static bool canMerge(const ConcurrentDSU &uf, unsigned x, unsigned y) {
size_t szx = uf.set_size(x), szy = uf.set_size(y);
const size_t hardthr = 2500;
// Global threshold - no cluster larger than hard threshold
if (szx + szy > hardthr)
return false;
// If one of the clusters is moderately large, than attach "almost" singletons
// only.
if ((szx > hardthr * 3 / 4 && szy > 50) ||
(szy > hardthr * 3 / 4 && szx > 50))
return false;
return true;
}
static void processBlockQuadratic(ConcurrentDSU &uf,
const std::vector<size_t> &block,
const KMerData &data,
unsigned tau) {
size_t blockSize = block.size();
for (size_t i = 0; i < blockSize; ++i) {
unsigned x = (unsigned)block[i];
hammer::KMer kmerx = data.kmer(x);
for (size_t j = i + 1; j < blockSize; j++) {
unsigned y = (unsigned)block[j];
hammer::KMer kmery = data.kmer(y);
if (uf.find_set(x) != uf.find_set(y) &&
canMerge(uf, x, y) &&
hamdistKMer(kmerx, kmery, tau) <= tau) {
uf.unite(x, y);
}
}
}
}
Clusterization CostructClusters() {
InitializeReadSequenceMap();
#pragma omp parallel for
for(size_t i = 0; i < reads_.size(); i++) {
auto candidates = read_index_.GetCandidatesFor(i);
for(auto it = candidates.begin(); it != candidates.end(); it++) {
size_t cluster1 = clusters_.find_set(i);
size_t cluster2 = clusters_.find_set(*it);
if(cluster1 != cluster2) {
string read_seq1 = read_seq_map_[cluster1];
string read_seq2 = read_seq_map_[cluster2];
auto comparison_result = seq_comparer_.SequencesMatch(read_seq1, read_seq2);
if(comparison_result.match) {
clusters_.unite(i, *it);
string superstring = GetSuperString(read_seq1, read_seq2, comparison_result);
size_t new_cluster = clusters_.find_set(i);
read_seq_map_[new_cluster] = superstring;
}
}
}
}
cout << clusters_.num_sets() << " clusters were constructed" << endl;
Clusterization result(reads_);
for(size_t i = 0; i < reads_.size(); i++)
result.Add(i, clusters_.find_set(i));
for(auto it = read_seq_map_.begin(); it != read_seq_map_.end(); it++) {
size_t cluster_id = clusters_.find_set(it->first);
if(it->first == cluster_id) {
result.AddClusterSequence(cluster_id, it->second, clusters_.set_size(cluster_id));
}
}
assert(result.ClustersSize() == clusters_.num_sets());
return result;
}
static bool canMerge(const ConcurrentDSU &uf, unsigned x, unsigned y) {
return (uf.set_size(x) + uf.set_size(y)) < 10000;
}