/// <summary>
/// Delete nodes marked for erosion. Update adjacent nodes to update their extension tables.
/// </summary>
/// <param name="graph">De Bruijn Graph.</param>
public void RemoveLowCoverageNodes(DeBruijnGraph graph)
{
//Basic strategy here, start at all reference nodes, go find everything that isn't in there
//and remove it.
DeBruijnGraph.ValidateGraph(graph);
//Mark all nodes as not visited
//Now visit everyone that is connected to the reference somehow
//Now mark any unvisited node for deletion.
if (Bio.CrossPlatform.Environment.GetRunningPlatform() != Bio.CrossPlatform.Environment.Platform.Mac)
{
Parallel.ForEach(graph.GetNodes(), new ParallelOptions()
{
MaxDegreeOfParallelism = Environment.ProcessorCount
}, x => {
if (x.KmerCount < CoverageCutOff)
{
x.MarkNodeForDelete();
}
});
Parallel.ForEach(
graph.GetNodes(),
(node) =>
{
node.RemoveMarkedExtensions();
});
}
else
{
foreach (var x in graph.GetNodes())
{
if (x.KmerCount < CoverageCutOff)
{
x.MarkNodeForDelete();
}
}
foreach (var node in
graph.GetNodes())
{
node.RemoveMarkedExtensions();
}
}
//Now to delete them, since they are not connected to anything we are keeping,
//no need to alter the graph structure
graph.RemoveMarkedNodes();
}
/// <summary>
/// Build contigs from graph. For contigs whose coverage is less than
/// the specified threshold, remove graph nodes belonging to them.
/// </summary>
/// <param name="deBruijnGraph">DeBruijn Graph.</param>
/// <param name="coverageThresholdForContigs">Coverage Threshold for contigs.</param>
/// <returns>Number of nodes removed.</returns>
public long RemoveLowCoverageContigs(DeBruijnGraph deBruijnGraph, double coverageThresholdForContigs)
{
if (deBruijnGraph == null)
{
throw new ArgumentNullException("deBruijnGraph");
}
if (coverageThresholdForContigs <= 0)
{
throw new ArgumentException("For removing low coverage contigs, coverage threshold should be a positive number");
}
this._coverageThreshold = coverageThresholdForContigs;
this._graph = deBruijnGraph;
DeBruijnGraph.ValidateGraph(deBruijnGraph);
this.ExcludeAmbiguousExtensions();
Parallel.ForEach(deBruijnGraph.GetNodes(), n => n.ComputeValidExtensions());
this.GetSimplePaths(false);
Parallel.ForEach(deBruijnGraph.GetNodes(), n => n.UndoAmbiguousExtensions());
return(deBruijnGraph.RemoveMarkedNodes());
}
/// <summary>
/// Delete nodes marked for erosion. Update adjacent nodes to update their extension tables.
/// </summary>
/// <param name="graph">De Bruijn Graph.</param>
public void RemoveUnconnectedNodes(DeBruijnGraph graph, IEnumerable <DeBruijnNode> referenceNodes)
{
//Basic strategy here, start at all reference nodes, go find everything that isn't in there
//and remove it.
DeBruijnGraph.ValidateGraph(graph);
//Mark all nodes as not visited
graph.SetNodeVisitState(false);
//Now visit everyone that is connected to the reference somehow
//This loop should spend basically all its time on the first node
foreach (DeBruijnNode node in referenceNodes)
{
if (node.IsVisited)
{
continue;
}
else
{
visitAllConnectedNodes(node);
}
}
//Now mark any unvisited node for deletion.
Parallel.ForEach(graph.GetNodes(), new ParallelOptions()
{
MaxDegreeOfParallelism = Environment.ProcessorCount
}, x => {
if (!x.IsVisited)
{
x.MarkNodeForDelete();
}
});
Parallel.ForEach(
graph.GetNodes(),
(node) =>
{
node.RemoveMarkedExtensions();
});
//Now to delete them, since they are not connected to anything we are keeping,
//no need to alter the graph structure
graph.RemoveMarkedNodes();
}
/// <summary>
/// Delete nodes marked for erosion. Update adjacent nodes to update their extension tables.
/// After nodes are deleted, some new end-points might be created. We need to check for
/// dangling links at these new points. This list is returned in the out parameter.
///
/// TODO: Perhaps refactor code so that the graph is only manipulated by itself?
/// Might make it easier to implement future performance improvements, or cost performance
/// </summary>
/// <param name="graph">De Bruijn Graph.</param>
private static IList <DeBruijnNode> RemoveErodedNodes(DeBruijnGraph graph)
{
bool eroded = graph.RemoveMarkedNodes() > 0;
IList <DeBruijnNode> graphNodes = null;
if (eroded)
{
graphNodes = graph.GetNodes().AsParallel().Where(n =>
{
bool wasEndPoint = (n.LeftExtensionNodesCount == 0 || n.RightExtensionNodesCount == 0);
n.RemoveMarkedExtensions();
// Check if this is a new end point.
return(wasEndPoint || (n.LeftExtensionNodesCount == 0 || n.RightExtensionNodesCount == 0));
}).ToList();
}
else
{
graphNodes = new List <DeBruijnNode>();
}
return(graphNodes);
}
/// <summary>
/// Delete nodes marked for erosion. Update adjacent nodes to update their extension tables.
/// </summary>
/// <param name="graph">De Bruijn Graph.</param>
public static int RemovePathologicalNodes(DeBruijnGraph graph)
{
//Basic strategy here, start at all reference nodes, go find everything that isn't in there
//and remove it.
DeBruijnGraph.ValidateGraph(graph);
var badSeq = Enumerable.Repeat((byte)'A', graph.KmerLength).ToArray();
var seq = new Bio.Sequence(Bio.Alphabets.DNA, badSeq, false);
var badkmer1 = KmerData32.GetKmers(seq, graph.KmerLength).First().KmerData;
badSeq = Enumerable.Repeat((byte)'G', graph.KmerLength).ToArray();
seq = new Bio.Sequence(Bio.Alphabets.DNA, badSeq, false);
var badkmer2 = KmerData32.GetKmers(seq, graph.KmerLength).First().KmerData;
var badNodeCount = 0;
foreach (var x in graph.GetNodes())
{
if (x.NodeValue.KmerData == badkmer1 ||
x.NodeValue.KmerData == badkmer2 ||
x.ContainsSelfReference)
{
x.MarkNodeForDelete();
Interlocked.Increment(ref badNodeCount);
}
}
foreach (var node in graph.GetNodes())
{
node.RemoveMarkedExtensions();
}
//Now to delete them, since they are not connected to anything we are keeping,
//no need to alter the graph structure
graph.RemoveMarkedNodes();
return(badNodeCount);
}
