/// <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(Node node)
{
if (node == null)
{
throw new ArgumentNullException("node");
}
// Get sequence index and validate.
int sequenceIndex = node.SequenceIndex;
if (sequenceIndex < 0 || sequenceIndex >= this.baseSequences.Count)
{
throw new ArgumentOutOfRangeException("node", Properties.Resource.KmerIndexOutOfRange);
}
// Get base sequence, position and validate.
return this.baseSequences[sequenceIndex];
}
/// <summary>
/// Check if input node is null.
/// </summary>
/// <param name="node">Input node.</param>
private static void ValidateNode(Node node)
{
if (node == null)
{
throw new ArgumentNullException("node");
}
}
/// <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(Node node, bool isSameOrientation)
{
ValidateNode(node);
Edge edge;
if (this.rightEndExtensionNodes.TryGetValue(node, out edge))
{
this.rightEndExtensionNodes[node].IsSameOrientation ^= isSameOrientation;
}
else
{
this.rightEndExtensionNodes[node] = new Edge(isSameOrientation);
}
}
/// <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
Node[] contigNodes = new Node[contigs.Count()];
Parallel.For(0, contigs.Count, (int ndx) => contigNodes[ndx] = new Node(contigs[ndx].Count, ndx));
GenerateContigAdjacency(contigs, kmerLength, contigNodes);
// Update graph with new nodes
this.baseSequences = new List<ISequence>(contigs);
this.kmerNodes = new HashSet<Node>(contigNodes);
}
/// <summary>
/// Checks for and adds edges between contigs
/// based on left, right kmer maps.
/// </summary>
/// <param name="contigNodes">Array of contig nodes.</param>
/// <param name="leftKmerMap">Map of left k-mer to contig nodes.</param>
/// <param name="rightKmerMap">Map of right k-mer to contig nodes.</param>
private static void AddContigGraphEdges(
Node[] contigNodes,
Dictionary<ISequence, List<int>> leftKmerMap,
Dictionary<ISequence, List<int>> rightKmerMap)
{
// Check and add left extensions. No locks used here since each iteration works with a different contigNode.
Parallel.ForEach(
leftKmerMap,
leftKmer =>
{
List<int> positions;
if (rightKmerMap.TryGetValue(leftKmer.Key, out positions))
{
foreach (int leftNodeIndex in leftKmer.Value)
{
foreach (int rightNodeIndex in positions)
{
contigNodes[leftNodeIndex].AddLeftEndExtension(contigNodes[rightNodeIndex], true);
}
}
}
if (leftKmerMap.TryGetValue(leftKmer.Key.GetReverseComplementedSequence(), out positions))
{
foreach (int leftNodeIndex in leftKmer.Value)
{
foreach (int rightNodeIndex in positions)
{
contigNodes[leftNodeIndex].AddLeftEndExtension(contigNodes[rightNodeIndex], false);
}
}
}
});
// Check and add right extensions. No locks used here since each iteration works with a different contigNode.
Parallel.ForEach(
rightKmerMap,
rightKmer =>
{
List<int> positions;
if (leftKmerMap.TryGetValue(rightKmer.Key, out positions))
{
foreach (int rightNodeIndex in rightKmer.Value)
{
foreach (int leftNodeIndex in positions)
{
contigNodes[rightNodeIndex].AddRightEndExtension(contigNodes[leftNodeIndex], true);
}
}
}
if (rightKmerMap.TryGetValue(rightKmer.Key.GetReverseComplementedSequence(), out positions))
{
foreach (int rightNodeIndex in rightKmer.Value)
{
foreach (int leftNodeIndex in positions)
{
contigNodes[rightNodeIndex].AddRightEndExtension(contigNodes[leftNodeIndex], false);
}
}
}
});
}
/// <summary>
/// Generate adjacency information between contig nodes
/// by computing overlapping regions between contig sequences.
/// </summary>
/// <param name="contigs">List of contig data.</param>
/// <param name="kmerLength">Kmer length.</param>
/// <param name="contigNodes">Array of contig nodes.</param>
private static void GenerateContigAdjacency(IList<ISequence> contigs, long kmerLength, Node[] contigNodes)
{
// Create dictionaries that map (k-1) left and right substrings of contigs to contig indexes.
Dictionary<ISequence, List<int>> leftKmerMap = new Dictionary<ISequence, List<int>>(new SequenceEqualityComparer());
Dictionary<ISequence, List<int>> rightKmerMap = new Dictionary<ISequence, List<int>>(new SequenceEqualityComparer());
Parallel.For(
0,
contigs.Count,
ndx =>
{
ISequence contig = contigs[ndx];
List<int> contigIndexes;
ISequence kmer;
if (contig.Count < kmerLength)
{
throw new ArgumentException(Properties.Resource.KmerLengthIsTooLong);
}
// update left map
kmer = contig.GetSubSequence(0, kmerLength - 1);
lock (leftKmerMap)
{
if (!leftKmerMap.TryGetValue(kmer, out contigIndexes))
{
contigIndexes = new List<int>();
leftKmerMap.Add(kmer, contigIndexes);
}
}
lock (contigIndexes)
{
contigIndexes.Add(ndx);
}
// Update right map
kmer = contig.GetSubSequence(contig.Count - (kmerLength - 1), kmerLength - 1);
lock (rightKmerMap)
{
if (!rightKmerMap.TryGetValue(kmer, out contigIndexes))
{
contigIndexes = new List<int>();
rightKmerMap.Add(kmer, contigIndexes);
}
}
lock (contigIndexes)
{
contigIndexes.Add(ndx);
}
});
AddContigGraphEdges(contigNodes, leftKmerMap, rightKmerMap);
}
/// <summary>
/// Performs Breadth First Search.
/// </summary>
/// <param name="node">Start Node.</param>
/// <param name="contigPairedReadMap">Map of all contigs having valid
/// mate pairs with given node contig.</param>
/// <returns>List of paths.</returns>
private List<ScaffoldPath> TraverseGraph(
Node node,
Dictionary<ISequence, IList<ValidMatePair>> contigPairedReadMap)
{
Queue<Paths> search = new Queue<Paths>();
List<Paths> paths = new List<Paths>();
this.LeftExtension(
new KeyValuePair<Node, Edge>(node, new Edge(false)),
search,
paths,
null,
contigPairedReadMap);
this.RightExtension(
new KeyValuePair<Node, Edge>(node, new Edge(true)),
search,
paths,
null,
contigPairedReadMap);
Paths parentPath;
while (search.Count != 0)
{
parentPath = search.Dequeue();
if (parentPath.NodeOrientation)
{
if (parentPath.CurrentNode.Value.IsSameOrientation)
{
this.RightExtension(
parentPath.CurrentNode,
search,
paths,
parentPath.FamilyTree,
contigPairedReadMap);
}
else
{
this.LeftExtension(
parentPath.CurrentNode,
search,
paths,
parentPath.FamilyTree,
contigPairedReadMap);
}
}
else if (parentPath.CurrentNode.Value.IsSameOrientation)
{
this.LeftExtension(
parentPath.CurrentNode,
search,
paths,
parentPath.FamilyTree,
contigPairedReadMap);
}
else
{
this.RightExtension(
parentPath.CurrentNode,
search,
paths,
parentPath.FamilyTree,
contigPairedReadMap);
}
}
return new List<ScaffoldPath>(paths.Select(t => t.FamilyTree));
}