static
float calc_energy(const Aln& aln, const std::string& s)
{
Vienna::eos_debug = -1;
return aln.energy_of_struct(s);
}
int
centroid_alifold(int argc, char* argv[])
{
std::vector<float> gamma;
std::vector<float> th;
std::vector<std::string> engine;
std::vector<float> mix_w;
std::string input;
std::vector<std::string> model;
float p_th=0.0;
std::string p_outname;
std::string outname;
std::string ps_outname;
uint max_bp_dist;
std::string param;
uint max_clusters;
uint num_samples=0;
uint seed;
//
int num_ea_samples = -1;
int max_mcc = -1;
// parse command line options
po::options_description desc("Options");
desc.add_options()
("help,h", "show this message")
#ifdef HAVE_LIBRNA
("engine,e", po::value<std::vector<std::string> >(&engine),
"specify the inference engine (default: \"McCaskill & Alifold\")")
#else
("engine,e", po::value<std::vector<std::string> >(&engine),
"specify the inference engine (default: \"CONTRAfold\")")
#endif
("mixture,w", po::value<std::vector<float> >(&mix_w), "mixture weights of inference engines")
("gamma,g", po::value<std::vector<float> >(&gamma), "weight of base pairs")
("threshold,t", po::value<std::vector<float> >(&th),
"thereshold of base pairs (this option overwrites 'gamma')")
//
("ea", po::value<int>(&num_ea_samples),
"compute (pseudo-)expected accuracy (pseudo if arg==0, sampling if arg>0; arg: # of sampling)")
("max-mcc", po::value<int>(&max_mcc),
"predict secondary structure by maximizing pseudo-expected MCC (arg: # of sampling)")
// added by M. Hamada
("mea", "run as an MEA estimator")
("noncanonical", "allow non-canonical base-pairs")
("constraints,C", "use structure constraints")
("output,o", po::value<std::string>(&outname),
"specify filename to output predicted secondary structures. If empty, use the standard output.")
("posteriors", po::value<float>(&p_th),
"output base-pairing probability matrices which contain base-pairing probabilities more than the given value.")
("posteriors-output", po::value<std::string>(&p_outname),
"specify filename to output base-pairing probability matrices. If empty, use the standard output.")
("postscript", po::value<std::string>(&ps_outname),
"draw predicted secondary structures with the postscript (PS) format")
/*("monochrome", "draw the postscript with monochrome")*/
("params", po::value<std::string>(¶m), "use the parameter file");
po::options_description opts_contrafold("Options for CONTRAfold model");
opts_contrafold.add_options()
#if 0 // HAVE_LIBRNA // move to hidden options
("alipf_fold", "use alipf_fold base-pairing probabilities rather than those of CONTRAfold model")
("pf_fold", "use pf_fold base-pairing probabilities rather than those of CONTRAfold model")
#endif
("max-dist,d", po::value<uint>(&max_bp_dist)->default_value(0),
"the maximum distance of base-pairs");
po::options_description opts_sampling("Options for sampling");
opts_sampling.add_options()
("sampling,s",
po::value<uint>(&num_samples),
"specify the number of samples to be generated for each sequence")
("max-clusters,c",
po::value<uint>(&max_clusters)->default_value(10),
"the maximum number of clusters for the stochastic sampling algorithm")
("seed",
po::value<uint>(&seed)->default_value(0),
"specify the seed for the random number generator (set this automatically if seed=0)");
po::options_description opts("Options");
opts.add_options()
#ifdef HAVE_LIBRNA
("alipf_fold", "use alipf_fold base-pairing probabilities rather than those of CONTRAfold model")
("pf_fold", "use pf_fold base-pairing probabilities rather than those of CONTRAfold model")
#endif
("aux", "use auxiliary base-pairing probabilities")
("monochrome", "draw the postscript with monochrome")
("seq-file", po::value<std::string>(&input), "training sequence filename")
("model-file", po::value<std::vector<std::string> >(&model), "model filename");
opts.add(desc);
opts.add(opts_contrafold);
opts.add(opts_sampling);
po::positional_options_description pd;
pd.add("seq-file", 1); pd.add("model-file", -1);
po::variables_map vm;
bool usage=false;
try {
po::parsed_options parsed =
po::command_line_parser(argc, argv).options(opts).positional(pd).run();
po::store(parsed, vm);
po::notify(vm);
} catch (...) {
usage=true;
}
if (usage || vm.count("help") || !vm.count("seq-file") ||
(vm.count("aux") && model.empty()))
{
std::string features("CONTRAfold");
#ifdef HAVE_LIBRNA
features += ", McCaskill";
features += ", Alifold";
#endif
features += ", pfold";
features += ", AUX";
std::cout << "CentroidAlifold v" << VERSION
<< " for predicting common RNA secondary structures" << std::endl
<< " (available engines: " << features << ")" << std::endl
<< "Usage:" << std::endl
<< " " << argv[0]
<< " [options] seq [bp_matrix ...]\n\n"
<< desc << std::endl
<< opts_contrafold << std::endl
<< opts_sampling << std::endl;
return 1;
}
BOOST_SPIRIT_CLASSIC_NS::file_iterator<> fi(input.c_str());
if (!fi)
{
perror(input.c_str());
return 1;
}
if (th.size()>0)
{
gamma.resize(th.size());
for (uint i=0; i!=th.size(); ++i)
gamma[i] = 1.0/th[i]-1.0;
}
if (gamma.size()==1 && gamma[0]<0.0)
{
float g[] = { 0.03125, 0.0625, 0.125, 0.25, 0.5, 1.0, 2.0, 4.0, 6.0,
8.0, 16.0, 32.0, 64.0, 128.0, 256.0, 512.0, 1024.0 };
gamma.resize(boost::size(g));
std::copy(boost::begin(g), boost::end(g), gamma.begin());
}
if (engine.empty())
{
#ifdef HAVE_LIBRNA
engine.push_back("McCaskill");
engine.push_back("Alifold");
#else
engine.push_back("CONTRAfold");
#endif
}
if (vm.count("pf_fold")) { engine.resize(1); engine[0]="McCaskill"; }
if (vm.count("alipf_fold")) { engine.resize(1); engine[0]="Alifold"; }
if (vm.count("aux")) { engine.resize(1); engine[0]="AUX"; }
FoldingEngine<Aln>* cf=NULL;
std::vector<FoldingEngine<Aln>*> cf_list(engine.size(), NULL);
std::vector<FoldingEngine<std::string>*> src_list(engine.size(), NULL);
for (uint i=0; i!=engine.size(); ++i)
{
if (engine[i]=="CONTRAfold")
{
if (gamma.empty()) gamma.push_back(vm.count("mea") ? 6.0 : 4.0);
src_list[i] = new CONTRAfoldModel(param, !vm.count("noncanonical"), max_bp_dist, seed, vm.count("mea"));
cf_list[i] = new AveragedModel(src_list[i], max_bp_dist, vm.count("mea"));
}
else if (engine[i]=="CONTRAfoldM")
{
if (gamma.empty()) gamma.push_back(vm.count("mea") ? 6.0 : 4.0);
cf_list[i] = new CONTRAfoldMultiModel(param, !vm.count("noncanonical"), max_bp_dist, seed, vm.count("mea"));
}
#ifdef HAVE_LIBRNA
else if (engine[i]=="McCaskill")
{
if (gamma.empty()) gamma.push_back(vm.count("mea") ? 6.0 : 2.0);
src_list[i] = new McCaskillModel(!vm.count("noncanonical"), max_bp_dist,
param.empty() ? NULL : param.c_str(),
seed, vm.count("mea"));
cf_list[i] = new AveragedModel(src_list[i], max_bp_dist, vm.count("mea"));
}
else if (engine[i]=="Alifold")
{
if (gamma.empty()) gamma.push_back(vm.count("mea") ? 6.0 : 2.0);
cf_list[i] = new AliFoldModel(!vm.count("noncanonical"), max_bp_dist,
param.empty() ? NULL : param.c_str(),
seed, vm.count("mea"));
}
#endif
else if (engine[i]=="pfold")
{
if (gamma.empty()) gamma.push_back(vm.count("mea") ? 6.0 : 1.0);
std::string pfold_bin_dir(getenv("PFOLD_BIN_DIR") ? getenv("PFOLD_BIN_DIR") : ".");
std::string awk_bin(getenv("AWK_BIN") ? getenv("AWK_BIN") : "mawk");
std::string sed_bin(getenv("SED_BIN") ? getenv("SED_BIN") : "sed");
cf_list[i] = new PfoldModel<Aln>(pfold_bin_dir, awk_bin, sed_bin, vm.count("mea"));
}
else if (engine[i]=="AUX")
{
if (gamma.empty()) gamma.push_back(vm.count("mea") ? 6.0 : 1.0);
src_list[i] = new AuxModel(model, vm.count("mea"));
cf_list[i] = new AveragedModel(src_list[i], 0, vm.count("mea"));
}
else
{
throw std::logic_error("unsupported engine");
}
}
if (engine.size()==1)
cf=cf_list[0];
else
{
if (gamma.empty()) gamma.push_back(vm.count("mea") ? 6.0 : 1.0);
std::vector<std::pair<FoldingEngine<Aln>*,float> > models;
for (uint i=0; i!=engine.size(); ++i)
{
if (engine.size()!=mix_w.size())
models.push_back(std::make_pair(cf_list[i], 1.0));
else
models.push_back(std::make_pair(cf_list[i], mix_w[i]));
}
cf = new MixtureModel<Aln>(models, vm.count("mea"));
}
std::ostream* out = &std::cout;
if (vm.count("output"))
{
out = new std::ofstream(outname.c_str());
if (out->fail())
{
perror(outname.c_str());
delete out;
return 1;
}
}
std::ostream* p_out = &std::cout;
if (vm.count("posteriors") && vm.count("posteriors-output") && !vm.count("sampling"))
{
p_out = new std::ofstream(p_outname.c_str());
if (p_out->fail())
{
perror(p_outname.c_str());
delete p_out;
return 1;
}
}
Aln aln;
uint n=0;
uint bytes=0;
uint total_bytes=0;
while ((bytes=aln.load(fi))>0)
{
n++;
total_bytes+=bytes;
std::list<std::string>::iterator seq;
for (seq=aln.seq().begin(); seq!=aln.seq().end(); ++seq)
{
std::replace(seq->begin(), seq->end(), 't', 'u');
std::replace(seq->begin(), seq->end(), 'T', 'U');
}
if (vm.count("constraints"))
{
std::cout << "Input constraints:" << std::endl;
std::string str;
std::getline(std::cin, str);
cf->set_constraint(str);
}
if (num_samples>0)
{
if (max_clusters>0)
cf->stochastic_fold(aln.name().front(), aln, num_samples, gamma, max_clusters, *out, p_outname, p_th);
else
cf->stochastic_fold(aln, num_samples, *out);
continue;
}
if (max_mcc>0)
{
cf->max_mcc_fold(aln.name().front(), aln, *out, max_mcc);
continue;
}
if (num_ea_samples>=0)
cf->centroid_fold(aln.name().front(), aln, gamma, *out, num_ea_samples);
else
cf->centroid_fold(aln.name().front(), aln, gamma, *out);
if (vm.count("posteriors")) cf->get_bp().save(*p_out, aln.consensus(), p_th);
if (!ps_outname.empty())
{
char buf[PATH_MAX];
if (n==1)
strncpy(buf, ps_outname.c_str(), sizeof(buf));
else
snprintf(buf, sizeof(buf), "%s-%d", ps_outname.c_str(), n-1);
cf->ps_plot(std::string(buf), aln, gamma[0], !vm.count("monochrome"));
}
}
if (fi!=fi.make_end())
std::cout << "parse error after " << total_bytes << " bytes were loaded" << std::endl;
if (engine.size()!=1 && cf) delete cf;
for (uint i=0; i!=cf_list.size(); ++i) if (cf_list[i]) delete cf_list[i];
for (uint i=0; i!=src_list.size(); ++i) if (src_list[i]) delete src_list[i];
if (out != &std::cout) delete out;
if (p_out != &std::cout) delete p_out;
return 0;
}
static
std::string get_seq(const Aln& aln) { return aln.consensus(); }
