//
// First time through, malloc and fill the variant list from the config table
//
static vmiEnumParameterP getVariantList() {
static vmiEnumParameterP list = NULL;
if (!list) {
Uns32 v = 1 + countVariants();
list = STYPE_CALLOC_N(vmiEnumParameter, v);
vmiEnumParameterP prm;
ppc32ConfigCP cfg;
Uns32 i;
for (i = 0, cfg = ppc32ConfigTable, prm = list;
cfg->name;
i++, cfg++, prm++) {
prm->name = cfg->name;
prm->value = i;
}
}
return list;
}
void BlockQuantify::count()
{
_impl->fasta_to_use.reset(new FastaFile(_impl->ref_fasta));
#ifdef DEBUG_BLOCKQUANTIFY
int lastpos = 0;
std::cerr << "starting block." << "\n";
#endif
auto current_bs_start = _impl->variants.begin();
std::string current_chr;
int current_bs = -1;
bool current_bs_valid = false;
// function to compute the QQ values for truth variants in the current
// benchmarking superlocus
const auto update_bs_qq = [this, ¤t_bs_start](BlockQuantifyImpl::variantlist_t::iterator to)
{
std::vector<float> tp_qqs;
for(auto cur = current_bs_start; cur != to; ++cur)
{
const float qqq = bcfhelpers::getFormatFloat(_impl->hdr, *cur, "QQ", 1);
if(std::isnan(qqq))
{
continue;
}
const std::string bd = bcfhelpers::getFormatString(_impl->hdr, *cur, "BD", 1);
// we want the scores of all TPs in this BS
if(bd == "TP")
{
tp_qqs.push_back(qqq);
}
}
float t_qq = bcfhelpers::missing_float();
if(!tp_qqs.empty())
{
t_qq = *(std::max_element(tp_qqs.begin(), tp_qqs.end()));
}
/** compute the median over all variants */
int fsize = bcf_hdr_nsamples(_impl->hdr);
float * fmt = (float*)calloc((size_t) fsize, sizeof(float));
for(auto cur = current_bs_start; cur != to; ++cur)
{
const std::string bd = bcfhelpers::getFormatString(_impl->hdr, *cur, "BD", 0);
bcf_get_format_float(_impl->hdr, *cur, "QQ", &fmt, &fsize);
if(bd != "TP")
{
fmt[0] = bcfhelpers::missing_float();
}
else
{
fmt[0] = t_qq;
}
bcf_update_format_float(_impl->hdr, *cur, "QQ", fmt, fsize);
}
free(fmt);
#ifdef DEBUG_BLOCKQUANTIFY
const int bs = bcfhelpers::getInfoInt(_impl->hdr, *current_bs_start, "BS", -1);
std::string values;
for(float x : tp_qqs)
{
values += std::to_string(x) + ",";
}
std::cerr << "BS: " << bs << " T_QQ = " << t_qq << " [" << values << "]" << "\n";
#endif
};
for(auto v_it = _impl->variants.begin(); v_it != _impl->variants.end(); ++v_it)
{
// update fields, must output GA4GH-compliant fields
countVariants(*v_it);
// determine benchmarking superlocus
const std::string vchr = bcfhelpers::getChrom(_impl->hdr, *v_it);
const int vbs = bcfhelpers::getInfoInt(_impl->hdr, *v_it, "BS");
if(!current_bs_valid)
{
current_bs = vbs;
current_chr = vchr;
current_bs_valid = true;
}
#ifdef DEBUG_BLOCKQUANTIFY
std::cerr << "current BS = " << current_bs << " vbs = " << vbs << "\n";
#endif
if( current_bs_start != v_it
&& (vbs != current_bs || vbs < 0 || vchr != current_chr))
{
update_bs_qq(v_it);
current_bs = vbs;
current_chr = vchr;
current_bs_start = v_it;
}
}
// write out final superlocus (if any)
update_bs_qq(_impl->variants.end());
for(auto & v : _impl->variants)
{
#ifdef DEBUG_BLOCKQUANTIFY
lastpos = v->pos;
#endif
// use BD and BVT to make ROCs
rocEvaluate(v);
}
#ifdef DEBUG_BLOCKQUANTIFY
std::cerr << "finished block " << lastpos << " - " << _impl->variants.size() << " records on thread " << std::this_thread::get_id() << "\n";
#endif
_impl->fasta_to_use.reset(nullptr);
}
void BlockQuantify::count()
{
_impl->fasta_to_use.reset(new FastaFile(_impl->ref_fasta));
#ifdef DEBUG_BLOCKQUANTIFY
int lastpos = 0;
std::cerr << "starting block." << "\n";
#endif
auto current_bs_start = _impl->variants.begin();
std::string current_chr;
int current_bs = -1;
bool current_bs_valid = false;
// function to compute the QQ values for truth variants in the current
// benchmarking superlocus
const auto update_bs_filters = [this, ¤t_bs_start](BlockQuantifyImpl::variantlist_t::iterator to)
{
std::set<int> bs_filters;
for(auto cur = current_bs_start; cur != to; ++cur)
{
for(int nf = 0; nf < (*cur)->d.n_flt; ++nf)
{
const int f = (*cur)->d.flt[nf];
if(f != bcf_hdr_id2int(_impl->hdr, BCF_DT_ID, "PASS"))
{
bs_filters.insert(f);
}
}
}
if(bs_filters.empty())
{
return;
}
for(auto cur = current_bs_start; cur != to; ++cur)
{
const std::string bdt = bcfhelpers::getFormatString(_impl->hdr, *cur, "BD", 0);
const std::string bvq = bcfhelpers::getFormatString(_impl->hdr, *cur, "BVT", 1);
// filter TPs where the query call in NOCALL
if(bdt == "TP" && bvq == "NOCALL")
{
for(auto f : bs_filters)
{
bcf_add_filter(_impl->hdr, *cur, f);
}
}
}
};
// function to compute the QQ values for truth variants in the current
// benchmarking superlocus
const auto update_bs_qq = [this, ¤t_bs_start](BlockQuantifyImpl::variantlist_t::iterator to)
{
std::vector<float> tp_qqs;
for(auto cur = current_bs_start; cur != to; ++cur)
{
const float qqq = bcfhelpers::getFormatFloat(_impl->hdr, *cur, "QQ", 1);
if(std::isnan(qqq))
{
continue;
}
const std::string bd = bcfhelpers::getFormatString(_impl->hdr, *cur, "BD", 1);
// we want the scores of all TPs in this BS
if(bd == "TP")
{
tp_qqs.push_back(qqq);
}
}
float t_qq = bcfhelpers::missing_float();
if(!tp_qqs.empty())
{
t_qq = *(std::min_element(tp_qqs.begin(), tp_qqs.end()));
}
/** compute the median over all variants */
int fsize = bcf_hdr_nsamples(_impl->hdr);
float * fmt = (float*)calloc((size_t) fsize, sizeof(float));
for(auto cur = current_bs_start; cur != to; ++cur)
{
const std::string bd = bcfhelpers::getFormatString(_impl->hdr, *cur, "BD", 0);
bcf_get_format_float(_impl->hdr, *cur, "QQ", &fmt, &fsize);
if(bd != "TP")
{
fmt[0] = bcfhelpers::missing_float();
}
else
{
const float qqq = bcfhelpers::getFormatFloat(_impl->hdr, *cur, "QQ", 1);
const std::string bd = bcfhelpers::getFormatString(_impl->hdr, *cur, "BD", 1);
if(bd == "TP" && !std::isnan(qqq))
{
fmt[0] = qqq;
}
else
{
fmt[0] = t_qq;
}
}
bcf_update_format_float(_impl->hdr, *cur, "QQ", fmt, fsize);
}
free(fmt);
#ifdef DEBUG_BLOCKQUANTIFY
const int bs = bcfhelpers::getInfoInt(_impl->hdr, *current_bs_start, "BS", -1);
std::string values;
for(float x : tp_qqs)
{
values += std::to_string(x) + ",";
}
std::cerr << "BS: " << bs << " T_QQ = " << t_qq << " [" << values << "]" << "\n";
#endif
};
const auto update_bs_conf_boundary_flag = [this, ¤t_bs_start](BlockQuantifyImpl::variantlist_t::iterator to)
{
static const int has_conf = 1;
static const int has_non_conf = 2;
int conf_non_conf = 0;
for(auto cur = current_bs_start; cur != to; ++cur)
{
const std::string regions = bcfhelpers::getInfoString(_impl->hdr, *cur, "Regions", "");
if(regions.find("CONF") == std::string::npos)
{
conf_non_conf |= has_non_conf;
}
else
{
conf_non_conf |= has_conf;
}
if(regions.find("TS_boundary") != std::string::npos)
{
conf_non_conf |= has_non_conf | has_conf;
}
}
for(auto cur = current_bs_start; cur != to; ++cur)
{
const std::string regions = bcfhelpers::getInfoString(_impl->hdr, *cur, "Regions", "");
if(conf_non_conf == (has_conf | has_non_conf))
{
if(regions.find("TS_boundary") == std::string::npos)
{
bcf_update_info_string(_impl->hdr,
*cur, "Regions",
(regions.empty() ? "TS_boundary" :
regions + ",TS_boundary").c_str());
}
}
else if(conf_non_conf == has_conf)
{
if(regions.find("TS_contained") == std::string::npos)
{
// also flag fully confident superloci
bcf_update_info_string(_impl->hdr,
*cur, "Regions",
(regions.empty() ? "TS_contained" :
regions + ",TS_contained").c_str());
}
}
}
};
for(auto v_it = _impl->variants.begin(); v_it != _impl->variants.end(); ++v_it)
{
// update fields, must output GA4GH-compliant fields
countVariants(*v_it);
// determine benchmarking superlocus
const std::string vchr = bcfhelpers::getChrom(_impl->hdr, *v_it);
const int vbs = bcfhelpers::getInfoInt(_impl->hdr, *v_it, "BS");
if(!current_bs_valid)
{
current_bs = vbs;
current_chr = vchr;
current_bs_valid = true;
}
#ifdef DEBUG_BLOCKQUANTIFY
std::cerr << "current BS = " << current_bs << " vbs = " << vbs << "\n";
#endif
if( current_bs_start != v_it
&& (vbs != current_bs || vbs < 0 || vchr != current_chr))
{
#ifdef DEBUG_BLOCKQUANTIFY
std::cerr << "finishing BS = " << current_bs << " vbs = " << vbs << "\n";
#endif
update_bs_qq(v_it);
update_bs_filters(v_it);
update_bs_conf_boundary_flag(v_it);
current_bs = vbs;
current_chr = vchr;
current_bs_start = v_it;
}
}
// do final superlocus (if any)
update_bs_qq(_impl->variants.end());
update_bs_filters(_impl->variants.end());
update_bs_conf_boundary_flag(_impl->variants.end());
for(auto & v : _impl->variants)
{
#ifdef DEBUG_BLOCKQUANTIFY
lastpos = v->pos;
#endif
// use BD and BVT to make ROCs
rocEvaluate(v);
}
#ifdef DEBUG_BLOCKQUANTIFY
std::cerr << "finished block " << lastpos << " - " << _impl->variants.size() << " records on thread " << std::this_thread::get_id() << "\n";
#endif
_impl->fasta_to_use.reset(nullptr);
}