int CAlignmentRefiner::Run(void)
{
CCdd cdd;
unsigned int alWidth;
unsigned int nBlocksFromAU;
string message;
// Get arguments
CArgs args = GetArgs();
string fname, err;
string basename = args["o"].AsString() + "_", suffix = ".cn3";
// Stream to results file, if provided, or cout
// (NOTE: "x_lg" is just a workaround for bug in SUN WorkShop 5.1 compiler)
ostream* x_lg = args["details"] ? &args["details"].AsOutputFile() : &cout;
ostream& detailsStream = *x_lg;
SetDiagStream(x_lg); // send all diagnostic messages to the same stream
// Set to info level here, not in main, to avoid info messages about missing log files.
SetDiagPostLevel(eDiag_Info);
// Set up details stream first...
if (args["details"]) {
string str;
detailsStream << args.Print(str) << endl << endl;
detailsStream << string(72, '=') << endl;
detailsStream.precision(2);
// if (args["qd"]) m_quietDetails = true;
}
// Get initial alignment (in a Cdd blob) from the input file;
// convert to an AlignmentUtility object.
if (!ReadCD(args["i"].AsString(), &cdd)) {
ERROR_MESSAGE_CL("error reading CD from input file " << args["i"].AsString());
return eRefinerResultCantReadCD;
}
AlignmentUtility* au = new AlignmentUtility(cdd.GetSequences(), cdd.GetSeqannot());
const BlockMultipleAlignment* bma = (au && au->Okay()) ? au->GetBlockMultipleAlignment() : NULL;
if (!bma) {
delete au;
ERROR_MESSAGE_CL("Found invalid alignment in CD " << args["i"].AsString());
return eRefinerResultAlignmentUtilityError;
}
nBlocksFromAU = bma->NAlignedBlocks();
alWidth = bma->AlignmentWidth();
TERSE_INFO_MESSAGE_CL("\nRows in alignment: " << bma->NRows());
TERSE_INFO_MESSAGE_CL("Alignment width : " << alWidth);
TERSE_INFO_MESSAGE_CL("Number of Aligned Blocks after IBM: " << nBlocksFromAU << "\n");
// Some general parameters...
m_nTrials = (unsigned) args["n"].AsInteger();
m_nCycles = args["nc"].AsInteger();
m_scoreDeviationThreshold = args["convScoreChange"].AsDouble();
m_quietMode = (args["q"]);
// if (m_quietMode) SetDiagPostLevel(eDiag_Error);
// Fill out data structure for leave-one-out parameters
// LOO is performed unless -no_LOO option used
RefinerResultCode looParamResult = ExtractLOOArgs(nBlocksFromAU, message);
if (looParamResult != eRefinerResultOK) {
ERROR_MESSAGE_CL(message);
return looParamResult;
}
// If using a fixed selection order based on row-scores of the
// initial alignment, no need to do multiple trials.
if (m_nTrials > 1 && m_loo.selectorCode != eRandomSelectionOrder) {
m_nTrials = 1;
WARNING_MESSAGE_CL("For deterministic row-selection order, multiple trials are redundant.\nSetting number of trials to one and continuing.\n");
}
//EchoSettings(detailsStream, true, false);
// Fill out data structure for block editing parameters.
// By default, edit blocks -- must explicitly skip with the -be_fix option.
RefinerResultCode beParamResult = ExtractBEArgs(nBlocksFromAU, message);
if (beParamResult != eRefinerResultOK) {
ERROR_MESSAGE_CL(message);
return beParamResult;
}
//EchoSettings(detailsStream, false, true);
EchoSettings(detailsStream, true, true);
if (!m_blockEdit.editBlocks && !m_loo.doLOO) {
ERROR_MESSAGE_CL("Nothing will happen as both LOO and block editing have been disabled. Stopping");
return eRefinerResultInconsistentArgumentCombination;
}
// Perform the refinement...
TRACE_MESSAGE_CL("Entering refiner engine...\n");
CBMARefinerEngine refinerEngine(m_loo, m_blockEdit, m_nCycles, m_nTrials, true, !m_quietMode, m_scoreDeviationThreshold);
RefinerResultCode result = refinerEngine.Refine(au, &detailsStream);
// Output final statistics and refined alignments
// Get results from all trials; use reverse iterator to get them
// out of the map in order of highest to lowest score.
unsigned int n = 0;
unsigned int trial;
unsigned int nToWrite = (m_nTrials > 1) ? args["nout"].AsInteger() : 1;
const RefinedAlignments& optimizedAlignments = refinerEngine.GetAllResults();
RefinedAlignmentsRevCIt rcit = optimizedAlignments.rbegin(), rend = optimizedAlignments.rend();
if (rcit != rend) {
detailsStream << endl << endl << "Original Alignment Score = " << refinerEngine.GetInitialScore() << endl;
detailsStream << endl << "Best Refined Alignments (in descending score order)\n\n";
} else {
detailsStream << endl << "No Refined Alignments found (?)\n\n";
}
for (; rcit != rend; ++rcit, ++n) {
trial = rcit->second.iteration;
if (rcit->second.au == NULL) {
detailsStream << "Problem in trial " << trial << " -> no refined alignment available." << endl << endl;
continue;
}
detailsStream << "Alignment " << n << ": Score = " << rcit->first << " (trial " << trial << ")" << endl;
if (n < nToWrite) {
err.erase();
cdd.SetSeqannot() = rcit->second.au->GetSeqAnnots();
// write output as a CD in a new file
fname = basename + NStr::UIntToString(n) + "_trial" + NStr::UIntToString(trial) + suffix;
detailsStream << " (written to file '" << fname << "')";
if (!WriteASNToFile(fname.c_str(), cdd, args["ob"].HasValue(), &err)) {
ERROR_MESSAGE_CL("error writing output file " << fname);
}
}
detailsStream << endl;
}
// Destructor of RefinerEngine cleans up map of optimized alignments
// once it goes out of scope.
// delete au;
// delete auOriginal;
if (args["details"]) args["details"].CloseFile();
return result;
}
