
#if 0
// Refining by subfamilies is disabled as it didn't give better
// results. I tried doing this before and after RefineHoriz.
// Should get back to this as it seems like this should work.
RefineSubfams(msa, GuideTree, g_uMaxIters - 2);
#endif
ValidateMuscleIds(msa);
ValidateMuscleIds(GuideTree);
//TextFile fileOut(g_pstrOutFileName, true);
//MHackEnd(msa);
//msa.ToFile(fileOut);
MuscleOutput(msa);
}
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
}
bool RefineSubfams(MSA &msa, const Tree &tree, unsigned uIters)
{
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 (g_bVerbose.get())
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.get(), g_Distance2.get(), g_Root2.get());
bool bAnyChangesThisSubfam;
if (g_bAnchors.get())
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 MakeRootMSA(const SeqVect &v, const Tree &GuideTree, ProgNode Nodes[],
MSA &a)
{
#if TRACE
Log("MakeRootMSA Tree=");
GuideTree.LogMe();
#endif
const unsigned uSeqCount = v.GetSeqCount();
unsigned uColCount = uInsane;
unsigned uSeqIndex = 0;
const unsigned uTreeNodeCount = GuideTree.GetNodeCount();
const unsigned uRootNodeIndex = GuideTree.GetRootNodeIndex();
const PWPath &RootPath = Nodes[uRootNodeIndex].m_Path;
const unsigned uRootColCount = RootPath.GetEdgeCount();
const unsigned uEstringSize = uRootColCount + 1;
short *Estring1 = new short[uEstringSize];
short *Estring2 = new short[uEstringSize];
SetProgressDesc("Root alignment");
unsigned uTreeNodeIndex = GetFirstNodeIndex(GuideTree);
do
{
Progress(uSeqIndex, uSeqCount);
unsigned uId = GuideTree.GetLeafId(uTreeNodeIndex);
const Seq &s = *(v[uId]);
Seq sRootE;
short *es = MakeRootSeqE(s, GuideTree, uTreeNodeIndex, Nodes, sRootE,
Estring1, Estring2);
Nodes[uTreeNodeIndex].m_EstringL = EstringNewCopy(es);
#if VALIDATE
Seq sRoot;
MakeRootSeq(s, GuideTree, uTreeNodeIndex, Nodes, sRoot);
if (!sRoot.Eq(sRootE))
{
Log("sRoot=");
sRoot.LogMe();
Log("sRootE=");
sRootE.LogMe();
Quit("Root seqs differ");
}
#if TRACE
Log("MakeRootSeq=\n");
sRoot.LogMe();
#endif
#endif
if (uInsane == uColCount)
{
uColCount = sRootE.Length();
a.SetSize(uSeqCount, uColCount);
}
else
{
assert(uColCount == sRootE.Length());
}
a.SetSeqName(uSeqIndex, s.GetName());
a.SetSeqId(uSeqIndex, uId);
for (unsigned uColIndex = 0; uColIndex < uColCount; ++uColIndex)
a.SetChar(uSeqIndex, uColIndex, sRootE[uColIndex]);
++uSeqIndex;
uTreeNodeIndex = GetNextNodeIndex(GuideTree, uTreeNodeIndex);
}
while (NULL_NEIGHBOR != uTreeNodeIndex);
delete[] Estring1;
delete[] Estring2;
ProgressStepsDone();
assert(uSeqIndex == uSeqCount);
}
void FindRoot(const Tree &tree, unsigned *ptruNode1, unsigned *ptruNode2,
double *ptrdLength1, double *ptrdLength2,
ROOT RootMethod)
{
#if TRACE
tree.LogMe();
#endif
if (tree.IsRooted())
Quit("FindRoot: tree already rooted");
const unsigned uNodeCount = tree.GetNodeCount();
const unsigned uLeafCount = tree.GetLeafCount();
if (uNodeCount < 2)
Quit("Root: don't support trees with < 2 edges");
EdgeInfo **EIs = new EdgeInfo *[uNodeCount];
for (unsigned uNodeIndex = 0; uNodeIndex < uNodeCount; ++uNodeIndex)
EIs[uNodeIndex] = new EdgeInfo[3];
EdgeList Edges;
for (unsigned uNodeIndex = 0; uNodeIndex < uNodeCount; ++uNodeIndex)
if (tree.IsLeaf(uNodeIndex))
{
unsigned uParent = tree.GetNeighbor1(uNodeIndex);
Edges.Add(uParent, uNodeIndex);
}
#if TRACE
Log("Edges: ");
Edges.LogMe();
#endif
// Main loop: iterate until all distances known
double dAllMaxDist = -1e20;
unsigned uMaxFrom = NULL_NEIGHBOR;
unsigned uMaxTo = NULL_NEIGHBOR;
for (;;)
{
EdgeList NextEdges;
#if TRACE
Log("\nTop of main loop\n");
Log("Edges: ");
Edges.LogMe();
Log("MDs:\n");
ListEIs(EIs, uNodeCount);
#endif
// For all edges
const unsigned uEdgeCount = Edges.GetCount();
if (0 == uEdgeCount)
break;
for (unsigned n = 0; n < uEdgeCount; ++n)
{
unsigned uNodeFrom;
unsigned uNodeTo;
Edges.GetEdge(n, &uNodeFrom, &uNodeTo);
CalcInfo(tree, uNodeFrom, uNodeTo, EIs);
#if TRACE
Log("Edge %u -> %u\n", uNodeFrom, uNodeTo);
#endif
const unsigned uNeighborCount = tree.GetNeighborCount(uNodeFrom);
for (unsigned i = 0; i < uNeighborCount; ++i)
{
const unsigned uNeighborIndex = tree.GetNeighbor(uNodeFrom, i);
if (!Known(tree, EIs, uNeighborIndex, uNodeFrom) &&
AllKnownOut(tree, EIs, uNeighborIndex, uNodeFrom))
NextEdges.Add(uNeighborIndex, uNodeFrom);
}
}
Edges.Copy(NextEdges);
}
#if TRACE
ListEIs(EIs, uNodeCount);
#endif
switch (RootMethod)
{
case ROOT_MidLongestSpan:
RootByMidLongestSpan(tree, EIs, ptruNode1, ptruNode2,
ptrdLength1, ptrdLength2);
break;
case ROOT_MinAvgLeafDist:
RootByMinAvgLeafDist(tree, EIs, ptruNode1, ptruNode2,
ptrdLength1, ptrdLength2);
break;
default:
Quit("Invalid RootMethod=%d", RootMethod);
}
for (unsigned uNodeIndex = 0; uNodeIndex < uNodeCount; ++uNodeIndex)
delete[] EIs[uNodeIndex];
delete[] EIs;
}
static void RootByMidLongestSpan(const Tree &tree, EdgeInfo **EIs,
unsigned *ptruNode1, unsigned *ptruNode2,
double *ptrdLength1, double *ptrdLength2)
{
const unsigned uNodeCount = tree.GetNodeCount();
unsigned uLeaf1 = NULL_NEIGHBOR;
unsigned uMostDistantLeaf = NULL_NEIGHBOR;
double dMaxDist = -VERY_LARGE_DOUBLE;
for (unsigned uNodeIndex = 0; uNodeIndex < uNodeCount; ++uNodeIndex)
{
if (!tree.IsLeaf(uNodeIndex))
continue;
const unsigned uNode2 = tree.GetNeighbor1(uNodeIndex);
if (NULL_NEIGHBOR == uNode2)
Quit("RootByMidLongestSpan: internal error 0");
const double dEdgeLength = tree.GetEdgeLength(uNodeIndex, uNode2);
const EdgeInfo &EI = EIs[uNodeIndex][0];
if (!EI.m_bSet)
Quit("RootByMidLongestSpan: internal error 1");
if (EI.m_uNode1 != uNodeIndex || EI.m_uNode2 != uNode2)
Quit("RootByMidLongestSpan: internal error 2");
const double dSpanLength = dEdgeLength + EI.m_dMaxDistToLeaf;
if (dSpanLength > dMaxDist)
{
dMaxDist = dSpanLength;
uLeaf1 = uNodeIndex;
uMostDistantLeaf = EI.m_uMostDistantLeaf;
}
}
if (NULL_NEIGHBOR == uLeaf1)
Quit("RootByMidLongestSpan: internal error 3");
const double dTreeHeight = dMaxDist/2.0;
unsigned uNode1 = uLeaf1;
unsigned uNode2 = tree.GetNeighbor1(uLeaf1);
double dAccumSpanLength = 0;
#if TRACE
Log("RootByMidLongestSpan: span=%u", uLeaf1);
#endif
for (;;)
{
const double dEdgeLength = tree.GetEdgeLength(uNode1, uNode2);
#if TRACE
Log("->%u(%g;%g)", uNode2, dEdgeLength, dAccumSpanLength);
#endif
if (dAccumSpanLength + dEdgeLength >= dTreeHeight)
{
*ptruNode1 = uNode1;
*ptruNode2 = uNode2;
*ptrdLength1 = dTreeHeight - dAccumSpanLength;
*ptrdLength2 = dEdgeLength - *ptrdLength1;
#if TRACE
{
const EdgeInfo &EI = EIs[uLeaf1][0];
Log("...\n");
Log("Midpoint: Leaf1=%u Leaf2=%u Node1=%u Node2=%u Length1=%g Length2=%g\n",
uLeaf1, EI.m_uMostDistantLeaf, *ptruNode1, *ptruNode2, *ptrdLength1, *ptrdLength2);
}
#endif
return;
}
if (tree.IsLeaf(uNode2))
Quit("RootByMidLongestSpan: internal error 4");
dAccumSpanLength += dEdgeLength;
const unsigned uSub = tree.GetNeighborSubscript(uNode1, uNode2);
const EdgeInfo &EI = EIs[uNode1][uSub];
if (!EI.m_bSet)
Quit("RootByMidLongestSpan: internal error 5");
uNode1 = uNode2;
uNode2 = EI.m_uMaxStep;
}
}
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;
}
static void RootByMinAvgLeafDist(const Tree &tree, EdgeInfo **EIs,
unsigned *ptruNode1, unsigned *ptruNode2,
double *ptrdLength1, double *ptrdLength2)
{
const unsigned uNodeCount = tree.GetNodeCount();
const unsigned uLeafCount = tree.GetLeafCount();
unsigned uNode1 = NULL_NEIGHBOR;
unsigned uNode2 = NULL_NEIGHBOR;
double dMinHeight = VERY_LARGE_DOUBLE;
double dBestLength1 = VERY_LARGE_DOUBLE;
double dBestLength2 = VERY_LARGE_DOUBLE;
for (unsigned uNodeIndex = 0; uNodeIndex < uNodeCount; ++uNodeIndex)
{
const unsigned uNeighborCount = tree.GetNeighborCount(uNodeIndex);
for (unsigned uSub = 0; uSub < uNeighborCount; ++uSub)
{
const unsigned uNeighborIndex = tree.GetNeighbor(uNodeIndex, uSub);
// Avoid visiting same edge a second time in reversed order.
if (uNeighborIndex < uNodeIndex)
continue;
const unsigned uSubRev = tree.GetNeighborSubscript(uNeighborIndex, uNodeIndex);
if (NULL_NEIGHBOR == uSubRev)
Quit("RootByMinAvgLeafDist, internal error 1");
// Get info for edges Node1->Node2 and Node2->Node1 (reversed)
const EdgeInfo &EI = EIs[uNodeIndex][uSub];
const EdgeInfo &EIRev = EIs[uNeighborIndex][uSubRev];
if (EI.m_uNode1 != uNodeIndex || EI.m_uNode2 != uNeighborIndex ||
EIRev.m_uNode1 != uNeighborIndex || EIRev.m_uNode2 != uNodeIndex)
Quit("RootByMinAvgLeafDist, internal error 2");
if (!EI.m_bSet)
Quit("RootByMinAvgLeafDist, internal error 3");
if (uLeafCount != EI.m_uLeafCount + EIRev.m_uLeafCount)
Quit("RootByMinAvgLeafDist, internal error 4");
const double dEdgeLength = tree.GetEdgeLength(uNodeIndex, uNeighborIndex);
if (dEdgeLength != tree.GetEdgeLength(uNeighborIndex, uNodeIndex))
Quit("RootByMinAvgLeafDist, internal error 5");
// Consider point p on edge 12 in tree (1=Node, 2=Neighbor).
//
// ----- ----
// | |
// 1----p--2
// | |
// ----- ----
//
// Define:
// ADLp = average distance to leaves to left of point p.
// ADRp = average distance to leaves to right of point p.
// L = edge length = distance 12
// x = distance 1p
// So distance p2 = L - x.
// Average distance from p to leaves on left of p is:
// ADLp = ADL1 + x
// Average distance from p to leaves on right of p is:
// ADRp = ADR2 + (L - x)
// To be a root, we require these two distances to be equal,
// ADLp = ADRp
// ADL1 + x = ADR2 + (L - x)
// Solving for x,
// x = (ADR2 - ADL1 + L)/2
// If 0 <= x <= L, we can place the root on edge 12.
const double ADL1 = EI.m_dTotalDistToLeaves / EI.m_uLeafCount;
const double ADR2 = EIRev.m_dTotalDistToLeaves / EIRev.m_uLeafCount;
const double x = (ADR2 - ADL1 + dEdgeLength)/2.0;
if (x >= 0 && x <= dEdgeLength)
{
const double dLength1 = x;
const double dLength2 = dEdgeLength - x;
const double dHeight1 = EI.m_dMaxDistToLeaf + dLength1;
const double dHeight2 = EIRev.m_dMaxDistToLeaf + dLength2;
const double dHeight = dHeight1 >= dHeight2 ? dHeight1 : dHeight2;
#if TRACE
Log("Candidate root Node1=%u Node2=%u Height=%g\n",
uNodeIndex, uNeighborIndex, dHeight);
#endif
if (dHeight < dMinHeight)
{
uNode1 = uNodeIndex;
uNode2 = uNeighborIndex;
dBestLength1 = dLength1;
dBestLength2 = dLength2;
dMinHeight = dHeight;
}
}
}
}
if (NULL_NEIGHBOR == uNode1 || NULL_NEIGHBOR == uNode2)
Quit("RootByMinAvgLeafDist, internal error 6");
#if TRACE
Log("Best root Node1=%u Node2=%u Length1=%g Length2=%g Height=%g\n",
uNode1, uNode2, dBestLength1, dBestLength2, dMinHeight);
#endif
*ptruNode1 = uNode1;
*ptruNode2 = uNode2;
*ptrdLength1 = dBestLength1;
*ptrdLength2 = dBestLength2;
}
