/// <summary>
/// Check if input node is null
/// </summary>
/// <param name="node">Input node</param>
private static void ValidateNode(DeBruijnNode node)
{
if (node == null)
{
throw new ArgumentNullException("node");
}
}
/// <summary>
/// Gets the sequence for kmer associated with input node.
/// Uses index and position information along with base sequence
/// to construct sequence.
/// There should be atleast one valid position in the node.
/// Since all positions indicate the same kmer sequence,
/// the position information from the first kmer is used
/// to construct the sequence
/// </summary>
/// <param name="node">Graph Node</param>
/// <returns>Sequence associated with input node</returns>
public ISequence GetNodeSequence(DeBruijnNode node)
{
if (node == null)
{
throw new ArgumentNullException("node");
}
// Get sequence index and validate
int sequenceIndex = node.SequenceIndex;
if (sequenceIndex < 0 || sequenceIndex >= _baseSequence.Count)
{
throw new ArgumentOutOfRangeException("node", Properties.Resource.KmerIndexOutOfRange);
}
// Get base sequence, position and validate
ISequence baseSequence = _baseSequence[sequenceIndex];
int position = node.KmerPosition;
if (position < 0 || position + node.KmerLength > baseSequence.Count)
{
throw new ArgumentOutOfRangeException("node", Properties.Resource.KmerPositionOutOfRange);
}
if (position == 0 && baseSequence.Count == node.KmerLength)
{
return(baseSequence);
}
return(baseSequence.Range(position, node.KmerLength));
}
/// <summary>
/// Makes extension edge corresponding to the node invalid,
/// after checking whether given node is part of left or right extensions.
/// Not Thread-safe. Use lock at caller if required.
/// </summary>
/// <param name="node">Node for which extension is to be made invalid</param>
public void MarkExtensionInvalid(DeBruijnNode node)
{
ValidateNode(node);
if (_rightEndExtensionNodes.ContainsKey(node))
{
_rightEndExtensionNodes[node].IsValid = false;
}
else if (_leftEndExtensionNodes.ContainsKey(node))
{
_leftEndExtensionNodes[node].IsValid = false;
}
}
/// <summary>
/// Add node with given orientation to right extension edges.
/// Not thread-safe. Use lock at caller if required.
/// </summary>
/// <param name="node">Node to add right-extension to</param>
/// <param name="isSameOrientation">Orientation of connecting edge</param>
public void AddRightEndExtension(DeBruijnNode node, bool isSameOrientation)
{
ValidateNode(node);
DeBruijnEdge edge;
if (_rightEndExtensionNodes.TryGetValue(node, out edge))
{
_rightEndExtensionNodes[node].IsSameOrientation ^= isSameOrientation;
}
else
{
_rightEndExtensionNodes[node] = new DeBruijnEdge(isSameOrientation);
}
}
/// <summary>
/// Removes edge corresponding to the node from appropriate data structure,
/// after checking whether given node is part of left or right extensions.
/// Thread-safe method
/// </summary>
/// <param name="node">Node for which extension is to be removed</param>
public void RemoveExtensionThreadSafe(DeBruijnNode node)
{
ValidateNode(node);
bool removed;
lock (_rightEndExtensionNodes)
{
removed = _rightEndExtensionNodes.Remove(node);
}
if (!removed)
{
lock (_leftEndExtensionNodes)
{
_leftEndExtensionNodes.Remove(node);
}
}
}
/// <summary>
/// Builds a contig graph from kmer graph using contig data information.
/// Creates a graph node for each contig, computes adjacency
/// for contig graph using edge information in kmer graph.
/// Finally, all kmer nodes are deleted from the graph.
/// </summary>
/// <param name="contigs">List of contig data</param>
/// <param name="kmerLength">Kmer length</param>
public void BuildContigGraph(IList <ISequence> contigs, int kmerLength)
{
if (contigs == null)
{
throw new ArgumentNullException("contigs");
}
if (kmerLength <= 0)
{
throw new ArgumentException(Properties.Resource.KmerLengthShouldBePositive);
}
// Create contig nodes
DeBruijnNode[] contigNodes = new DeBruijnNode[contigs.Count];
Parallel.For(0, contigs.Count, ndx => contigNodes[ndx] = new DeBruijnNode(contigs[ndx].Count, ndx));
GenerateContigAdjacency(contigs, kmerLength, contigNodes);
// Update graph with new nodes
_baseSequence = new List <ISequence>(contigs);
_kmerNodes = new HashSet <DeBruijnNode>(contigNodes);
}
/// <summary>
/// Generate adjacency information between nodes
/// by computing overlapping regions between sequences.
/// </summary>
/// <param name="kmerNodeMap">Graph Nodes mapped by sequence string</param>
/// <param name="kmerLength">Kmer Length</param>
private void GenerateAdjacency(Dictionary <string, KmerDataGraphNodePair> kmerNodeMap, int kmerLength)
{
// Nothing to do if there are no nodes.
if (kmerNodeMap.Count == 0)
{
return;
}
// All nodes have sequences of equal length.
// Hence, obtained once and stored for optimization.
using (ThreadLocal <Tuple <char[], char[]> > kmerBuilders = new ThreadLocal <Tuple <char[], char[]> >
(() => Tuple.Create <char[], char[]>(new char[kmerLength], new char[kmerLength])))
{
Parallel.ForEach(kmerNodeMap, nodeValue =>
{
bool orientation = nodeValue.Value.KeyHasSameOrientation;
string kmerString;
string kmerStringRC;
if (orientation)
{
kmerString = nodeValue.Key;
kmerStringRC = kmerString.GetReverseComplement(kmerBuilders.Value.Item1);
}
else
{
kmerStringRC = nodeValue.Key;
kmerString = kmerStringRC.GetReverseComplement(kmerBuilders.Value.Item1);
}
DeBruijnNode node = nodeValue.Value.Node;
char[] nextKmer = kmerBuilders.Value.Item1;
char[] nextKmerRC = kmerBuilders.Value.Item2;
// Query its possible four extensions in the right (forward)
// If it exists, set 'right' edge information in current node
// The kmer sequence of right extension nodes should either start with
// (k-1) length right-substring or ends with its reverse complement.
// Get required substring from current node's sequence, and
// add a dummy character to make length equal to k.
kmerString.CopyTo(1, nextKmer, 0, kmerLength - 1); // right sub-string
kmerStringRC.CopyTo(0, nextKmerRC, 1, kmerLength - 1); // reverse-complement
for (int i = 0; i < DnaSymbols.Length; i++)
{
nextKmer[kmerLength - 1] = DnaSymbols[i]; // replace last character with dnaChar
KmerDataGraphNodePair nextNode;
if (kmerNodeMap.TryGetValue(new string(nextKmer), out nextNode)) // check if the kmer exists
{
// Add right extension with orientation set to true
// Ok to use unsafe add method since each parallel thread works with a different node
node.AddRightEndExtension(nextNode.Node, nextNode.KeyHasSameOrientation);
}
else
{
nextKmerRC[0] = DnaSymbolsComplement[i];
if (kmerNodeMap.TryGetValue(new string(nextKmerRC), out nextNode))
{
// Add right extension with orientation set to false
// Ok to use unsafe add method since each parallel thread works with a different node
node.AddRightEndExtension(nextNode.Node, !nextNode.KeyHasSameOrientation);
}
}
}
// Repeat above exercise for left extensions
// The kmer sequence of left extension nodes should either end with
// (k-1) length left-substring or ends with its reverse complement.
// Get required substring from current node's sequence, and
// add a dummy character to make length equal to k.
kmerString.CopyTo(0, nextKmer, 1, kmerLength - 1);
kmerStringRC.CopyTo(1, nextKmerRC, 0, kmerLength - 1);
for (int i = 0; i < DnaSymbols.Length; i++)
{
nextKmer[0] = DnaSymbols[i]; // replace first character with new DNA character
KmerDataGraphNodePair nextNode;
if (kmerNodeMap.TryGetValue(new string(nextKmer), out nextNode)) // check if the kmer exists
{
// Add left extension with orientation set to true
node.AddLeftEndExtension(nextNode.Node, nextNode.KeyHasSameOrientation);
}
else
{
nextKmerRC[kmerLength - 1] = DnaSymbolsComplement[i];
if (kmerNodeMap.TryGetValue(new string(nextKmerRC), out nextNode))
{
// Add left extension with orientation set to false
node.AddLeftEndExtension(nextNode.Node, !nextNode.KeyHasSameOrientation);
}
}
}
});
}
}
/// <summary>
/// Initializes a new instance of the DeBruijnPath class with specified node.
/// </summary>
/// <param name="node">Graph node</param>
public DeBruijnPath(DeBruijnNode node)
{
_path = new List <DeBruijnNode> {
node
};
}
