void Caller::call_node_pileup(const NodePileup& pileup) {
_node = _graph->get_node(pileup.node_id());
assert(_node != NULL);
_node_calls.clear();
char def_char = _leave_uncalled ? '.' : '-';
_node_calls.assign(_node->sequence().length(), Genotype(def_char, def_char));
// todo: parallelize this loop
// process each base in pileup individually
for (int i = 0; i < pileup.base_pileup_size(); ++i) {
if (pileup.base_pileup(i).num_bases() >= _min_depth &&
pileup.base_pileup(i).num_bases() <= _max_depth) {
call_base_pileup(pileup, i);
}
}
// add nodes and edges created when making calls to the output graph
// (_side_map gets updated)
create_node_calls(pileup);
_visited_nodes.insert(_node->id());
}
void Caller::call_base_pileup(const NodePileup& np, int64_t offset) {
const BasePileup& bp = np.base_pileup(offset);
// parse the pilueup structure
vector<pair<int, int> > base_offsets;
vector<pair<int, int> > indel_offsets;
Pileups::parse_base_offsets(bp, base_offsets, indel_offsets, indel_offsets);
// compute top two most frequent bases and their counts
char top_base;
int top_count;
char second_base;
int second_count;
compute_top_frequencies(bp, base_offsets, top_base, top_count, second_base, second_count);
// note first and second base will be upper case too
char ref_base = ::toupper(bp.ref_base());
// test against thresholding heuristics
if ((double)(top_count + second_count) / (double)base_offsets.size() >= _min_frac) {
// compute max likelihood snp genotype. it will be one of the three combinations
// of the top two bases (we don't care about case here)
pair<char, char> g = mp_snp_genotype(bp, base_offsets, top_base, second_base);
// update the node calls
if (top_count >= _min_support) {
if (g.first != ref_base) {
_node_calls[offset].first = g.first;
} else {
_node_calls[offset].first = '.';
}
}
if (second_count >= _min_support) {
if (g.second != ref_base && g.second != g.first) {
_node_calls[offset].second = g.second;
} else {
_node_calls[offset].second = '.';
}
}
}
}
