static inline
KOccurRun* fetch_ref_contact(const GraphCache *cache, uint32_t pathid,
const PathRefRun *ref_runs,
KOccurRunBuffer *runbuf)
{
// Get path
const GCachePath *path = graph_cache_path(cache, pathid);
const GCacheStep *steps = graph_cache_step(cache, path->first_step);
size_t num_steps = path->num_steps;
// Get runs along the ref
PathRefRun ref_run = ref_runs[pathid];
size_t num_runs = ref_run.num_runs;
KOccurRun *koruns = runbuf->data + ref_run.first_runid;
koruns_sort_by_qoffset(koruns, num_runs);
// Set qoffset to be the kmer offset in the path
size_t r, s, offset = 0;
for(s = r = 0; s < num_steps; s++) {
for(; r < num_runs && koruns[r].qoffset == s; r++) {
koruns[r].qoffset = offset;
}
if(r == num_runs) break;
const GCacheSnode *snode = graph_cache_snode(cache, steps[s].supernode);
offset += snode->num_nodes;
}
return koruns;
}
// If `pickup_new_runs` is true we pick up runs starting at this supernode
static inline bool gcrawler_stop_at_ref_covg(const GraphCache *cache,
const GCacheStep *step,
BreakpointCaller *caller,
KOccurRunBuffer *koruns,
KOccurRunBuffer *koruns_ended,
bool pickup_new_runs)
{
GCacheSnode *snode = graph_cache_snode(cache, step->supernode);
GCachePath *path = graph_cache_path(cache, step->pathid);
const dBNode *nodes = graph_cache_first_node(cache, snode);
bool forward = (step->orient == FORWARD);
// Use index of last step as qoffset
size_t qoffset = path->num_steps-1;
// Kmer occurance runs are added to koruns_3p_ended only if they end and are
// longer than the mininum length in kmers (caller->min_ref_nkmers)
kograph_filter_extend(caller->kograph,
nodes, snode->num_nodes, forward,
caller->min_ref_nkmers, qoffset,
koruns, koruns_ended,
pickup_new_runs);
size_t i, min_run_qoffset = SIZE_MAX, min_ended_run_qoffset = SIZE_MAX;
for(i = 0; i < koruns->len; i++)
min_run_qoffset = MIN2(min_run_qoffset, koruns->data[i].qoffset);
// Stop if all our earliest runs have finished
for(i = 0; i < koruns_ended->len; i++) {
min_ended_run_qoffset = MIN2(min_ended_run_qoffset, koruns_ended->data[i].qoffset);
}
// Continue if...
return min_run_qoffset <= min_ended_run_qoffset;
}
// Constructs a path of supernodes (SupernodePath)
// `wlk` GraphWalker should be set to go at `node`
// `rptwlk` RepeatWalker should be clear
// `jmpfunc` is called with each supernode traversed and if it returns true
// we continue crawling, otherwise we stop. If NULL assume always true
// returns pathid in GraphCache
uint32_t graph_crawler_load_path(GraphCache *cache, dBNode node,
GraphWalker *wlk, RepeatWalker *rptwlk,
bool (*jmpfunc)(GraphCache *_cache,
GCacheStep *_step, void *_arg),
void *arg)
{
size_t i;
uint32_t stepid, pathid = graph_cache_new_path(cache);
ctx_assert(db_nodes_are_equal(wlk->node, node));
for(i = 0; ; i++)
{
stepid = graph_cache_new_step(cache, node);
GCacheStep *step = graph_cache_step(cache, stepid);
GCacheSnode *snode = graph_cache_snode(cache, step->supernode);
// Traverse to the end of the supernode
walk_supernode_end(cache, snode, step->orient, wlk);
if(jmpfunc != NULL && !jmpfunc(cache, step, arg)) break;
// Find next node
uint8_t num_edges;
const dBNode *next_nodes;
Nucleotide next_bases[4];
if(step->orient == FORWARD) {
num_edges = snode->num_next;
next_nodes = snode->next_nodes;
binary_seq_unpack_byte(next_bases, snode->next_bases);
}
else {
num_edges = snode->num_prev;
next_nodes = snode->prev_nodes;
binary_seq_unpack_byte(next_bases, snode->prev_bases);
}
// Traverse to next supernode
if(!graph_walker_next_nodes(wlk, num_edges, next_nodes, next_bases) ||
!rpt_walker_attempt_traverse(rptwlk, wlk)) break;
node = wlk->node;
}
return pathid;
}
static void write_bubbles_to_file(BubbleCaller *caller)
{
// Loop over supernodes checking if they are 3p flanks
size_t snode_count = graph_cache_num_snodes(&caller->cache);
GCacheSnode *snode;
size_t i;
for(i = 0; i < snode_count; i++)
{
snode = graph_cache_snode(&caller->cache, i);
find_bubbles_ending_with(caller, snode);
if(caller->spp_forward.len > 1)
print_bubble(caller, caller->spp_forward.b, caller->spp_forward.len);
if(caller->spp_reverse.len > 1)
print_bubble(caller, caller->spp_reverse.b, caller->spp_reverse.len);
}
}
void graph_crawler_reset_rpt_walker(RepeatWalker *rptwlk,
const GraphCache *cache, uint32_t pathid)
{
rpt_walker_fast_clear(rptwlk, NULL, 0);
const GCachePath *path = graph_cache_path(cache, pathid);
const GCacheStep *step = graph_cache_step(cache, path->first_step), *endstep;
const GCacheSnode *snode;
const dBNode *node0, *node1;
// Loop over supernodes in the path
for(endstep = step + path->num_steps; step < endstep; step++)
{
// We don't care about orientation here
snode = graph_cache_snode(cache, step->supernode);
node0 = graph_cache_first_node(cache, snode);
node1 = graph_cache_last_node(cache, snode);
rpt_walker_fast_clear_single_node(rptwlk, *node0);
rpt_walker_fast_clear_single_node(rptwlk, *node1);
}
}
// Potential bubble - filter ref and duplicate alleles
static void print_bubble(BubbleCaller *caller,
GCacheStep **steps, size_t num_paths)
{
const BubbleCallingPrefs prefs = caller->prefs;
const dBGraph *db_graph = caller->db_graph;
GCacheSnode *snode;
size_t i;
dBNodeBuffer *flank5p = &caller->flank5p;
if(flank5p->len == 0)
{
// Haven't fetched 5p flank yet
// flank5p[0] already contains the first node
flank5p->len = 1;
supernode_extend(flank5p, prefs.max_flank_len, db_graph);
db_nodes_reverse_complement(flank5p->b, flank5p->len);
}
//
// Print Bubble
//
// write to string buffer then flush to gzFile
StrBuf *sbuf = &caller->output_buf;
strbuf_reset(sbuf);
// Temporary node buffer to use
dBNodeBuffer *pathbuf = &caller->pathbuf;
db_node_buf_reset(pathbuf);
// Get bubble number (threadsafe num_bubbles_ptr++)
size_t id = __sync_fetch_and_add((volatile size_t*)caller->num_bubbles_ptr, 1);
// This can be set to anything without a '.' in it
const char prefix[] = "call";
// 5p flank
// strbuf_sprintf(sbuf, ">bubble.%s%zu.5pflank kmers=%zu\n", prefix, id, flank5p->len);
strbuf_append_str(sbuf, ">bubble.");
strbuf_append_str(sbuf, prefix);
strbuf_append_ulong(sbuf, id);
strbuf_append_str(sbuf, ".5pflank kmers=");
strbuf_append_ulong(sbuf, flank5p->len);
strbuf_append_char(sbuf, '\n');
branch_to_str(flank5p->b, flank5p->len, true, sbuf, db_graph);
// 3p flank
db_node_buf_reset(pathbuf);
snode = graph_cache_snode(&caller->cache, steps[0]->supernode);
graph_cache_snode_fetch_nodes(&caller->cache, snode, steps[0]->orient, pathbuf);
// strbuf_sprintf(sbuf, ">bubble.%s%zu.3pflank kmers=%zu\n", prefix, id, pathbuf->len);
strbuf_append_str(sbuf, ">bubble.");
strbuf_append_str(sbuf, prefix);
strbuf_append_ulong(sbuf, id);
strbuf_append_str(sbuf, ".3pflank kmers=");
strbuf_append_ulong(sbuf, pathbuf->len);
strbuf_append_char(sbuf, '\n');
branch_to_str(pathbuf->b, pathbuf->len, false, sbuf, db_graph);
// Print alleles
for(i = 0; i < num_paths; i++)
{
db_node_buf_reset(pathbuf);
graph_cache_step_fetch_nodes(&caller->cache, steps[i], pathbuf);
// strbuf_sprintf(sbuf, ">bubble.%s%zu.branch.%zu kmers=%zu\n",
// prefix, id, i, pathbuf->len);
strbuf_append_str(sbuf, ">bubble.");
strbuf_append_str(sbuf, prefix);
strbuf_append_ulong(sbuf, id);
strbuf_append_str(sbuf, ".branch.");
strbuf_append_ulong(sbuf, i);
strbuf_append_str(sbuf, " kmers=");
strbuf_append_ulong(sbuf, pathbuf->len);
strbuf_append_char(sbuf, '\n');
branch_to_str(pathbuf->b, pathbuf->len, false, sbuf, db_graph);
}
strbuf_append_char(sbuf, '\n');
ctx_assert(strlen(sbuf->b) == sbuf->end);
// lock, print, unlock
pthread_mutex_lock(caller->out_lock);
gzwrite(caller->gzout, sbuf->b, sbuf->end);
pthread_mutex_unlock(caller->out_lock);
}
