suffix_mng_t *suffix_mng_new(sa_genome3_t *genome) {
suffix_mng_t *p = (suffix_mng_t *) calloc(1, sizeof(suffix_mng_t));
int num_chroms = genome->num_chroms;
char *name;
Container *subject = (Container *) malloc(sizeof(Container));
bl_containerInit(subject, num_chroms, sizeof(char *));
linked_list_t **suffix_lists = (linked_list_t **) malloc (sizeof(linked_list_t *) * num_chroms);
for (int i = 0; i < num_chroms; i++) {
suffix_lists[i] = linked_list_new(COLLECTION_MODE_ASYNCHRONIZED);
name = calloc(64, sizeof(char));
sprintf(name, "%i", i);
bl_containerAdd(subject, &name);
}
p->num_seeds = 0;
p->num_chroms = num_chroms;
p->subject = subject;
p->suffix_lists = suffix_lists;
return p;
}
/*----------------------------- bl_slExtractPoints -----------------------------
*
* @brief extracts all start and end points from the matches,
* returns a sorted list according to the position in the sequence,
* runs in O(n log n) (consists of one sort and two linear scans),
* @parameter array of slmatch_t, assumed to be presorted by
* start position on sequence
* @author Christian Otto
*
*/
Container* bl_slExtractPoints(slmatch_t *fragments, Uint size) {
int i, j, *space = NULL;
Uint *sorted;
Container *points;
point_t point;
slmatch_t *amatch, *bmatch;
/* sort the indexes of the container src by sequence position of end points */
sorted = quickSort(space, fragments, size, cmp_slmatch_end_quick, NULL);
/* initialize data structures */
points = (Container *) malloc(sizeof(Container));
bl_containerInit(points, 1000, sizeof(point_t));
/* traverse all start points according to the order */
for (i = 0, j = 0; i < size; i++) {
amatch = fragments + i;
bmatch = fragments + sorted[j];
/* if end point is before sequence pos of next start point */
while (FSTART_S(amatch) > FEND_S(bmatch)) {
point.x = FEND_S(bmatch);
point.y = FEND_Q(bmatch);
point.index = sorted[j];
point.start = 0;
bl_containerAdd(points, &point);
bmatch = fragments + sorted[++j];
}
point.x = FSTART_S(amatch);
point.y = FSTART_Q(amatch);
point.index = i;
point.start = 1;
bl_containerAdd(points, &point);
}
/* assumes that end point is always later than start point */
while (j < size) {
bmatch = fragments + sorted[j];
point.x = FEND_S(bmatch);
point.y = FEND_Q(bmatch);
point.index = sorted[j];
point.start = 0;
bl_containerAdd(points, &point);
j++;
}
free(sorted);
return points;
}
void FragmentList::addFragMatchRC(uint32_t refStart, uint32_t refEnd,
uint32_t queryStart, uint32_t queryEnd) {
slmatch_t frag;
bl_slmatchInit(&frag, 0);
frag.p = refStart;
frag.q = refEnd;
frag.i = queryStart;
frag.j = queryEnd;
frag.scr = static_cast<double>(queryEnd) - queryStart;
bl_containerAdd(fragmentsRC, &frag);
}
/*------------------------------ bl_slChainLin ---------------------------------
*
* @brief creates chains of non overlapping fragments from a list of
* fragments with the highest score using linear gap costs (L1)
* @parameter fragments - the list of fragments to be chained,
* lambda - cost of aligning gap with character on sequence,
* eps - cost of aligning gap with character on query
* @author Christian Otto
*
*/
void bl_slChainLin(slmatch_t *fragments, Uint size, double eps, double lambda, int maxgap) {
Uint i, *trans, *sorted, *space = NULL;
int pred, succ;
PairSint t;
point_t *point;
slmatch_t *current, *first;
slchain_t *chain, *cand, **prev, **predmatch, **succmatch;
#ifndef BINTREE
VebTree veb;
#else
BinTree bin;
#endif
Container *points;
VQueue refs;
/* no chaining if size is zero */
if (size == 0) {
return;
}
/* sorting (if already sorted -> only one scan required) */
qsort(fragments, size, sizeof(slmatch_t), cmp_slmatch_qsort);
/* preinitialize */
points = bl_slExtractPoints(fragments, size);
/* sorting by query position (required for vebtree/bintree) */
trans = (Uint *) malloc(sizeof(Uint) * bl_containerSize(points));
sorted = quickSort(space, points->contspace, bl_containerSize(points),
cmp_slmatch_trans_first_y, NULL);
for (i = 0; i < bl_containerSize(points); i++) {
trans[sorted[i]] = i;
}
/* get t.a and t.b */
t.a = ((point_t*) bl_containerGet(points, bl_containerSize(points) - 1))->x;
t.b = ((point_t*) bl_containerGet(points,
sorted[bl_containerSize(points) - 1]))->y;
free(sorted);
/* initializations for chaining itself */
prev = (slchain_t **) malloc(sizeof(slchain_t *) * size);
memset(prev, 0, sizeof(slchain_t*) * size);
#ifndef BINTREE
bl_vebtreeInit(&veb, bl_containerSize(points), sizeof(slchain_t *));
#else
bl_bintreeInit(&bin, bl_containerSize(points), sizeof(slchain_t *));
#endif
bl_vqueueInit(&refs, 2 * size, sizeof(slchain_t *));
/* traverse all points (sorted by sequence start position) */
for (i = 0; i < bl_containerSize(points); i++) {
point = (point_t *) bl_containerGet(points, i);
/* point is the start point of fragment point->index (in fragments) */
if (point->start == 1) {
/* get current match from fragments */
current = fragments + point->index;
/* RMQ in vebtree/bintree by y-coordinate */
#ifndef BINTREE
pred = bl_vebtreePred(&veb, trans[i]);
predmatch = (slchain_t **) bl_vebtreeGetData(&veb, pred);
#else
pred = bl_bintreePred(&bin, trans[i]);
predmatch = (slchain_t **) bl_bintreeGet(&bin, pred);
#endif
if (predmatch != NULL && !MAXGAP(current, *predmatch)) {
/* only take prev if it increases chain score */
if ((*predmatch)->scr >= (double) GLIN(current, *predmatch)) {
prev[point->index] = *predmatch;
}
/*
* addition due to clusters of local chains
* (differs from Abouelhoda2004)
*/
else if (current->scr >= (double) GLIN(current, *predmatch)) {
first = *(slmatch_t **) bl_containerGet((*predmatch)->matches, 0);
if (first->chain != NULL &&
((slchain_t *)first->chain)->scr < (*predmatch)->scr +
current->scr - GLIN(current, *predmatch)) {
chain = (slchain_t *) malloc(sizeof(slchain_t));
bl_slchainInit(chain);
bl_slchainCopy(chain, *predmatch);
chain->j = FEND_Q(current) - FSTART_Q(*predmatch) + 1;
chain->q = FEND_S(current) - FSTART_S(*predmatch) + 1;
chain->scr += current->scr - (double) GLIN(current, *predmatch);
bl_containerAdd(chain->matches, ¤t);
bl_slchainpDestruct(&first->chain);
first->chain = chain;
}
}
}
}
/* point is the end point of fragment point->index */
else {
current = fragments + point->index;
chain = (slchain_t *) malloc(sizeof(slchain_t));
/*
* chaining candidate found
* (can be only better chain for prev or best chain for current as well)
*/
if (prev[point->index] != NULL) {
cand = prev[point->index];
bl_slchainInit(chain);
chain->i = FSTART_Q(cand);
chain->j = FEND_Q(current) - FSTART_Q(cand) + 1;
chain->p = FSTART_S(cand);
chain->q = FEND_S(current) - FSTART_S(cand) + 1;
chain->scr = current->scr + cand->scr - (double) GLIN(current, cand);
bl_containerMerge(chain->matches, cand->matches);
bl_containerAdd(chain->matches, ¤t);
if (current->subject != cand->subject) {
DBG("slchain.c: Attempt to chain fragments of different \
reference sequences.\nExit forced.\n\n", NULL);
exit(-1);
}
chain->subject = current->subject;
/* update best chain of cluster */
first = *(slmatch_t **) bl_containerGet(chain->matches, 0);
if (first->chain != NULL) {
if (((slchain_t *) first->chain)->scr <= chain->scr) {
slchain_t *tmp = (slchain_t *) malloc(sizeof(slchain_t));
bl_slchainInit(tmp);
bl_slchainCopy(tmp, chain);
bl_slchainpDestruct(&first->chain);
first->chain = (void *) tmp;
}
}
}
/* no chaining candidate found */
else {
void suffix_mng_create_cals(fastq_read_t *read, int min_area, int strand,
sa_index3_t *sa_index, array_list_t *cal_list,
suffix_mng_t *p) {
if (!p) return;
if (!p->suffix_lists) return;
if (p->num_seeds <= 0) return;
int read_area, chrom;
seed_t *seed;
seed_cal_t *cal;
linked_list_t *seed_list;
claspinfo_t info;
bl_claspinfoInit(&info);
// initialization
info.fragments = (Container *) malloc(sizeof(Container));
bl_containerInit(info.fragments, p->num_seeds, sizeof(slmatch_t));
info.subject = p->subject;
slmatch_t frag;
linked_list_t *suffix_list;
for (unsigned int i = 0; i < p->num_chroms; i++) {
suffix_list = p->suffix_lists[i];
if (suffix_list) {
for (linked_list_item_t *item = suffix_list->first;
item != NULL;
item = item->next) {
seed = item->item;
bl_slmatchInit(&frag, 0);
frag.i = seed->read_start;
frag.j = seed->read_end - seed->read_start + 1;
frag.p = seed->genome_start;
frag.q = seed->genome_end - seed->genome_start + 1;
frag.scr = seed->genome_end - seed->genome_start + 1;
frag.subject = seed->chromosome_id;
bl_containerAdd(info.fragments, &frag);
}
}
}
// sort fragments
qsort(info.fragments->contspace, bl_containerSize(info.fragments),
sizeof(slmatch_t), cmp_slmatch_qsort);
int begin = 0;
for (int i = 1; i <= bl_containerSize(info.fragments); i++){
// end of fragments list or different database sequence
// --> process fragment[begin]...fragment[i-1], write output
// and free chains (less memory consumption with large input files)
if (i == bl_containerSize(info.fragments) ||
((slmatch_t *) bl_containerGet(info.fragments, begin))->subject !=
((slmatch_t *) bl_containerGet(info.fragments, i))->subject){
if (info.chainmode == SOP){
// only use chaining without clustering if no ids are specified
bl_slClusterSop((slmatch_t *) info.fragments->contspace + begin, i - begin,
info.epsilon, info.lambda, info.maxgap);
}
else {
bl_slClusterLin((slmatch_t *) info.fragments->contspace + begin, i - begin,
info.epsilon, info.lambda, info.maxgap);
}
for (int j = begin; j < i; j++) {
slmatch_t *match = (slmatch_t *) bl_containerGet(info.fragments, j);
if (match->chain) {
slchain_t *chain = (slchain_t *) match->chain;
if (chain->scr >= info.minscore &&
bl_containerSize(chain->matches) >= info.minfrag) {
chrom = atoi(*(char **) bl_containerGet(info.subject, chain->subject));
read_area = 0;
seed_list = linked_list_new(COLLECTION_MODE_ASYNCHRONIZED);
for (int k = 0; k < bl_containerSize(chain->matches); k++){
slmatch_t *frag = *(slmatch_t **) bl_containerGet(chain->matches, k);
seed = seed_new(frag->i, frag->i + frag->j - 1, frag->p, frag->p + frag->q - 1);
seed->chromosome_id = chrom;
seed->strand = strand;
read_area += frag->j;
cigar_append_op(frag->j, '=', &seed->cigar);
linked_list_insert_last(seed, seed_list);
}
// extend seeds
cal = seed_cal_new(chrom, strand, chain->p, chain->p + chain->q - 1, seed_list);
cal->read = read;
extend_seeds(cal, sa_index);
seed_cal_update_info(cal);
if (cal->read_area >= min_area) {
array_list_insert(cal, cal_list);
} else {
seed_cal_free(cal);
}
}
bl_slchainDestruct(chain);
free(chain);
match->chain = NULL;
}
} // END OF for (j = begin; j < i; j++)
begin = i;
} // END OF if (i == bl_containerSize(info.fragments) ||
} // END OF for (i = 1; i <= bl_containerSize(info.fragments); i++)
// destruct everything
info.subject = NULL;
bl_claspinfoDestruct(&info);
// finally, clear suffix manager
suffix_mng_clear(p);
}
