@Test(dataProvider = "MeetsMinSequenceData")
public void testSplitterCompleteCycle(
final List<String> mids,
final int minSeqLength,
final boolean prefixMeets,
final boolean suffixMeets) {
final SeqGraph graph = new SeqGraph(11);
final SeqVertex top = new SeqVertex("AAAAAAAA");
final SeqVertex bot = new SeqVertex("GGGGGGGG");
final List<SeqVertex> v = new ArrayList<>();
for (final String s : mids) {
v.add(new SeqVertex(s));
}
graph.addVertices(v.toArray(new SeqVertex[v.size()]));
graph.addVertices(top, bot);
for (final SeqVertex vi : v) {
graph.addEdge(top, vi);
graph.addEdge(vi, bot);
}
final SharedVertexSequenceSplitter splitter = new SharedVertexSequenceSplitter(graph, v);
Assert.assertEquals(
splitter.meetsMinMergableSequenceForPrefix(minSeqLength), prefixMeets, "Prefix failed");
Assert.assertEquals(
splitter.meetsMinMergableSequenceForSuffix(minSeqLength), suffixMeets, "Suffix failed");
Assert.assertEquals(
splitter.meetsMinMergableSequenceForEitherPrefixOrSuffix(minSeqLength),
suffixMeets || prefixMeets,
"Either prefix or suffix failed");
}
@Test(dataProvider = "PrefixSuffixData")
public void testSplitter(
final List<String> strings, int expectedPrefixLen, int expectedSuffixLen) {
final SeqGraph graph = new SeqGraph(11);
final List<SeqVertex> v = new ArrayList<>();
for (final String s : strings) {
v.add(new SeqVertex(s));
}
graph.addVertices(v.toArray(new SeqVertex[v.size()]));
final String expectedPrefix = strings.get(0).substring(0, expectedPrefixLen);
final String expectedSuffix =
strings.get(0).substring(strings.get(0).length() - expectedSuffixLen);
final SharedVertexSequenceSplitter splitter = new SharedVertexSequenceSplitter(graph, v);
splitter.split();
Assert.assertEquals(splitter.prefixV.getSequenceString(), expectedPrefix);
Assert.assertEquals(splitter.suffixV.getSequenceString(), expectedSuffix);
Assert.assertTrue(splitter.splitGraph.outDegreeOf(splitter.prefixV) <= strings.size());
Assert.assertEquals(splitter.splitGraph.inDegreeOf(splitter.prefixV), 0);
Assert.assertTrue(splitter.splitGraph.inDegreeOf(splitter.suffixV) <= strings.size());
Assert.assertEquals(splitter.splitGraph.outDegreeOf(splitter.suffixV), 0);
for (final SeqVertex mid : splitter.newMiddles) {
Assert.assertNotNull(splitter.splitGraph.getEdge(splitter.prefixV, mid));
Assert.assertNotNull(splitter.splitGraph.getEdge(mid, splitter.suffixV));
}
}
/**
* Can we safely split up the vertices in toMerge?
*
* @param graph a graph
* @param bot a vertex whose incoming vertices we want to split
* @param toMerge the set of vertices we'd be splitting up
* @return true if we can safely split up toMerge
*/
private boolean safeToSplit(
final SeqGraph graph, final SeqVertex bot, final Collection<SeqVertex> toMerge) {
final Set<SeqVertex> outgoingOfBot = new HashSet<SeqVertex>(graph.outgoingVerticesOf(bot));
for (final SeqVertex m : toMerge) {
final Set<BaseEdge> outs = graph.outgoingEdgesOf(m);
if (m == bot || outs.size() != 1 || !graph.outgoingVerticesOf(m).contains(bot))
// m == bot => don't allow self cycles in the graph
return false;
if (outgoingOfBot.contains(m))
// forbid cycles from bottom -> mid
return false;
}
return true;
}
// recursively initialize nodeDests from start node
private void initNodeDests(SeqPhase phase, SeqNode n) throws Xcept {
if (nodeDests.get(n) != null) return;
if (stopAtFFTs && n.fft() != null) return;
LinkedHashSet<SeqNode> dests = new LinkedHashSet<SeqNode>();
nodeDests.put(n, dests);
Iterator<SeqEdge> es = graph.getEdgesFrom(n);
while (es.hasNext()) {
SeqEdge e = es.next();
SeqNode n1 = e.dest();
dests.add(n1);
if (phase != graph.getPhase(n1)) continue;
if (stopAtFFTs && n1.fft() != null) continue;
initNodeDests(phase, n1);
}
}
/**
* Would factoring out this suffix result in elimating the reference source vertex?
*
* @param graph the graph
* @param commonSuffix the common suffix of all toSplits
* @param toSplits the list of vertices we're are trying to split
* @return true if toSplit contains the reference source and this ref source has all and only the
* bases of commonSuffix
*/
private boolean wouldEliminateRefSource(
final SeqGraph graph, final SeqVertex commonSuffix, final Collection<SeqVertex> toSplits) {
for (final SeqVertex toSplit : toSplits) {
if (graph.isRefSource(toSplit)) return toSplit.length() == commonSuffix.length();
}
return false;
}
// constructor
public SeqGraphXcons(
SeqGraph graph,
LinkedHashSet<SeqNode> startNodes,
LinkedHashSet<SeqNode> stopNodes,
boolean stopAtFFTs)
throws Xcept {
this.graph = graph;
this.startNodes = startNodes;
this.stopNodes = stopNodes;
this.stopAtFFTs = stopAtFFTs;
System.err.println("SeqGraphXCons stopAtFFTs=" + stopAtFFTs);
nodeDests = new Hashtable<SeqNode, LinkedHashSet<SeqNode>>();
Iterator<SeqNode> ns = startNodes.iterator();
while (ns.hasNext()) {
SeqNode n = ns.next();
SeqPhase p = graph.getPhase(n);
initNodeDests(p, n);
}
System.err.println(" nodeDests=" + nodeDests);
}
/**
* Simple single-function interface to split and then update a graph
*
* @param graph the graph containing the vertices in toMerge
* @param v The bottom node whose incoming vertices we'd like to split
* @return true if some useful splitting was done, false otherwise
*/
public boolean split(final SeqGraph graph, final SeqVertex v) {
if (graph == null) throw new IllegalArgumentException("graph cannot be null");
if (v == null) throw new IllegalArgumentException("v cannot be null");
if (!graph.vertexSet().contains(v))
throw new IllegalArgumentException("graph doesn't contain vertex v " + v);
final Collection<SeqVertex> toSplit = graph.incomingVerticesOf(v);
if (toSplit.size() < 2)
// Can only split at least 2 vertices
return false;
else if (!safeToSplit(graph, v, toSplit)) {
return false;
} else {
final SeqVertex suffixVTemplate = commonSuffix(toSplit);
if (suffixVTemplate.isEmpty()) {
return false;
} else if (wouldEliminateRefSource(graph, suffixVTemplate, toSplit)) {
return false;
} else if (allVerticesAreTheCommonSuffix(suffixVTemplate, toSplit)) {
return false;
} else {
final List<BaseEdge> edgesToRemove = new LinkedList<BaseEdge>();
// graph.printGraph(new File("split.pre_" + v.getSequenceString() + "." +
// counter + ".dot"), 0);
for (final SeqVertex mid : toSplit) {
// create my own copy of the suffix
final SeqVertex suffixV = new SeqVertex(suffixVTemplate.getSequence());
graph.addVertex(suffixV);
final SeqVertex prefixV = mid.withoutSuffix(suffixV.getSequence());
final BaseEdge out = graph.outgoingEdgeOf(mid);
final SeqVertex incomingTarget;
if (prefixV == null) {
// this node is entirely explained by suffix
incomingTarget = suffixV;
} else {
incomingTarget = prefixV;
graph.addVertex(prefixV);
graph.addEdge(prefixV, suffixV, new BaseEdge(out.isRef(), 1));
edgesToRemove.add(out);
}
graph.addEdge(suffixV, graph.getEdgeTarget(out), out.copy());
for (final BaseEdge in : graph.incomingEdgesOf(mid)) {
graph.addEdge(graph.getEdgeSource(in), incomingTarget, in.copy());
edgesToRemove.add(in);
}
}
graph.removeAllVertices(toSplit);
graph.removeAllEdges(edgesToRemove);
// graph.printGraph(new File("split.post_" + v.getSequenceString() + "." +
// counter++ + ".dot"), 0);
return true;
}
}
}
@Test(dataProvider = "CompleteCycleData")
public void testSplitterCompleteCycle(
final List<String> strings, final boolean hasTop, final boolean hasBot) {
final SeqGraph graph = new SeqGraph(11);
int edgeWeight = 1;
final SeqVertex top = hasTop ? new SeqVertex("AAAAAAAA") : null;
final SeqVertex bot = hasBot ? new SeqVertex("GGGGGGGG") : null;
final List<SeqVertex> v = new ArrayList<>();
for (final String s : strings) {
v.add(new SeqVertex(s));
}
graph.addVertices(v.toArray(new SeqVertex[v.size()]));
final SeqVertex first = v.get(0);
if (hasTop) {
graph.addVertex(top);
for (final SeqVertex vi : v) graph.addEdge(top, vi, new BaseEdge(vi == first, edgeWeight++));
}
if (hasBot) {
graph.addVertex(bot);
for (final SeqVertex vi : v) graph.addEdge(vi, bot, new BaseEdge(vi == first, edgeWeight++));
}
final Set<String> haplotypes = new HashSet<>();
final KBestHaplotypeFinder originalPaths =
new KBestHaplotypeFinder((SeqGraph) graph.clone(), graph.getSources(), graph.getSinks());
for (final KBestHaplotype path : originalPaths) haplotypes.add(new String(path.bases()));
final SharedVertexSequenceSplitter splitter = new SharedVertexSequenceSplitter(graph, v);
splitter.split();
if (PRINT_GRAPHS)
graph.printGraph(
new File(Utils.join("_", strings) + "_" + hasTop + "_" + hasBot + ".original.dot"), 0);
if (PRINT_GRAPHS)
splitter.splitGraph.printGraph(
new File(Utils.join("_", strings) + "_" + hasTop + "_" + hasBot + ".split.dot"), 0);
splitter.updateGraph(top, bot);
if (PRINT_GRAPHS)
graph.printGraph(
new File(Utils.join("_", strings) + "_" + hasTop + "_" + hasBot + ".updated.dot"), 0);
final KBestHaplotypeFinder splitPaths =
new KBestHaplotypeFinder(graph, graph.getSources(), graph.getSinks());
for (final KBestHaplotype path : splitPaths) {
final String h = new String(path.bases());
Assert.assertTrue(haplotypes.contains(h), "Failed to find haplotype " + h);
}
final List<byte[]> sortedOriginalPaths = new ArrayList<>(originalPaths.size());
for (final KBestHaplotype kbh : originalPaths.unique()) sortedOriginalPaths.add(kbh.bases());
Collections.sort(sortedOriginalPaths, BaseUtils.BASES_COMPARATOR);
final List<byte[]> sortedSplitPaths = new ArrayList<>(splitPaths.size());
for (final KBestHaplotype kbh : splitPaths.unique()) sortedSplitPaths.add(kbh.bases());
Collections.sort(sortedSplitPaths, BaseUtils.BASES_COMPARATOR);
Assert.assertEquals(
sortedSplitPaths,
sortedOriginalPaths,
Utils.join("_", strings) + "_" + hasTop + "_" + hasBot);
}