stList *chooseMatching_greedy(stList *edges, int64_t nodeNumber) {
/*
* Greedily picks the edge from the list such that each node has at most one edge.
*/
//First clone the list..
edges = stList_copy(edges, NULL);
stSortedSet *seen = getEmptyNodeOrEdgeSetWithCleanup();
stList *matching = stList_construct();
//Sort the adjacency pairs..
stList_sort(edges, chooseMatching_greedyP);
double strength = INT64_MAX;
while (stList_length(edges) > 0) {
stIntTuple *edge = stList_pop(edges);
double d = stIntTuple_get(edge, 2);
assert(d <= strength);
strength = d;
if(!nodeInSet(seen, stIntTuple_get(edge, 0)) && !nodeInSet(seen, stIntTuple_get(edge, 1))) {
addNodeToSet(seen, stIntTuple_get(edge, 0));
addNodeToSet(seen, stIntTuple_get(edge, 1));
stList_append(matching,edge);
}
}
assert(stList_length(edges) == 0);
stList_destruct(edges);
stSortedSet_destruct(seen);
return matching;
}
static stList *mergeSubstrings(stList *substrings, int64_t proximityToMerge) {
/*
* Merge set of substrings into fewer substrings, if they overlap by less than proximityToMerge
*/
stList *mergedSubstrings = stList_construct3(0, (void (*)(void *)) substring_destruct);
if (stList_length(substrings) == 0) {
return mergedSubstrings;
}
stList_sort(substrings, (int (*)(const void *, const void *)) substring_cmp);
Substring *pSubsequence = substring_clone(stList_get(substrings, 0));
stList_append(mergedSubstrings, pSubsequence);
for (int64_t i = 1; i < stList_length(substrings); i++) {
Substring *substring = stList_get(substrings, i);
if (pSubsequence->name == substring->name
&& pSubsequence->start + pSubsequence->length + proximityToMerge >= substring->start) { //Merge
if (pSubsequence->start + pSubsequence->length < substring->start + substring->length) {
pSubsequence->length = substring->start + substring->length - pSubsequence->start;
}
} else {
pSubsequence = substring_clone(substring);
stList_append(mergedSubstrings, pSubsequence);
}
}
return mergedSubstrings;
}
void stCaf_addAdjacencies(Flower *flower) {
//Build a list of caps.
stList *list = stList_construct();
Flower_EndIterator *endIterator = flower_getEndIterator(flower);
End *end;
while ((end = flower_getNextEnd(endIterator)) != NULL) {
End_InstanceIterator *instanceIterator = end_getInstanceIterator(end);
Cap *cap;
while ((cap = end_getNext(instanceIterator)) != NULL) {
if (!cap_getStrand(cap)) {
cap = cap_getReverse(cap);
}
stList_append(list, cap);
}
end_destructInstanceIterator(instanceIterator);
}
flower_destructEndIterator(endIterator);
assert(stList_length(list) % 2 == 0);
//Sort the list of caps.
stList_sort(list, (int(*)(const void *, const void *)) addAdjacenciesPP);
//Now make the adjacencies.
for (int64_t i = 1; i < stList_length(list); i += 2) {
Cap *cap = stList_get(list, i - 1);
Cap *cap2 = stList_get(list, i);
cap_makeAdjacent(cap, cap2);
}
//Clean up.
stList_destruct(list);
}
void stTree_sortChildren(stTree *root, int cmpFn(stTree *a, stTree *b)) {
sortChildrenCmpFn = cmpFn;
stList_sort(root->nodes, sortChildrenListCmpFn);
sortChildrenCmpFn = NULL;
for (int i = 0; i < stTree_getChildNumber(root); i++) {
stTree_sortChildren(stTree_getChild(root, i), cmpFn);
}
}
static AdjacencySwitch *getBest2EdgeAdjacencySwitch(stList *components,
stSortedSet *allAdjacencyEdges) {
/*
* Look for the two lowest value adjacency edges in all current edges that are in a separate component and returns them as an adjacency switch
* with now new adjacency edges.
*/
/*
* Get lowest scoring edge for each component.
*/
stList *lowestScoringEdgeFromEachComponent = stList_construct();
for (int64_t i = 0; i < stList_length(components); i++) {
stList_append(lowestScoringEdgeFromEachComponent,
getLowestScoringEdge(stList_get(components, i)));
}
/*
* Get two lowest scoring edges.
*/
stList_sort(lowestScoringEdgeFromEachComponent,
getBest2EdgeAdjacencySwitchP);
stIntTuple *lowestScoreEdge1 = stList_get(
lowestScoringEdgeFromEachComponent, 0);
stIntTuple *lowestScoreEdge2 = stList_get(
lowestScoringEdgeFromEachComponent, 1);
assert(lowestScoreEdge1 != lowestScoreEdge2);
stList_destruct(lowestScoringEdgeFromEachComponent); //Cleanup
stIntTuple *newEdge1 = getWeightedEdgeFromSet(
stIntTuple_get(lowestScoreEdge1, 0),
stIntTuple_get(lowestScoreEdge2, 0), allAdjacencyEdges);
stIntTuple *newEdge2 = getWeightedEdgeFromSet(
stIntTuple_get(lowestScoreEdge1, 1),
stIntTuple_get(lowestScoreEdge2, 1), allAdjacencyEdges);
if (newEdge1 == NULL) {
assert(newEdge2 == NULL);
newEdge1 = getWeightedEdgeFromSet(
stIntTuple_get(lowestScoreEdge1, 0),
stIntTuple_get(lowestScoreEdge2, 1), allAdjacencyEdges);
newEdge2 = getWeightedEdgeFromSet(
stIntTuple_get(lowestScoreEdge1, 1),
stIntTuple_get(lowestScoreEdge2, 0), allAdjacencyEdges);
}
assert(newEdge1 != NULL);
assert(newEdge2 != NULL);
return adjacencySwitch_construct(
lowestScoreEdge1,
lowestScoreEdge2,
newEdge1,
newEdge2,
stIntTuple_get(lowestScoreEdge1, 2)
+ stIntTuple_get(lowestScoreEdge2, 2));
}
void test_stList_sort(CuTest *testCase) {
setup();
stList_sort(list, (int (*)(const void *, const void *))strcmp);
CuAssertTrue(testCase, stList_length(list) == stringNumber);
CuAssertStrEquals(testCase, "five", stList_get(list, 0));
CuAssertStrEquals(testCase, "four", stList_get(list, 1));
CuAssertStrEquals(testCase, "one", stList_get(list, 2));
CuAssertStrEquals(testCase, "three", stList_get(list, 3));
CuAssertStrEquals(testCase, "two", stList_get(list, 4));
teardown();
}
/* build a list of blocks, sorted by the root components */
static stList *buildRootSorted(struct malnSet *malnSet) {
stList *sorted = stList_construct();
struct malnBlkSetIterator *iter = malnBlkSet_getIterator(malnSet->blks);
struct malnBlk *blk;
while ((blk = malnBlkSetIterator_getNext(iter)) != NULL) {
stList_append(sorted, blk);
}
malnBlkSetIterator_destruct(iter);
stList_sort(sorted, blkCmpRootComp);
return sorted;
}
/* Get a list of components that overlap the specified guide range and are in
* blocks not flagged as dying and matches treeLoc filters. Return NULL if no
* overlaps. List is sorted by ascending width, which helps the merge
* efficiency. */
stList *malnSet_getOverlappingComps(struct malnSet *malnSet, struct Seq *seq, int chromStart, int chromEnd, unsigned treeLocFilter) {
if (malnSet->compRangeMap == NULL) {
buildRangeTree(malnSet);
}
stList *overComps = NULL;
for (struct range *rng = genomeRangeTreeAllOverlapping(malnSet->compRangeMap, seq->orgSeqName, chromStart, chromEnd); rng != NULL; rng = rng->next) {
for (struct slRef *compRef = rng->val; compRef != NULL; compRef = compRef->next) {
struct malnComp *comp = compRef->val;
if (keepOverlap(comp, seq, chromStart, chromEnd, treeLocFilter)) {
if (overComps == NULL) {
overComps = stList_construct();
}
stList_append(overComps, comp);
}
}
}
// sort so tests are reproducible
if (overComps != NULL) {
stList_sort(overComps, sortCompListCmpFn);
}
return overComps;
}
int main(int argc, char *argv[]) {
// Parse arguments
if (argc != 3) {
usage(argv);
return 1;
}
// You would load a custom HMM here if you wanted using
// hmm_getStateMachine (see the realign code)
StateMachine *stateMachine = stateMachine5_construct(fiveState);
PairwiseAlignmentParameters *parameters = pairwiseAlignmentBandingParameters_construct();
stHash *targetSequences = readFastaFile(argv[1]);
stHash *querySequences = readFastaFile(argv[2]);
// For each query sequence, align it against all target sequences.
stHashIterator *queryIt = stHash_getIterator(querySequences);
char *queryHeader;
while ((queryHeader = stHash_getNext(queryIt)) != NULL) {
char *querySeq = stHash_search(querySequences, queryHeader);
stHashIterator *targetIt = stHash_getIterator(targetSequences);
char *targetHeader;
while ((targetHeader = stHash_getNext(targetIt)) != NULL) {
char *targetSeq = stHash_search(targetSequences, targetHeader);
// Here we should try both the target sequence and its
// reverse-complemented version
// Aligns the sequences.
// If you have alignment constraints (anchors) you should
// replace this with getAlignedPairsUsingAnchors.
stList *alignedPairs = getAlignedPairs(stateMachine, targetSeq,
querySeq, parameters,
true, true);
// Takes into account the probability of aligning to a
// gap, by transforming the posterior probability into the
// AMAP objective function (see Schwartz & Pachter, 2007).
alignedPairs = reweightAlignedPairs2(alignedPairs, strlen(targetSeq),
strlen(querySeq),
parameters->gapGamma);
// I think this calculates the optimal ordered set of
// alignments from the unordered set of aligned pairs, not
// completely sure.
alignedPairs = filterPairwiseAlignmentToMakePairsOrdered(alignedPairs,
targetSeq,
querySeq,
// This parameter says that the minimum posterior probability we will accept has to be at least 0.9.
0.9);
// After this the "aligned pairs" data structure changes,
// which is a little sketchy. It's just so that the
// alignment can be printed properly.
stList_mapReplace(alignedPairs, convertToAnchorPair, NULL);
stList_sort(alignedPairs, (int (*)(const void *, const void *)) stIntTuple_cmpFn);
struct PairwiseAlignment *alignment = convertAlignedPairsToPairwiseAlignment(targetHeader, queryHeader,
0, strlen(targetSeq), strlen(querySeq), alignedPairs);
// Output the cigar string
cigarWrite(stdout, alignment, 0);
stList_destruct(alignedPairs);
destructPairwiseAlignment(alignment);
}
stHash_destructIterator(targetIt);
}
stHash_destructIterator(queryIt);
// Clean up
stHash_destruct(targetSequences);
stHash_destruct(querySequences);
pairwiseAlignmentBandingParameters_destruct(parameters);
stateMachine_destruct(stateMachine);
}
