/// <summary>
/// Depth first search (DFS) algorithm (recursive version).
/// Visit the subtree nodes from the current node, and
/// mark visited nodes by DFS with 'true'.
/// </summary>
/// <param name="node">root node of the subtree</param>
/// <param name="mark">Bool list of all nodes (in a tree)</param>
public static void DepthFirstSearch(BinaryGuideTreeNode node, List <bool> mark)
{
if (node == null)
{
throw new ArgumentNullException("node");
}
mark[node.ID] = true;
while (node.LeftChildren != null)
{
if (mark[node.LeftChildren.ID] == false)
{
DepthFirstSearch(node.LeftChildren, mark);
}
else
{
break;
}
}
while (node.RightChildren != null)
{
if (mark[node.RightChildren.ID] == false)
{
DepthFirstSearch(node.RightChildren, mark);
}
else
{
break;
}
}
}
/// <summary>
/// Compare two (sub)trees from root to leaves,
/// find the top node position that the subtrees of this node are
/// different between the two trees
///
/// Normally nodeA is the root of newly generated tree A, and nodeB is the root
/// of old tree B. This method returns the top node in tree A, so that the subtree
/// of this node will be re-aligned.
/// </summary>
/// <param name="nodeA">root of (sub)tree</param>
/// <param name="nodeB">root of (sub)tree</param>
public static BinaryGuideTreeNode FindSmallestTreeDifference(BinaryGuideTreeNode nodeA, BinaryGuideTreeNode nodeB)
{
if (nodeA == null)
{
throw new ArgumentNullException("nodeA");
}
if (nodeB == null)
{
throw new ArgumentNullException("nodeB");
}
//TODO the order of left/right child nodes do not have to be the same to be identical
if (nodeA != nodeB)
{
return(nodeA);
}
else
{
while (nodeA.LeftChildren != null && nodeB.LeftChildren != null)
{
return(FindSmallestTreeDifference(nodeA.LeftChildren, nodeB.LeftChildren));
}
while (nodeA.RightChildren != null && nodeB.RightChildren != null)
{
return(FindSmallestTreeDifference(nodeA.RightChildren, nodeB.RightChildren));
}
}
return(null);
}
/// <summary>
/// Create a tree with an assigned tree ID
/// </summary>
/// <param name="id">zero-based tree ID</param>
public BinaryGuideTree(int id)
{
_id = id;
_root = new BinaryGuideTreeNode();
_numberOfNodes = 1;
_numberOfLeaves = 1;
}
/// <summary>
/// Combine cluster nextA and nextB into a new cluster
/// </summary>
private void CreateCluster()
{
BinaryGuideTreeNode _node = new BinaryGuideTreeNode(++_currentClusterID);
// link the two nodes nextA and nextB with the new node
_node.LeftChildren = Nodes[_nextA];
_node.RightChildren = Nodes[_nextB];
Nodes[_nextA].Parent = _node;
Nodes[_nextB].Parent = _node;
// use the leftmost leave's sequenceID
_node.SequenceID = Nodes[_nextA].SequenceID;
Nodes.Add(_node);
BinaryGuideTreeEdge _edge1 = new BinaryGuideTreeEdge(Nodes[_nextA].ID);
BinaryGuideTreeEdge _edge2 = new BinaryGuideTreeEdge(Nodes[_nextB].ID);
_edge1.ParentNode = _node;
_edge2.ParentNode = _node;
_edge1.ChildNode = Nodes[_nextA];
_edge2.ChildNode = Nodes[_nextB];
// the length of the edge is the percent identity of two node sequences
// or the average of identities between two sets of sequences
//_edge1.Length = KimuraDistanceScoreCalculator.calculateDistanceScore(
// seqs[nodes[next1].sequenceID], seqs[nodes[next2].sequenceID]);
// modified: define kimura distance as sequence distance
_edge1.Length = _smallestDistance;
_edge2.Length = _smallestDistance;
_edges.Add(_edge1);
_edges.Add(_edge2);
}
/// <summary>
/// Create a tree using an existing node as root
/// </summary>
/// <param name="r">a BinaryGuideTreeNode</param>
public BinaryGuideTree(BinaryGuideTreeNode r)
{
if (r == null)
{
throw new ArgumentNullException("r");
}
_id = r.ID;
_root = r;
_numberOfNodes = 1;
_numberOfLeaves = 1;
}
/// <summary>
/// Construct a tree by hierarchical clustering method.
///
/// The node list is already generated in the hierarchical clustering method
/// and the root will be the last node in the list
/// </summary>
/// <param name="hCluster">hierarcical clustering class object</param>
public BinaryGuideTree(IHierarchicalClustering hCluster)
{
if (hCluster == null)
{
throw new ArgumentException("null Hierarchical clustering class");
}
if (hCluster.Nodes.Count == 0)
{
throw new ArgumentException("empty node list in Hierarchical clustering class");
}
_nodes = hCluster.Nodes;
_edges = hCluster.Edges;
_root = hCluster.Nodes[hCluster.Nodes.Count - 1];
_numberOfNodes = hCluster.Nodes.Count;
_numberOfLeaves = (_numberOfNodes + 2) / 2;
}
/// <summary>
/// Combine cluster nextA and nextB into a new cluster
/// </summary>
/// <param name="distanceMatrix">distance matrix</param>
private void CreateCluster(IDistanceMatrix distanceMatrix)
{
BinaryGuideTreeNode node = new BinaryGuideTreeNode(++_currentClusterID);
// link the two nodes nextA and nextB with the new node
node.LeftChildren = Nodes[_nextA];
node.RightChildren = Nodes[_nextB];
Nodes[_nextA].Parent = node;
Nodes[_nextB].Parent = node;
// use the leftmost leave's sequenceID
int next = Math.Min(_nextA, _nextB);
node.SequenceID = Nodes[next].SequenceID;
_indexToCluster[node.SequenceID] = _currentClusterID;
Nodes.Add(node);
// Add edges
BinaryGuideTreeEdge edgeA = new BinaryGuideTreeEdge(Nodes[_nextA].ID);
BinaryGuideTreeEdge edgeB = new BinaryGuideTreeEdge(Nodes[_nextB].ID);
edgeA.ParentNode = node;
edgeB.ParentNode = node;
edgeA.ChildNode = Nodes[_nextA];
edgeB.ChildNode = Nodes[_nextB];
Nodes[_nextA].ParentEdge = edgeA;
Nodes[_nextB].ParentEdge = edgeB;
// the length of the edge is the percent identity of two node sequences
// or the average of identities between two sets of sequences
//_edge1.Length = KimuraDistanceScoreCalculator.calculateDistanceScore(
// seqs[nodes[nextA].sequenceID], seqs[nodes[nextB].sequenceID]);
// modified: define kimura distance as sequence distance
edgeA.Length = _smallestDistance;
edgeB.Length = _smallestDistance;
_edges.Add(edgeA);
_edges.Add(edgeB);
}
/// <summary>
/// Use the input node as root, return the subtree *leaf* nodes of this node,
/// and keep them in the alignment order by progressive aligner.
/// </summary>
/// <param name="node">root node in a (sub)tree</param>
public List <BinaryGuideTreeNode> ExtractSubTreeLeafNodes(BinaryGuideTreeNode node)
{
// mark whether the node is descendent of this node
List <bool> mark = new List <bool>(_numberOfNodes);
for (int i = 0; i < _numberOfNodes; ++i)
{
mark.Add(false);
}
// Mark the subtree nodes by depth first search algorithm
DepthFirstSearch(node, mark);
List <BinaryGuideTreeNode> result = new List <BinaryGuideTreeNode>();
for (int i = 0; i < mark.Count; ++i)
{
if (mark[i] && _nodes[i].IsLeaf)
{
result.Add(_nodes[i]);
}
}
return(result);
}
/// <summary>
/// Construct a tree with one root node
/// </summary>
public BinaryGuideTree()
{
_root = new BinaryGuideTreeNode();
_numberOfNodes = 1;
_numberOfLeaves = 1;
}
/// <summary>
/// Construct an edge with an assigned ID.
/// </summary>
/// <param name="id">zero-based edge ID</param>
public BinaryGuideTreeEdge(int id)
{
_id = id;
ParentNode = null;
ChildNode = null;
}
/// <summary>
/// Create a tree using an existing node as root
/// </summary>
/// <param name="r">a BinaryGuideTreeNode</param>
public BinaryGuideTree(BinaryGuideTreeNode r)
{
if (r == null)
{
throw new ArgumentNullException("r");
}
_id = r.ID;
_root = r;
_numberOfNodes = 1;
_numberOfLeaves = 1;
}
/// <summary>
/// Main pregressive alignment algorithm aligns a set of sequences guided by
/// a binary tree.
/// </summary>
/// <param name="sequences">input sequences</param>
/// <param name="tree">a binary guide tree</param>
public void Align(IList <ISequence> sequences, BinaryGuideTree tree)
{
SequenceWeighting sequenceWeighting = null;
if (PAMSAMMultipleSequenceAligner.UseWeights)
{
sequenceWeighting = new SequenceWeighting(tree);
/*
* for (int i = 0; i < sequenceWeighting.Weights.Length; ++i)
* {
* sequenceWeighting.Weights[i] = 1;
* }
*/
}
if (sequences.Count == 0)
{
throw new ArgumentException("Empty set of sequences");
}
IAlphabet alphabet = sequences[0].Alphabet;
Parallel.For(1, sequences.Count, PAMSAMMultipleSequenceAligner.parallelOption, i =>
{
if (!Alphabets.CheckIsFromSameBase(sequences[i].Alphabet, alphabet))
{
throw new ArgumentException("Inconsistent sequence alphabet");
}
});
if (PAMSAMMultipleSequenceAligner.UseWeights)
{
// Generate profile for leaf nodes
Parallel.For(0, sequences.Count, PAMSAMMultipleSequenceAligner.parallelOption, i =>
{
tree.Nodes[i].ProfileAlignment = ProfileAlignment.GenerateProfileAlignment(sequences[i], sequenceWeighting.Weights[i]);
tree.Nodes[i].Weight = sequenceWeighting.Weights[i];
});
}
else
{
// Generate profile for leaf nodes
Parallel.For(0, sequences.Count, PAMSAMMultipleSequenceAligner.parallelOption, i =>
{
tree.Nodes[i].ProfileAlignment = ProfileAlignment.GenerateProfileAlignment(sequences[i]);
});
}
// Iterate internal nodes;
// as defined in the tree, the last node is the root
for (int i = sequences.Count; i < tree.Nodes.Count; ++i)
{
if (tree.Nodes[i].NeedReAlignment)
{
// pull out its children
_nodeA = tree.Nodes[i].LeftChildren;
_nodeB = tree.Nodes[i].RightChildren;
if (PAMSAMMultipleSequenceAligner.UseWeights)
{
_profileAligner.Weights = new float[2];
_profileAligner.Weights[0] = _nodeA.Weight;
_profileAligner.Weights[1] = _nodeB.Weight;
tree.Nodes[i].Weight = _nodeA.Weight + _nodeB.Weight;
}
// align two profiles
ProfileAlignment result = null;
if (_nodeA.ProfileAlignment.NumberOfSequences < _nodeB.ProfileAlignment.NumberOfSequences)
{
result = (ProfileAlignment)_profileAligner.Align(
_nodeA.ProfileAlignment, _nodeB.ProfileAlignment);
// assign aligned profiles to the current internal node
tree.Nodes[i].ProfileAlignment = result;
// generate eString for the children nodes
_nodeA.EString = _profileAligner.GenerateEString(_profileAligner.AlignedA);
_nodeB.EString = _profileAligner.GenerateEString(_profileAligner.AlignedB);
}
else
{
result = (ProfileAlignment)_profileAligner.Align(
_nodeB.ProfileAlignment, _nodeA.ProfileAlignment);
// assign aligned profiles to the current internal node
tree.Nodes[i].ProfileAlignment = result;
// generate eString for the children nodes
_nodeA.EString = _profileAligner.GenerateEString(_profileAligner.AlignedB);
_nodeB.EString = _profileAligner.GenerateEString(_profileAligner.AlignedA);
}
// children node profiles can be deleted
_nodeA.ProfileAlignment.Clear();
_nodeB.ProfileAlignment.Clear();
}
}
// Convert original unaligned sequences to aligned ones by applying alignment paths in eStrings
try
{
_alignedSequences = new List <ISequence>(sequences.Count);
}
catch (OutOfMemoryException ex)
{
throw new Exception("Out of memory", ex.InnerException);
}
for (int i = 0; i < sequences.Count; ++i)
{
_alignedSequences.Add(null);
}
Parallel.For(0, sequences.Count, PAMSAMMultipleSequenceAligner.parallelOption, i =>
{
ISequence seq = sequences[i];
BinaryGuideTreeNode node;
node = tree.Nodes[i];
while (!node.IsRoot)
{
seq = _profileAligner.GenerateSequenceFromEString(node.EString, seq);
node = node.Parent;
}
_alignedSequences[i] = seq;
});
}
/// <summary>
/// Construct an empty edge.
/// </summary>
public BinaryGuideTreeEdge()
{
_id = -1;
ParentNode = null;
ChildNode = null;
}
/// <summary>
/// Construct an empty edge.
/// </summary>
public BinaryGuideTreeEdge()
{
_id = -1;
ParentNode = null;
ChildNode = null;
}
/// <summary>
/// Construct an edge with an assigned ID.
/// </summary>
/// <param name="id">zero-based edge ID</param>
public BinaryGuideTreeEdge(int id)
{
_id = id;
ParentNode = null;
ChildNode = null;
}
/// <summary>
/// Construct a tree by hierarchical clustering method.
///
/// The node list is already generated in the hierarchical clustering method
/// and the root will be the last node in the list
/// </summary>
/// <param name="hCluster">hierarcical clustering class object</param>
public BinaryGuideTree(IHierarchicalClustering hCluster)
{
if (hCluster == null)
{
throw new ArgumentException("null Hierarchical clustering class");
}
if (hCluster.Nodes.Count == 0)
{
throw new ArgumentException("empty node list in Hierarchical clustering class");
}
_nodes = hCluster.Nodes;
_edges = hCluster.Edges;
_root = hCluster.Nodes[hCluster.Nodes.Count - 1];
_numberOfNodes = hCluster.Nodes.Count;
_numberOfLeaves = (_numberOfNodes + 2) / 2;
}
/// <summary>
/// Main pregressive alignment algorithm aligns a set of sequences guided by
/// a binary tree.
/// </summary>
/// <param name="sequences">input sequences</param>
/// <param name="tree">a binary guide tree</param>
public void Align(IList<ISequence> sequences, BinaryGuideTree tree)
{
SequenceWeighting sequenceWeighting = null;
if (PAMSAMMultipleSequenceAligner.UseWeights)
{
sequenceWeighting = new SequenceWeighting(tree);
/*
for (int i = 0; i < sequenceWeighting.Weights.Length; ++i)
{
sequenceWeighting.Weights[i] = 1;
}
*/
}
if (sequences.Count==0)
{
throw new ArgumentException("Empty set of sequences");
}
IAlphabet alphabet = sequences[0].Alphabet;
Parallel.For(1, sequences.Count, PAMSAMMultipleSequenceAligner.ParallelOption, i =>
{
if (!Alphabets.CheckIsFromSameBase(sequences[i].Alphabet, alphabet))
{
throw new ArgumentException("Inconsistent sequence alphabet");
}
});
if (PAMSAMMultipleSequenceAligner.UseWeights)
{
// Generate profile for leaf nodes
Parallel.For(0, sequences.Count, PAMSAMMultipleSequenceAligner.ParallelOption, i =>
{
tree.Nodes[i].ProfileAlignment = ProfileAlignment.GenerateProfileAlignment(sequences[i], sequenceWeighting.Weights[i]);
tree.Nodes[i].Weight = sequenceWeighting.Weights[i];
});
}
else
{
// Generate profile for leaf nodes
Parallel.For(0, sequences.Count, PAMSAMMultipleSequenceAligner.ParallelOption, i =>
{
tree.Nodes[i].ProfileAlignment = ProfileAlignment.GenerateProfileAlignment(sequences[i]);
});
}
// Iterate internal nodes;
// as defined in the tree, the last node is the root
for (int i = sequences.Count; i < tree.Nodes.Count; ++i)
{
if (tree.Nodes[i].NeedReAlignment)
{
// pull out its children
_nodeA = tree.Nodes[i].LeftChildren;
_nodeB = tree.Nodes[i].RightChildren;
if (PAMSAMMultipleSequenceAligner.UseWeights)
{
_profileAligner.Weights = new float[2];
_profileAligner.Weights[0] = _nodeA.Weight;
_profileAligner.Weights[1] = _nodeB.Weight;
tree.Nodes[i].Weight = _nodeA.Weight + _nodeB.Weight;
}
// align two profiles
ProfileAlignment result = null;
if (_nodeA.ProfileAlignment.NumberOfSequences < _nodeB.ProfileAlignment.NumberOfSequences)
{
result = (ProfileAlignment)_profileAligner.Align(
_nodeA.ProfileAlignment, _nodeB.ProfileAlignment);
// assign aligned profiles to the current internal node
tree.Nodes[i].ProfileAlignment = result;
// generate eString for the children nodes
_nodeA.EString = _profileAligner.GenerateEString(_profileAligner.AlignedA);
_nodeB.EString = _profileAligner.GenerateEString(_profileAligner.AlignedB);
}
else
{
result = (ProfileAlignment)_profileAligner.Align(
_nodeB.ProfileAlignment, _nodeA.ProfileAlignment);
// assign aligned profiles to the current internal node
tree.Nodes[i].ProfileAlignment = result;
// generate eString for the children nodes
_nodeA.EString = _profileAligner.GenerateEString(_profileAligner.AlignedB);
_nodeB.EString = _profileAligner.GenerateEString(_profileAligner.AlignedA);
}
// children node profiles can be deleted
_nodeA.ProfileAlignment.Clear();
_nodeB.ProfileAlignment.Clear();
}
}
// Convert original unaligned sequences to aligned ones by applying alignment paths in eStrings
try
{
_alignedSequences = new List<ISequence>(sequences.Count);
}
catch (OutOfMemoryException ex)
{
throw new Exception("Out of memory", ex.InnerException);
}
for (int i=0; i<sequences.Count;++i)
{
_alignedSequences.Add(null);
}
Parallel.For(0, sequences.Count, PAMSAMMultipleSequenceAligner.ParallelOption, i =>
{
ISequence seq = sequences[i];
BinaryGuideTreeNode node;
node = tree.Nodes[i];
while (!node.IsRoot)
{
seq = _profileAligner.GenerateSequenceFromEString(node.EString, seq);
node = node.Parent;
}
_alignedSequences[i] = seq;
});
}
/// <summary>
/// Use the input node as root, return the subtree *leaf* nodes of this node,
/// and keep them in the alignment order by progressive aligner.
/// </summary>
/// <param name="node">root node in a (sub)tree</param>
public List<BinaryGuideTreeNode> ExtractSubTreeLeafNodes(BinaryGuideTreeNode node)
{
// mark whether the node is descendent of this node
List<bool> mark = new List<bool>(_numberOfNodes);
for (int i = 0; i < _numberOfNodes; ++i)
{
mark.Add(false);
}
// Mark the subtree nodes by depth first search algorithm
DepthFirstSearch(node, mark);
List<BinaryGuideTreeNode> result = new List<BinaryGuideTreeNode>();
for (int i = 0; i < mark.Count; ++i)
{
if (mark[i] && _nodes[i].IsLeaf)
{
result.Add(_nodes[i]);
}
}
return result;
}
/// <summary>
/// Create a tree with an assigned tree ID
/// </summary>
/// <param name="id">zero-based tree ID</param>
public BinaryGuideTree(int id)
{
_id = id;
_root = new BinaryGuideTreeNode();
_numberOfNodes = 1;
_numberOfLeaves = 1;
}
/// <summary>
/// Depth first search (DFS) algorithm (recursive version).
/// Visit the subtree nodes from the current node, and
/// mark visited nodes by DFS with 'true'.
/// </summary>
/// <param name="node">root node of the subtree</param>
/// <param name="mark">Bool list of all nodes (in a tree)</param>
public static void DepthFirstSearch(BinaryGuideTreeNode node, List<bool> mark)
{
if (node == null)
{
throw new ArgumentNullException("node");
}
mark[node.ID] = true;
while (node.LeftChildren != null)
{
if (mark[node.LeftChildren.ID] == false)
{
DepthFirstSearch(node.LeftChildren, mark);
}
else
{
break;
}
}
while (node.RightChildren != null)
{
if (mark[node.RightChildren.ID] == false)
{
DepthFirstSearch(node.RightChildren, mark);
}
else
{
break;
}
}
}
/// <summary>
/// Compare two (sub)trees from root to leaves,
/// find the top node position that the subtrees of this node are
/// different between the two trees
///
/// Normally nodeA is the root of newly generated tree A, and nodeB is the root
/// of old tree B. This method returns the top node in tree A, so that the subtree
/// of this node will be re-aligned.
/// </summary>
/// <param name="nodeA">root of (sub)tree</param>
/// <param name="nodeB">root of (sub)tree</param>
public static BinaryGuideTreeNode FindSmallestTreeDifference(BinaryGuideTreeNode nodeA, BinaryGuideTreeNode nodeB)
{
if (nodeA == null)
{
throw new ArgumentNullException("nodeA");
}
if (nodeB == null)
{
throw new ArgumentNullException("nodeB");
}
//TODO the order of left/right child nodes do not have to be the same to be identical
if (nodeA != nodeB)
{
return nodeA;
}
else
{
while (nodeA.LeftChildren != null && nodeB.LeftChildren != null)
{
return FindSmallestTreeDifference(nodeA.LeftChildren, nodeB.LeftChildren);
}
while (nodeA.RightChildren != null && nodeB.RightChildren != null)
{
return FindSmallestTreeDifference(nodeA.RightChildren, nodeB.RightChildren);
}
}
return null;
}
/// <summary>
/// Construct a tree with one root node
/// </summary>
public BinaryGuideTree()
{
_root = new BinaryGuideTreeNode();
_numberOfNodes = 1;
_numberOfLeaves = 1;
}
