void GraphOutput::load_nodes_extremities(string linear_seqs_name)
{
kmer_links.clear(); // PIERRE: reset previous stored kmer links
Bank *Nodes = new Bank((char *)linear_seqs_name.c_str());
long nb_nodes = first_id_els.node; //PIERRE;
char * rseq;
int readlen;
sizeKmer--; // nodes extremities overlap by (k-1)-mers, so let's extract (k-1)-mers
while (Nodes->get_next_seq(&rseq,&readlen))
{
kmer_type left_kmer, right_kmer, left_kmer_fw, left_kmer_rc, right_kmer_fw, right_kmer_rc;
left_kmer = extractKmerFromRead(rseq,0,&left_kmer_fw,&left_kmer_rc, false);
right_kmer = extractKmerFromRead(rseq,readlen-sizeKmer,&right_kmer_fw,&right_kmer_rc, false);
Strand left_strand = (left_kmer == left_kmer_fw)?FW:RC;
Strand right_strand = (right_kmer == right_kmer_fw)?FW:RC;
kmer_links[left_kmer].insert(node_strand(nb_nodes, left_strand, LEFT));
kmer_links[right_kmer].insert(node_strand(nb_nodes, right_strand, RIGHT));
nb_nodes++;
}
Nodes->close();
delete Nodes;
sizeKmer++; // make sure to restore k
}
Set *load_false_positives()
{
int64_t NbInsertedKmers = 0;
char * rseq;
int readlen;
kmer_type kmer, graine, graine_revcomp;
Bank *FalsePositives = new Bank(return_file_name(false_positive_kmers_file));
// alloc false positives with the just the right estimated size
uint64_t nbFP = countFP(FalsePositives);
FPSet *fp = new FPSet(nbFP);
while (FalsePositives->get_next_seq(&rseq,&readlen))
{
kmer = extractKmerFromRead(rseq,0,&graine,&graine_revcomp);
fp->insert(kmer);
NbInsertedKmers++;
if ((NbInsertedKmers%table_print_frequency)==0) fprintf (stderr,(char*)"%cInsert false positive Kmers in hash table %lld",13,NbInsertedKmers);
}
fp->finalize(); // always call this when finishing to create a FPSet
fprintf (stderr,"\nInserted %lld false positive kmers in the hash structure.\n\n",NbInsertedKmers);
print_size_summary(fp);
return fp;
}
id_els GraphOutput::construct_graph(string linear_seqs_name) // PIERRE: i need to know the last nb_nodes
{
Bank *Nodes = new Bank((char *)linear_seqs_name.c_str());
id_els nb_els = first_id_els; //Alexan: stucture for print id elements in graph output
char * rseq;
int readlen;
Nodes->rewind_all();
sizeKmer--; // nodes extremities overlap by (k-1)-mers, so let's extract (k-1)-mers
// for each node, output all the out-edges (in-edges will correspond to out-edges of neighbors)
while (Nodes->get_next_seq(&rseq,&readlen))
{
kmer_type left_kmer, right_kmer, left_kmer_fw, left_kmer_rc, right_kmer_fw, right_kmer_rc;
set<node_strand>::iterator it;
left_kmer = extractKmerFromRead(rseq,0,&left_kmer_fw,&left_kmer_rc, false);
right_kmer = extractKmerFromRead(rseq,readlen-sizeKmer,&right_kmer_fw,&right_kmer_rc, false);
Strand left_strand = (left_kmer == left_kmer_fw)?FW:RC;
Strand right_strand = (right_kmer == right_kmer_fw)?FW:RC;
// left edges (are revcomp extensions)
for (it = kmer_links[left_kmer].begin(); it != kmer_links[left_kmer].end(); it++)
{
long cur_node = it->node;
Strand cur_strand = it->strand;
LeftOrRight cur_left_or_right = it->left_or_right;
if (cur_node ==nb_els.node) // prevent self loops on same kmer
if (readlen == sizeKmer)
continue;
string label = "R";
if (cur_left_or_right == LEFT)
{
if (cur_strand != left_strand)
label+=(string)"F";
else
continue;
}
else
{
if (cur_strand == left_strand)
label+=(string)"R";
else
continue;
}
print_edge(nb_els.edge, nb_els.node,cur_node,label);
nb_els.edge++;
}
// right edges
for (it = kmer_links[right_kmer].begin(); it != kmer_links[right_kmer].end(); it++)
{
long cur_node = it->node;
Strand cur_strand = it->strand;
LeftOrRight cur_left_or_right = it->left_or_right;
if (cur_node == nb_els.node) // prevent self loops on same kmer
if (readlen == sizeKmer)
continue;
string label = "F";
if (cur_left_or_right == LEFT)
{
if (cur_strand == right_strand)
label+=(string)"F";
else
continue;
}
else
{
if (cur_strand != right_strand)
label+=(string)"R";
else
continue;
}
print_edge(nb_els.edge, nb_els.node,cur_node,label);
nb_els.edge++;
}
//nodes
print_node(nb_els.node, rseq);
nb_els.node++;
}
sizeKmer++; // make sure to restore k
Nodes->close();
delete Nodes;
return nb_els;
}
Set *load_false_positives_cascading4()
{
int64_t NbInsertedKmers;
char * rseq;
int readlen;
kmer_type kmer, graine, graine_revcomp;
// **** Initialize B2, B3, B4 and T4 ****
Bank *FalsePositives = new Bank(return_file_name(false_positive_kmers_file));
uint64_t nbFP = countFP(FalsePositives);
FPSetCascading4 *fp = new FPSetCascading4;
fp->bloom2 = new Bloom((uint64_t)(nbFP * NBITS_PER_KMER));
fp->bloom2->set_number_of_hash_func((int)floorf(0.7*NBITS_PER_KMER));
uint64_t estimated_T2_size = max((int)ceilf(nbkmers_solid * (double)powf((double)0.62, (double)NBITS_PER_KMER)), 1);
uint64_t estimated_T3_size = max((int)ceilf(nbFP * (double)powf((double)0.62, (double)NBITS_PER_KMER)) ,1);
fp->bloom3 = new Bloom((uint64_t)(estimated_T2_size * NBITS_PER_KMER));
fp->bloom3->set_number_of_hash_func((int)floorf(0.7*NBITS_PER_KMER));
fp->bloom4 = new Bloom((uint64_t)(estimated_T3_size * NBITS_PER_KMER));
fp->bloom4->set_number_of_hash_func((int)floorf(0.7*NBITS_PER_KMER));
// **** Insert the false positives in B2 ****
NbInsertedKmers = 0;
while (FalsePositives->get_next_seq(&rseq,&readlen))
{
kmer = extractKmerFromRead(rseq,0,&graine,&graine_revcomp);
fp->bloom2->add(kmer);
NbInsertedKmers++;
if ((NbInsertedKmers%table_print_frequency)==0)
fprintf (stderr,"%cInsert false positive B2 %lld",13,NbInsertedKmers);
}
fprintf (stderr,"%cInsert false positive B2 %lld", 13,NbInsertedKmers);
FalsePositives->close();
printf("\nInserted %lld (estimated, %lld) kmers in B2.\n", NbInsertedKmers, nbFP);
// **** Insert false positives in B3 and write T2
int addKmers = 0;
NbInsertedKmers = 0;
FILE *T2_file = fopen(return_file_name("t2_kmers"), "w+"); // We will read this file later, when filling T4
BinaryBank *SolidKmers = new BinaryBank(return_file_name(solid_kmers_file),sizeof(kmer),0);
while(SolidKmers->read_element(&kmer))
{
if (fp->bloom2->contains(kmer))
{
if (!fwrite(&kmer, sizeof(kmer), 1, T2_file))
{
printf("error: can't fwrite (disk full?)\n");
exit(1);
}
fp->bloom3->add(kmer);
addKmers++;
}
NbInsertedKmers++;
if ((NbInsertedKmers% table_print_frequency)==0)
fprintf (stderr,(char*)"%cInsert false positive B3 %lld",13,NbInsertedKmers);
}
fprintf (stderr,(char*)"%cInsert false positive B3 %lld",13,NbInsertedKmers);
SolidKmers->close();
printf("\nInserted %lld (estimated, %llu) kmers in B3.\n", addKmers, estimated_T2_size);
// **** Insert false positives in B4 (we could write T3, but it's not necessary)
FalsePositives = new Bank(return_file_name(false_positive_kmers_file));
NbInsertedKmers = 0;
addKmers = 0;
while (FalsePositives->get_next_seq(&rseq,&readlen))
{
kmer = extractKmerFromRead(rseq,0,&graine,&graine_revcomp);
if (fp->bloom3->contains(kmer))
{
fp->bloom4->add(kmer);
addKmers++;
}
NbInsertedKmers++;
if ((NbInsertedKmers%table_print_frequency)==0)
fprintf (stderr,"%cInsert false positive B4 %lld",13,NbInsertedKmers);
}
fprintf (stderr,"%cInsert false positive B4 %lld", 13,NbInsertedKmers);
FalsePositives->close();
printf("\nInserted %lld (estimated, %lld) kmers in B4.\n", addKmers, estimated_T3_size);
// **** Count and insert false positives in T4
rewind(T2_file);
addKmers = 0;
while (fread(&kmer, sizeof(kmer), 1, T2_file))
if (fp->bloom4->contains(kmer))
addKmers++;
fp->false_positives = new FPSet(addKmers);
rewind(T2_file);
addKmers = 0;
NbInsertedKmers = 0;
while (fread(&kmer, sizeof(kmer), 1, T2_file))
{
if (fp->bloom4->contains(kmer))
{
fp->false_positives->insert(kmer);
addKmers++;
}
NbInsertedKmers++;
if ((NbInsertedKmers%table_print_frequency)==0)
fprintf (stderr,"%cInsert false positive T4 %lld",13,NbInsertedKmers);
}
fp->false_positives->finalize();
fprintf (stderr,"%cInsert false positive T4 %lld", 13,NbInsertedKmers);
fclose(T2_file);
printf("\nInserted %lld (estimated, %lld) kmers in T4.\n\n", addKmers, (uint64_t)fp->false_positives->capacity());
print_size_summary(fp);
return fp;
}