/**
* @param arg_fileName: a string character
* @param arg_geneFile: which gene file to use
* @param arg_geneName: which gene we are interested. (NOTE: only first one gene
* is used).
*/
SEXP impl_readBGENToListByGene(SEXP arg_fileName, SEXP arg_geneFile,
SEXP arg_geneName) {
// begin
std::string FLAG_fileName = CHAR(STRING_ELT(arg_fileName, 0));
std::string FLAG_geneFile = CHAR(STRING_ELT(arg_geneFile, 0));
std::string FLAG_geneName = CHAR(STRING_ELT(arg_geneName, 0));
OrderedMap<std::string, std::string> geneRange;
loadGeneFile(FLAG_geneFile, FLAG_geneName, &geneRange);
std::string range;
int n = geneRange.size();
for (int i = 0; i < n; ++i) {
if (range.size() > 0) {
range += ",";
}
range += geneRange.valueAt(i);
}
REprintf("range = %s\n", range.c_str());
BGenFile bin(FLAG_fileName.c_str());
if (range.size())
bin.setRangeList(range.c_str());
else {
error("Please provide a valid gene name before we can continue.\n");
};
return readBGEN2List(&bin);
} // end readBGEN2List
/**
* @param arg_fileName: a string character
* @param arg_geneFile: which gene file to use
* @param arg_geneName: which gene we are interested. (just allow One gene
* name).
*/
SEXP impl_readBGENToMatrixByGene(SEXP arg_fileName, SEXP arg_geneFile,
SEXP arg_geneName) {
SEXP ans = R_NilValue;
std::string FLAG_fileName = CHAR(STRING_ELT(arg_fileName, 0));
std::string FLAG_geneFile = CHAR(STRING_ELT(arg_geneFile, 0));
std::vector<std::string> FLAG_geneName;
extractStringArray(arg_geneName, &FLAG_geneName);
if (FLAG_fileName.size() == 0) {
error("Please provide BGEN file name");
}
if (FLAG_geneName.size() && FLAG_geneFile.size() == 0) {
error("Please provide gene file name when extract genotype by gene");
}
int nGene = FLAG_geneName.size();
Rprintf("%d region to be extracted.\n", nGene);
int numAllocated = 0;
// allocate return value
PROTECT(ans = allocVector(VECSXP, nGene));
numAllocated++;
setListNames(FLAG_geneName, &ans);
OrderedMap<std::string, std::string> geneRange;
loadGeneFile(FLAG_geneFile, FLAG_geneName, &geneRange);
for (int i = 0; i < nGene; ++i) {
// REprintf("range = %s\n", FLAG_geneName[i].c_str());
const std::string& range = geneRange[FLAG_geneName[i]];
// Rprintf( "range = %s\n", range.c_str());
BGenFile bin(FLAG_fileName);
if (range.size())
bin.setRangeList(range.c_str());
else {
warning("Gene name [ %s ] does not exists in provided gene file",
FLAG_geneName[i].c_str());
UNPROTECT(numAllocated);
return (ans);
};
// real working part
SET_VECTOR_ELT(ans, i, readBGEN2Matrix(&bin));
}
UNPROTECT(numAllocated);
return ans;
}
int main(int argc, char** argv) {
PARSE_PARAMETER(argc, argv);
if (FLAG_help) {
PARAMETER_HELP();
return 0;
}
welcome();
PARAMETER_STATUS();
if (FLAG_REMAIN_ARG.size() > 0) {
fprintf(stderr, "Unparsed arguments: ");
for (unsigned int i = 0; i < FLAG_REMAIN_ARG.size(); i++) {
fprintf(stderr, " %s", FLAG_REMAIN_ARG[i].c_str());
}
exit(1);
}
if (!FLAG_outPrefix.size()) FLAG_outPrefix = "rvtest";
if ((FLAG_inVcf.empty() ? 0 : 1) + (FLAG_inBgen.empty() ? 0 : 1) +
(FLAG_inKgg.empty() ? 0 : 1) !=
1) {
fprintf(stderr,
"Please provide one type of input file using: --inVcf, --inBgen or "
"--inKgg\n");
exit(1);
}
// check new version
if (!FLAG_noweb) {
VersionChecker ver;
if (ver.retrieveRemoteVersion("http://zhanxw.com/rvtests/version") < 0) {
fprintf(stderr,
"Retrieve remote version failed, use '--noweb' to skip.\n");
} else {
ver.setLocalVersion(VERSION);
if (ver.isRemoteVersionNewer()) {
fprintf(stderr, "New version of rvtests is available:");
ver.printRemoteContent();
}
}
}
// start logging
Logger _logger((FLAG_outPrefix + ".log").c_str());
logger = &_logger;
logger->info("Program version: %s", VERSION);
logger->infoToFile("Git Version: %s", GIT_VERSION);
logger->infoToFile("Parameters BEGIN");
PARAMETER_INSTANCE().WriteToFile(logger->getHandle());
logger->infoToFile("Parameters END");
logger->sync();
// set up multithreading
#ifdef _OPENMP
if (FLAG_numThread <= 0) {
fprintf(stderr, "Invalid number of threads [ %d ], reset to single thread",
FLAG_numThread);
omp_set_num_threads(1);
} else if (FLAG_numThread > omp_get_max_threads()) {
int maxThreads = omp_get_max_threads();
fprintf(stderr,
"Reduced your specified number of threads to the maximum of system "
"limit [ %d ]",
maxThreads);
omp_set_num_threads(maxThreads);
} else if (FLAG_numThread == 1) {
// need to set to one thread, otherwise all CPUs may be used
omp_set_num_threads(1);
} else {
logger->info("Set number of threads = [ %d ]", FLAG_numThread);
omp_set_num_threads(FLAG_numThread);
}
#endif
// start analysis
time_t startTime = time(0);
logger->info("Analysis started at: %s", currentTime().c_str());
GenotypeExtractor* ge = NULL;
if (!FLAG_inVcf.empty()) {
ge = new VCFGenotypeExtractor(FLAG_inVcf);
} else if (!FLAG_inBgen.empty()) {
ge = new BGenGenotypeExtractor(FLAG_inBgen, FLAG_inBgenSample);
} else if (!FLAG_inKgg.empty()) {
ge = new KGGGenotypeExtractor(FLAG_inKgg);
} else {
assert(false);
}
// set range filters here
ge->setRangeList(FLAG_rangeList.c_str());
ge->setRangeFile(FLAG_rangeFile.c_str());
// set people filters here
if (FLAG_peopleIncludeID.size() || FLAG_peopleIncludeFile.size()) {
ge->excludeAllPeople();
ge->includePeople(FLAG_peopleIncludeID.c_str());
ge->includePeopleFromFile(FLAG_peopleIncludeFile.c_str());
}
ge->excludePeople(FLAG_peopleExcludeID.c_str());
ge->excludePeopleFromFile(FLAG_peopleExcludeFile.c_str());
if (!FLAG_siteFile.empty()) {
ge->setSiteFile(FLAG_siteFile);
logger->info("Restrict analysis based on specified site file [ %s ]",
FLAG_siteFile.c_str());
}
if (FLAG_siteDepthMin > 0) {
ge->setSiteDepthMin(FLAG_siteDepthMin);
logger->info("Set site depth minimum to %d", FLAG_siteDepthMin);
}
if (FLAG_siteDepthMax > 0) {
ge->setSiteDepthMax(FLAG_siteDepthMax);
logger->info("Set site depth maximum to %d", FLAG_siteDepthMax);
}
if (FLAG_siteMACMin > 0) {
ge->setSiteMACMin(FLAG_siteMACMin);
logger->info("Set site minimum MAC to %d", FLAG_siteDepthMin);
}
if (FLAG_annoType != "") {
ge->setAnnoType(FLAG_annoType.c_str());
logger->info("Set annotype type filter to %s", FLAG_annoType.c_str());
}
std::vector<std::string> vcfSampleNames;
ge->getPeopleName(&vcfSampleNames);
logger->info("Loaded [ %zu ] samples from genotype files",
vcfSampleNames.size());
DataLoader dataLoader;
dataLoader.setPhenotypeImputation(FLAG_imputePheno);
dataLoader.setCovariateImputation(FLAG_imputeCov);
if (FLAG_multiplePheno.empty()) {
dataLoader.loadPhenotype(FLAG_pheno, FLAG_mpheno, FLAG_phenoName);
// // load phenotypes
// std::map<std::string, double> phenotype;
// if (FLAG_pheno.empty()) {
// logger->error("Cannot do association when phenotype is missing!");
// return -1;
// }
// // check if alternative phenotype columns are used
// if (!FLAG_mpheno.empty() && !FLAG_phenoName.empty()) {
// logger->error("Please specify either --mpheno or --pheno-name");
// return -1;
// }
// if (!FLAG_mpheno.empty()) {
// int col = atoi(FLAG_mpheno);
// int ret = loadPedPhenotypeByColumn(FLAG_pheno.c_str(), &phenotype,
// col);
// if (ret < 0) {
// logger->error("Loading phenotype failed!");
// return -1;
// }
// } else if (!FLAG_phenoName.empty()) {
// int ret = loadPedPhenotypeByHeader(FLAG_pheno.c_str(), &phenotype,
// FLAG_phenoName.c_str());
// if (ret < 0) {
// logger->error("Loading phenotype failed!");
// return -1;
// }
// } else {
// int col = 1; // default use the first phenotype
// int ret = loadPedPhenotypeByColumn(FLAG_pheno.c_str(), &phenotype,
// col);
// if (ret < 0) {
// logger->error("Loading phenotype failed!");
// return -1;
// }
// }
// logger->info("Loaded [ %zu ] sample phenotypes.", phenotype.size());
// rearrange phenotypes
// drop samples from phenotype or vcf
matchPhenotypeAndVCF("missing phenotype", &dataLoader, ge);
// // phenotype names (vcf sample names) arranged in the same order as in
// VCF
// std::vector<std::string> phenotypeNameInOrder;
// std::vector<double>
// phenotypeInOrder; // phenotype arranged in the same order as in VCF
// rearrange(phenotype, vcfSampleNames, &vcfSampleToDrop,
// &phenotypeNameInOrder,
// &phenotypeInOrder, FLAG_imputePheno);
// if (vcfSampleToDrop.size()) {
// // exclude this sample from parsing VCF
// ge->excludePeople(vcfSampleToDrop);
// // output dropped samples
// for (size_t i = 0; i < vcfSampleToDrop.size(); ++i) {
// if (i == 0)
// logger->warn(
// "Total [ %zu ] samples are dropped from VCF file due to missing
// "
// "phenotype",
// vcfSampleToDrop.size());
// if (i >= 10) {
// logger->warn(
// "Skip outputting additional [ %d ] samples with missing "
// "phenotypes.",
// ((int)vcfSampleToDrop.size() - 10));
// break;
// }
// logger->warn("Drop sample [ %s ] from VCF file due to missing
// phenotype",
// (vcfSampleToDrop)[i].c_str());
// }
// // logger->warn("Drop %zu sample from VCF file since we don't have
// their
// // phenotypes", vcfSampleToDrop.size());
// }
// if (phenotypeInOrder.size() != phenotype.size()) {
// logger->warn(
// "Drop [ %d ] samples from phenotype file due to missing genotypes
// from "
// "VCF files",
// (int)(phenotype.size() - phenotypeInOrder.size()));
// // We may output these samples by comparing keys of phenotype and
// // phenotypeNameInOrder
// }
dataLoader.loadCovariate(FLAG_cov, FLAG_covName);
matchCovariateAndVCF("missing covariate", &dataLoader, ge);
// // load covariate
// Matrix covariate;
// HandleMissingCov handleMissingCov = COVARIATE_DROP;
// if (FLAG_imputeCov) {
// handleMissingCov = COVARIATE_IMPUTE;
// }
// if (FLAG_cov.empty() && !FLAG_covName.empty()) {
// logger->info("Use phenotype file as covariate file [ %s ]",
// FLAG_pheno.c_str());
// FLAG_cov = FLAG_pheno;
// }
// if (!FLAG_cov.empty()) {
// logger->info("Begin to read covariate file.");
// std::vector<std::string> columnNamesInCovariate;
// std::set<std::string> sampleToDropInCovariate;
// int ret = loadCovariate(FLAG_cov.c_str(), phenotypeNameInOrder,
// FLAG_covName.c_str(), handleMissingCov,
// &covariate,
// &columnNamesInCovariate,
// &sampleToDropInCovariate);
// if (ret < 0) {
// logger->error("Load covariate file failed !");
// exit(1);
// }
// // drop phenotype samples
// if (!sampleToDropInCovariate.empty()) {
// int idx = 0;
// int n = phenotypeNameInOrder.size();
// for (int i = 0; i < n; ++i) {
// if (sampleToDropInCovariate.count(phenotypeNameInOrder[i]) !=
// 0) { // need to drop
// continue;
// }
// phenotypeNameInOrder[idx] = phenotypeNameInOrder[i];
// phenotypeInOrder[idx] = phenotypeInOrder[i];
// idx++;
// }
// phenotypeNameInOrder.resize(idx);
// phenotypeInOrder.resize(idx);
// logger->warn(
// "[ %zu ] sample phenotypes are dropped due to lacking
// covariates.",
// sampleToDropInCovariate.size());
// }
// // drop vcf samples;
// for (std::set<std::string>::const_iterator iter =
// sampleToDropInCovariate.begin();
// iter != sampleToDropInCovariate.end(); ++iter) {
// ge->excludePeople(iter->c_str());
// }
// }
} else {
dataLoader.loadMultiplePhenotype(FLAG_multiplePheno, FLAG_pheno, FLAG_cov);
matchPhenotypeAndVCF("missing phenotype", &dataLoader, ge);
matchCovariateAndVCF("missing covariate", &dataLoader, ge);
}
dataLoader.loadSex();
if (FLAG_sex) {
dataLoader.useSexAsCovariate();
matchCovariateAndVCF("missing sex", &dataLoader, ge);
}
// // load sex
// std::vector<int> sex;
// if (loadSex(FLAG_pheno, phenotypeNameInOrder, &sex)) {
// logger->error("Cannot load sex of samples from phenotype file");
// exit(1);
// }
// if (FLAG_sex) { // append sex in covariate
// std::vector<int> index; // mark missing samples
// int numMissing = findMissingSex(sex, &index);
// logger->info("Futher exclude %d samples with missing sex", numMissing);
// removeByIndex(index, &sex);
// excludeSamplesByIndex(index, &ge, &phenotypeNameInOrder,
// &phenotypeInOrder,
// &covariate);
// appendToMatrix("Sex", sex, &covariate);
// }
if (!FLAG_condition.empty()) {
dataLoader.loadMarkerAsCovariate(FLAG_inVcf, FLAG_condition);
matchCovariateAndVCF("missing in conditioned marker(s)", &dataLoader, ge);
}
// // load conditional markers
// if (!FLAG_condition.empty()) {
// Matrix geno;
// std::vector<std::string> rowLabel;
// if (loadMarkerFromVCF(FLAG_inVcf, FLAG_condition, &rowLabel, &geno) < 0)
// {
// logger->error("Load conditional markers [ %s ] from [ %s ] failed.",
// FLAG_condition.c_str(), FLAG_inVcf.c_str());
// exit(1);
// }
// if (appendGenotype(&covariate, phenotypeNameInOrder, geno, rowLabel) < 0)
// {
// logger->error(
// "Failed to combine conditional markers [ %s ] from [ %s ] failed.",
// FLAG_condition.c_str(), FLAG_inVcf.c_str());
// exit(1);
// }
// }
dataLoader.checkConstantCovariate();
// // check if some covariates are constant for all samples
// // e.g. user may include covariate "1" in addition to intercept
// // in such case, we will give a fatal error
// for (int i = 0; i < covariate.cols; ++i) {
// std::set<double> s;
// s.clear();
// for (int j = 0; j < covariate.rows; ++j) {
// s.insert(covariate(j,i));
// }
// if (s.size() == 1) {
// logger->error(
// "Covariate [ %s ] equals [ %g ] for all samples, cannot fit "
// "model...\n",
// covariate.GetColumnLabel(i), *s.begin());
// exit(1);
// }
// }
g_SummaryHeader = new SummaryHeader;
g_SummaryHeader->recordCovariate(dataLoader.getCovariate());
// record raw phenotype
g_SummaryHeader->recordPhenotype("Trait",
dataLoader.getPhenotype().extractCol(0));
// adjust phenotype
// bool binaryPhenotype;
if (FLAG_qtl) {
// binaryPhenotype = false;
dataLoader.setTraitType(DataLoader::PHENOTYPE_QTL);
logger->info("-- Force quantitative trait mode -- ");
} else {
if (dataLoader.detectPhenotypeType() == DataLoader::PHENOTYPE_BINARY) {
logger->warn("-- Enabling binary phenotype mode -- ");
dataLoader.setTraitType(DataLoader::PHENOTYPE_BINARY);
} else {
dataLoader.setTraitType(DataLoader::PHENOTYPE_QTL);
}
// binaryPhenotype = isBinaryPhenotype(phenotypeInOrder);
// if (binaryPhenotype) {
// logger->warn("-- Enabling binary phenotype mode -- ");
// convertBinaryPhenotype(&phenotypeInOrder);
// }
}
if (FLAG_useResidualAsPhenotype) {
dataLoader.useResidualAsPhenotype();
g_SummaryHeader->recordEstimation(dataLoader.getEstimation());
}
// // use residual as phenotype
// if (FLAG_useResidualAsPhenotype) {
// if (binaryPhenotype) {
// logger->warn(
// "WARNING: Skip transforming binary phenotype, although you want to
// "
// "use residual as phenotype!");
// } else {
// if (covariate.cols > 0) {
// LinearRegression lr;
// Vector pheno;
// Matrix covAndInt;
// copy(phenotypeInOrder, &pheno);
// copyCovariateAndIntercept(covariate.rows, covariate, &covAndInt);
// if (!lr.FitLinearModel(covAndInt, pheno)) {
// logger->error(
// "Cannot fit model: [ phenotype ~ 1 + covariates ], now use the
// "
// "original phenotype");
// } else {
// const int n = lr.GetResiduals().Length();
// for (int i = 0; i < n; ++i) {
// phenotypeInOrder[i] = lr.GetResiduals()[i];
// }
// covariate.Dimension(0, 0);
// logger->info(
// "DONE: Fit model [ phenotype ~ 1 + covariates ] and model "
// "residuals will be used as responses.");
// }
// } else { // no covaraites
// centerVector(&phenotypeInOrder);
// logger->info("DONE: Use residual as phenotype by centerng it");
// }
// }
// }
if (FLAG_inverseNormal) {
dataLoader.inverseNormalizePhenotype();
g_SummaryHeader->setInverseNormalize(FLAG_inverseNormal);
}
// // phenotype transformation
// if (FLAG_inverseNormal) {
// if (binaryPhenotype) {
// logger->warn(
// "WARNING: Skip transforming binary phenotype, although you required
// "
// "inverse normalization!");
// } else {
// logger->info("Now applying inverse normalize transformation.");
// inverseNormalizeLikeMerlin(&phenotypeInOrder);
// g_SummaryHeader->setInverseNormalize(FLAG_inverseNormal);
// logger->info("DONE: inverse normalization transformation finished.");
// }
// }
g_SummaryHeader->recordPhenotype("AnalyzedTrait",
dataLoader.getPhenotype().extractCol(0));
if (dataLoader.getPhenotype().nrow() == 0) {
logger->fatal("There are 0 samples with valid phenotypes, quitting...");
exit(1);
}
// if (phenotypeInOrder.empty()) {
// logger->fatal("There are 0 samples with valid phenotypes, quitting...");
// exit(1);
// }
logger->info("Analysis begins with [ %d ] samples...",
dataLoader.getPhenotype().nrow());
//////////////////////////////////////////////////////////////////////////////
// prepare each model
bool singleVariantMode = FLAG_modelSingle.size() || FLAG_modelMeta.size();
bool groupVariantMode = (FLAG_modelBurden.size() || FLAG_modelVT.size() ||
FLAG_modelKernel.size());
if (singleVariantMode && groupVariantMode) {
logger->error("Cannot support both single variant and region based tests");
exit(1);
}
ModelManager modelManager(FLAG_outPrefix);
// set up models in qtl/binary modes
if (dataLoader.isBinaryPhenotype()) {
modelManager.setBinaryOutcome();
matchPhenotypeAndVCF("missing phenotype (not case/control)", &dataLoader,
ge);
} else {
modelManager.setQuantitativeOutcome();
}
// create models
modelManager.create("single", FLAG_modelSingle);
modelManager.create("burden", FLAG_modelBurden);
modelManager.create("vt", FLAG_modelVT);
modelManager.create("kernel", FLAG_modelKernel);
modelManager.create("meta", FLAG_modelMeta);
if (FLAG_outputRaw) {
modelManager.create("outputRaw", "dump");
}
const std::vector<ModelFitter*>& model = modelManager.getModel();
const std::vector<FileWriter*>& fOuts = modelManager.getResultFile();
const size_t numModel = model.size();
// TODO: optimize this to avoid data copying
Matrix phenotypeMatrix;
Matrix covariate;
toMatrix(dataLoader.getPhenotype(), &phenotypeMatrix);
toMatrix(dataLoader.getCovariate(), &covariate);
// determine VCF file reading pattern
// current support:
// * line by line ( including range selection)
// * gene by gene
// * range by range
std::string rangeMode = "Single";
if (FLAG_geneFile.size() && (FLAG_setFile.size() || FLAG_setList.size())) {
logger->error("Cannot specify both gene file and set file.");
exit(1);
}
if (!FLAG_gene.empty() && FLAG_geneFile.empty()) {
logger->error("Please provide gene file for gene bases analysis.");
exit(1);
}
OrderedMap<std::string, RangeList> geneRange;
if (FLAG_geneFile.size()) {
rangeMode = "Gene";
int ret =
loadGeneFile(FLAG_geneFile.c_str(), FLAG_gene.c_str(), &geneRange);
if (ret < 0 || geneRange.size() == 0) {
logger->error("Error loading gene file or gene list is empty!");
return -1;
} else {
logger->info("Loaded [ %zu ] genes.", geneRange.size());
}
}
if (!FLAG_set.empty() && FLAG_setFile.empty()) {
logger->error("Please provide set file for set bases analysis.");
exit(1);
}
if (FLAG_setFile.size()) {
rangeMode = "Range";
int ret = loadRangeFile(FLAG_setFile.c_str(), FLAG_set.c_str(), &geneRange);
if (ret < 0 || geneRange.size() == 0) {
logger->error("Error loading set file or set list is empty!");
return -1;
} else {
logger->info("Loaded [ %zu ] set to tests.", geneRange.size());
}
}
if (FLAG_setList.size()) {
rangeMode = "Range";
int ret = appendListToRange(FLAG_setList, &geneRange);
if (ret < 0) {
logger->error("Error loading set list or set list is empty!");
return -1;
}
}
DataConsolidator dc;
dc.setSex(&dataLoader.getSex());
dc.setFormula(&dataLoader.getFormula());
dc.setGenotypeCounter(ge->getGenotypeCounter());
// load kinshp if needed by family models
if (modelManager.hasFamilyModel() ||
(!FLAG_modelMeta.empty() && !FLAG_kinship.empty())) {
logger->info("Family-based model specified. Loading kinship file...");
// process auto kinship
if (dc.setKinshipSample(dataLoader.getPhenotype().getRowName()) ||
dc.setKinshipFile(DataConsolidator::KINSHIP_AUTO, FLAG_kinship) ||
dc.setKinshipEigenFile(DataConsolidator::KINSHIP_AUTO,
FLAG_kinshipEigen) ||
dc.loadKinship(DataConsolidator::KINSHIP_AUTO)) {
logger->error(
"Failed to load autosomal kinship (you may use vcf2kinship to "
"generate one).");
exit(1);
}
if (dc.setKinshipFile(DataConsolidator::KINSHIP_X, FLAG_xHemiKinship) ||
dc.setKinshipEigenFile(DataConsolidator::KINSHIP_X,
FLAG_xHemiKinshipEigen) ||
dc.loadKinship(DataConsolidator::KINSHIP_X)) {
logger->warn(
"Autosomal kinship loaded, but no hemizygote region kinship "
"provided, some sex chromosome tests will be skipped.");
// keep the program going
}
} else if (!FLAG_kinship.empty() && FLAG_modelMeta.empty()) {
logger->info(
"Family-based model not specified. Options related to kinship will be "
"ignored here.");
}
// set imputation method
if (FLAG_impute.empty()) {
logger->info("Impute missing genotype to mean (by default)");
dc.setStrategy(DataConsolidator::IMPUTE_MEAN);
} else if (FLAG_impute == "mean") {
logger->info("Impute missing genotype to mean");
dc.setStrategy(DataConsolidator::IMPUTE_MEAN);
} else if (FLAG_impute == "hwe") {
logger->info("Impute missing genotype by HWE");
dc.setStrategy(DataConsolidator::IMPUTE_HWE);
} else if (FLAG_impute == "drop") {
logger->info("Drop missing genotypes");
dc.setStrategy(DataConsolidator::DROP);
}
dc.setPhenotypeName(dataLoader.getPhenotype().getRowName());
// set up par region
ParRegion parRegion(FLAG_xLabel, FLAG_xParRegion);
dc.setParRegion(&parRegion);
// genotype will be extracted and stored
if (FLAG_freqUpper > 0) {
ge->setSiteFreqMax(FLAG_freqUpper);
logger->info("Set upper minor allele frequency limit to %g",
FLAG_freqUpper);
}
if (FLAG_freqLower > 0) {
ge->setSiteFreqMin(FLAG_freqLower);
logger->info("Set lower minor allele frequency limit to %g",
FLAG_freqLower);
}
// handle sex chromosome
ge->setParRegion(&parRegion);
ge->setSex(&dataLoader.getSex());
// use dosage instead GT
if (!FLAG_dosageTag.empty()) {
ge->setDosageTag(FLAG_dosageTag);
logger->info("Use dosage genotype from VCF flag %s.",
FLAG_dosageTag.c_str());
}
// multi-allelic sites will be treats as ref/alt1, ref/alt2, ref/alt3..
// instead of ref/alt1 (biallelic)
if (FLAG_multiAllele) {
ge->enableMultiAllelicMode();
logger->info("Enable analysis using multiple allelic models");
}
// genotype QC options
if (FLAG_indvDepthMin > 0) {
ge->setGDmin(FLAG_indvDepthMin);
logger->info("Minimum GD set to %d (or marked as missing genotype).",
FLAG_indvDepthMin);
}
if (FLAG_indvDepthMax > 0) {
ge->setGDmax(FLAG_indvDepthMax);
logger->info("Maximum GD set to %d (or marked as missing genotype).",
FLAG_indvDepthMax);
}
if (FLAG_indvQualMin > 0) {
ge->setGQmin(FLAG_indvQualMin);
logger->info("Minimum GQ set to %d (or marked as missing genotype).",
FLAG_indvQualMin);
}
// e.g. check colinearity and correlations between predictors
dc.preRegressionCheck(phenotypeMatrix, covariate);
// prepare PLINK files for BoltLMM model
if (!FLAG_boltPlink.empty()) {
if (dc.prepareBoltModel(FLAG_boltPlink,
dataLoader.getPhenotype().getRowName(),
dataLoader.getPhenotype())) {
logger->error(
"Failed to prepare inputs for BOLT-LMM association test model with "
"this prefix [ %s ]!",
FLAG_boltPlink.c_str());
exit(1);
}
}
logger->info("Analysis started");
Result& buf = dc.getResult();
Matrix& genotype = dc.getOriginalGenotype();
// we have three modes:
// * single variant reading, single variant test
// * range variant reading, single variant test
// * range variant reading, group variant test
if (rangeMode == "Single" && singleVariantMode) { // use line by line mode
buf.addHeader("CHROM");
buf.addHeader("POS");
if (FLAG_outputID) {
buf.addHeader("ID");
}
buf.addHeader("REF");
buf.addHeader("ALT");
buf.addHeader("N_INFORMATIVE");
// output headers
for (size_t m = 0; m < model.size(); m++) {
model[m]->writeHeader(fOuts[m], buf);
}
int variantProcessed = 0;
while (true) {
buf.clearValue();
int ret = ge->extractSingleGenotype(&genotype, &buf);
if (ret == GenotypeExtractor::FILE_END) { // reach file end
break;
}
if (ret == GenotypeExtractor::FAIL_FILTER) {
continue;
}
if (ret != GenotypeExtractor::SUCCEED) {
logger->error("Extract genotype failed at site: %s:%s!",
buf["CHROM"].c_str(), buf["POS"].c_str());
continue;
}
if (genotype.cols == 0) {
logger->warn("Extract [ %s:%s ] has 0 variants, skipping",
buf["CHROM"].c_str(), buf["POS"].c_str());
continue;
}
++variantProcessed;
dc.consolidate(phenotypeMatrix, covariate, genotype);
buf.updateValue("N_INFORMATIVE", toString(genotype.rows));
// logger->info("Test variant at site: %s:%s!",
// buf["CHROM"].c_str(), buf["POS"].c_str());
// fit each model
for (size_t m = 0; m != numModel; m++) {
model[m]->reset();
model[m]->fit(&dc);
model[m]->writeOutput(fOuts[m], buf);
}
}
logger->info("Analyzed [ %d ] variants", variantProcessed);
} else if (rangeMode != "Single" &&
singleVariantMode) { // read by gene/range model, single variant
// test
buf.addHeader(rangeMode);
buf.addHeader("CHROM");
buf.addHeader("POS");
if (FLAG_outputID) {
buf.addHeader("ID");
}
buf.addHeader("REF");
buf.addHeader("ALT");
buf.addHeader("N_INFORMATIVE");
// output headers
for (size_t m = 0; m < numModel; m++) {
model[m]->writeHeader(fOuts[m], buf);
}
std::string geneName;
RangeList rangeList;
int variantProcessed = 0;
for (size_t i = 0; i < geneRange.size(); ++i) {
geneRange.at(i, &geneName, &rangeList);
ge->setRange(rangeList);
while (true) {
buf.clearValue();
int ret = ge->extractSingleGenotype(&genotype, &buf);
if (ret == GenotypeExtractor::FILE_END) { // reach end of this region
break;
}
if (ret == GenotypeExtractor::FAIL_FILTER) {
continue;
}
if (ret != GenotypeExtractor::SUCCEED) {
logger->error("Extract genotype failed for gene %s!",
geneName.c_str());
continue;
}
if (genotype.cols == 0) {
logger->warn("Gene %s has 0 variants, skipping", geneName.c_str());
continue;
}
++variantProcessed;
dc.consolidate(phenotypeMatrix, covariate, genotype);
buf.updateValue(rangeMode, geneName);
buf.updateValue("N_INFORMATIVE", genotype.rows);
// #pragma omp parallel for
for (size_t m = 0; m != numModel; m++) {
model[m]->reset();
model[m]->fit(&dc);
model[m]->writeOutput(fOuts[m], buf);
}
}
}
logger->info("Analyzed [ %d ] variants from [ %d ] genes/regions",
variantProcessed, (int)geneRange.size());
} else if (rangeMode != "Single" &&
groupVariantMode) { // read by gene/range mode, group variant
// test
buf.addHeader(rangeMode);
buf.addHeader("RANGE");
buf.addHeader("N_INFORMATIVE");
buf.addHeader("NumVar");
buf.addHeader("NumPolyVar");
// output headers
for (size_t m = 0; m < numModel; m++) {
model[m]->writeHeader(fOuts[m], buf);
}
std::string geneName;
RangeList rangeList;
int variantProcessed = 0;
ge->enableAutoMerge();
for (size_t i = 0; i < geneRange.size(); ++i) {
geneRange.at(i, &geneName, &rangeList);
ge->setRange(rangeList);
buf.clearValue();
int ret = ge->extractMultipleGenotype(&genotype);
if (ret != GenotypeExtractor::SUCCEED) {
logger->error("Extract genotype failed for gene %s!", geneName.c_str());
continue;
}
if (genotype.cols == 0) {
logger->info("Gene %s has 0 variants, skipping", geneName.c_str());
continue;
}
variantProcessed += genotype.cols; // genotype is people by marker
dc.consolidate(phenotypeMatrix, covariate, genotype);
buf.updateValue(rangeMode, geneName);
buf.updateValue("RANGE", rangeList.toString());
buf.updateValue("N_INFORMATIVE", genotype.rows);
buf.updateValue("NumVar", genotype.cols);
buf.updateValue("NumPolyVar",
dc.getFlippedToMinorPolymorphicGenotype().cols);
// #ifdef _OPENMP
// #pragma omp parallel for
// #endif
for (size_t m = 0; m != numModel; m++) {
model[m]->reset();
model[m]->fit(&dc);
model[m]->writeOutput(fOuts[m], buf);
}
}
logger->info("Analyzed [ %d ] variants from [ %d ] genes/regions",
variantProcessed, (int)geneRange.size());
} else {
logger->error(
"Unsupported reading mode and test modes! (need more parameters?)");
exit(1);
}
// Resource cleaning up
modelManager.close();
delete g_SummaryHeader;
time_t endTime = time(0);
logger->info("Analysis ends at: %s", currentTime().c_str());
int elapsedSecond = (int)(endTime - startTime);
logger->info("Analysis took %d seconds", elapsedSecond);
fputs("RVTESTS finished successfully\n", stdout);
return 0;
}
int main(int argc, char** argv) {
////////////////////////////////////////////////
BEGIN_PARAMETER_LIST(pl)
ADD_PARAMETER_GROUP(pl, "Basic Input/Output")
ADD_STRING_PARAMETER(pl, inVcf, "--inVcf", "Input VCF File")
ADD_STRING_PARAMETER(pl, outPrefix, "--out", "Output prefix")
ADD_BOOL_PARAMETER(pl, outputRaw, "--outputRaw",
"Output genotypes, phenotype, covariates(if any) and "
"collapsed genotype to tabular files")
ADD_PARAMETER_GROUP(pl, "Specify Covariate")
ADD_STRING_PARAMETER(pl, cov, "--covar", "Specify covariate file")
ADD_STRING_PARAMETER(
pl, covName, "--covar-name",
"Specify the column name in covariate file to be included in analysis")
ADD_BOOL_PARAMETER(pl, sex, "--sex",
"Include sex (5th column in the PED file) as a covariate")
ADD_PARAMETER_GROUP(pl, "Specify Phenotype")
ADD_STRING_PARAMETER(pl, pheno, "--pheno", "Specify phenotype file")
ADD_BOOL_PARAMETER(pl, inverseNormal, "--inverseNormal",
"Transform phenotype like normal distribution")
ADD_BOOL_PARAMETER(
pl, useResidualAsPhenotype, "--useResidualAsPhenotype",
"Fit covariate ~ phenotype, use residual to replace phenotype")
ADD_STRING_PARAMETER(pl, mpheno, "--mpheno",
"Specify which phenotype column to read (default: 1)")
ADD_STRING_PARAMETER(pl, phenoName, "--pheno-name",
"Specify which phenotype column to read by header")
ADD_BOOL_PARAMETER(pl, qtl, "--qtl", "Treat phenotype as quantitative trait")
ADD_STRING_PARAMETER(
pl, multiplePheno, "--multiplePheno",
"Specify aa template file for analyses of more than one phenotype")
ADD_PARAMETER_GROUP(pl, "Specify Genotype")
ADD_STRING_PARAMETER(pl, dosageTag, "--dosage",
"Specify which dosage tag to use. (e.g. EC or DS)")
ADD_PARAMETER_GROUP(pl, "Chromosome X Options")
ADD_STRING_PARAMETER(pl, xLabel, "--xLabel",
"Specify X chromosome label (default: 23|X)")
ADD_STRING_PARAMETER(pl, xParRegion, "--xParRegion",
"Specify PAR region (default: hg19), can be build "
"number e.g. hg38, b37; or specify region, e.g. "
"'60001-2699520,154931044-155260560'")
ADD_PARAMETER_GROUP(pl, "People Filter")
ADD_STRING_PARAMETER(pl, peopleIncludeID, "--peopleIncludeID",
"List IDs of people that will be included in study")
ADD_STRING_PARAMETER(
pl, peopleIncludeFile, "--peopleIncludeFile",
"From given file, set IDs of people that will be included in study")
ADD_STRING_PARAMETER(pl, peopleExcludeID, "--peopleExcludeID",
"List IDs of people that will be included in study")
ADD_STRING_PARAMETER(
pl, peopleExcludeFile, "--peopleExcludeFile",
"From given file, set IDs of people that will be included in study")
ADD_PARAMETER_GROUP(pl, "Site Filter")
ADD_STRING_PARAMETER(
pl, rangeList, "--rangeList",
"Specify some ranges to use, please use chr:begin-end format.")
ADD_STRING_PARAMETER(
pl, rangeFile, "--rangeFile",
"Specify the file containing ranges, please use chr:begin-end format.")
ADD_STRING_PARAMETER(pl, siteFile, "--siteFile",
"Specify the file containing sites to include, please "
"use \"chr pos\" format.")
ADD_INT_PARAMETER(
pl, siteDepthMin, "--siteDepthMin",
"Specify minimum depth(inclusive) to be included in analysis")
ADD_INT_PARAMETER(
pl, siteDepthMax, "--siteDepthMax",
"Specify maximum depth(inclusive) to be included in analysis")
ADD_INT_PARAMETER(pl, siteMACMin, "--siteMACMin",
"Specify minimum Minor Allele Count(inclusive) to be "
"included in analysis")
ADD_STRING_PARAMETER(pl, annoType, "--annoType",
"Specify annotation type that is followed by ANNO= in "
"the VCF INFO field, regular expression is allowed ")
ADD_PARAMETER_GROUP(pl, "Genotype Filter")
ADD_INT_PARAMETER(
pl, indvDepthMin, "--indvDepthMin",
"Specify minimum depth(inclusive) of a sample to be included in analysis")
ADD_INT_PARAMETER(
pl, indvDepthMax, "--indvDepthMax",
"Specify maximum depth(inclusive) of a sample to be included in analysis")
ADD_INT_PARAMETER(
pl, indvQualMin, "--indvQualMin",
"Specify minimum depth(inclusive) of a sample to be included in analysis")
ADD_PARAMETER_GROUP(pl, "Association Model")
ADD_STRING_PARAMETER(pl, modelSingle, "--single",
"Single variant tests, choose from: score, wald, exact, "
"famScore, famLrt, famGrammarGamma, firth")
ADD_STRING_PARAMETER(pl, modelBurden, "--burden",
"Burden tests, choose from: cmc, zeggini, mb, exactCMC, "
"rarecover, cmat, cmcWald")
ADD_STRING_PARAMETER(pl, modelVT, "--vt",
"Variable threshold tests, choose from: price, analytic")
ADD_STRING_PARAMETER(
pl, modelKernel, "--kernel",
"Kernal-based tests, choose from: SKAT, KBAC, FamSKAT, SKATO")
ADD_STRING_PARAMETER(pl, modelMeta, "--meta",
"Meta-analysis related functions to generate summary "
"statistics, choose from: score, cov, dominant, "
"recessive")
ADD_PARAMETER_GROUP(pl, "Family-based Models")
ADD_STRING_PARAMETER(pl, kinship, "--kinship",
"Specify a kinship file for autosomal analysis, use "
"vcf2kinship to generate")
ADD_STRING_PARAMETER(pl, xHemiKinship, "--xHemiKinship",
"Provide kinship for the chromosome X hemizygote region")
ADD_STRING_PARAMETER(pl, kinshipEigen, "--kinshipEigen",
"Specify eigen decomposition results of a kinship file "
"for autosomal analysis")
ADD_STRING_PARAMETER(
pl, xHemiKinshipEigen, "--xHemiKinshipEigen",
"Specify eigen decomposition results of a kinship file for X analysis")
ADD_PARAMETER_GROUP(pl, "Grouping Unit ")
ADD_STRING_PARAMETER(pl, geneFile, "--geneFile",
"Specify a gene file (for burden tests)")
ADD_STRING_PARAMETER(pl, gene, "--gene", "Specify which genes to test")
ADD_STRING_PARAMETER(pl, setList, "--setList",
"Specify a list to test (for burden tests)")
ADD_STRING_PARAMETER(pl, setFile, "--setFile",
"Specify a list file (for burden tests, first 2 "
"columns: setName chr:beg-end)")
ADD_STRING_PARAMETER(pl, set, "--set",
"Specify which set to test (1st column)")
ADD_PARAMETER_GROUP(pl, "Frequency Cutoff")
/*ADD_BOOL_PARAMETER(pl, freqFromFile, "--freqFromFile", "Obtain frequency
* from external file")*/
// ADD_BOOL_PARAMETER(pl, freqFromControl, "--freqFromControl", "Calculate
// frequency from case samples")
ADD_DOUBLE_PARAMETER(
pl, freqUpper, "--freqUpper",
"Specify upper minor allele frequency bound to be included in analysis")
ADD_DOUBLE_PARAMETER(
pl, freqLower, "--freqLower",
"Specify lower minor allele frequency bound to be included in analysis")
ADD_PARAMETER_GROUP(pl, "Missing Data")
ADD_STRING_PARAMETER(
pl, impute, "--impute",
"Impute missing genotype (default:mean): mean, hwe, and drop")
ADD_BOOL_PARAMETER(
pl, imputePheno, "--imputePheno",
"Impute phenotype to mean of those have genotypes but no phenotypes")
ADD_BOOL_PARAMETER(pl, imputeCov, "--imputeCov",
"Impute each covariate to its mean, instead of drop "
"samples with missing covariates")
ADD_PARAMETER_GROUP(pl, "Conditional Analysis")
ADD_STRING_PARAMETER(pl, condition, "--condition",
"Specify markers to be conditions (specify range)")
ADD_PARAMETER_GROUP(pl, "Auxiliary Functions")
ADD_BOOL_PARAMETER(pl, noweb, "--noweb", "Skip checking new version")
ADD_BOOL_PARAMETER(pl, help, "--help", "Print detailed help message")
END_PARAMETER_LIST(pl);
pl.Read(argc, argv);
if (FLAG_help) {
pl.Help();
return 0;
}
welcome();
pl.Status();
if (FLAG_REMAIN_ARG.size() > 0) {
fprintf(stderr, "Unparsed arguments: ");
for (unsigned int i = 0; i < FLAG_REMAIN_ARG.size(); i++) {
fprintf(stderr, " %s", FLAG_REMAIN_ARG[i].c_str());
}
exit(1);
}
if (!FLAG_outPrefix.size()) FLAG_outPrefix = "rvtest";
REQUIRE_STRING_PARAMETER(FLAG_inVcf,
"Please provide input file using: --inVcf");
// check new version
if (!FLAG_noweb) {
VersionChecker ver;
if (ver.retrieveRemoteVersion("http://zhanxw.com/rvtests/version") < 0) {
fprintf(stderr,
"Retrieve remote version failed, use '--noweb' to skip.\n");
} else {
ver.setLocalVersion(VERSION);
if (ver.isRemoteVersionNewer()) {
fprintf(stderr, "New version of rvtests is available:");
ver.printRemoteContent();
}
}
}
// start logging
Logger _logger((FLAG_outPrefix + ".log").c_str());
logger = &_logger;
logger->info("Program version: %s", VERSION);
logger->infoToFile("Git Version: %s", GIT_VERSION);
logger->infoToFile("Parameters BEGIN");
pl.WriteToFile(logger->getHandle());
logger->infoToFile("Parameters END");
logger->sync();
// start analysis
time_t startTime = time(0);
logger->info("Analysis started at: %s", currentTime().c_str());
GenotypeExtractor ge(FLAG_inVcf);
// set range filters here
ge.setRangeList(FLAG_rangeList.c_str());
ge.setRangeFile(FLAG_rangeFile.c_str());
// set people filters here
if (FLAG_peopleIncludeID.size() || FLAG_peopleIncludeFile.size()) {
ge.excludeAllPeople();
ge.includePeople(FLAG_peopleIncludeID.c_str());
ge.includePeopleFromFile(FLAG_peopleIncludeFile.c_str());
}
ge.excludePeople(FLAG_peopleExcludeID.c_str());
ge.excludePeopleFromFile(FLAG_peopleExcludeFile.c_str());
if (FLAG_siteDepthMin > 0) {
ge.setSiteDepthMin(FLAG_siteDepthMin);
logger->info("Set site depth minimum to %d", FLAG_siteDepthMin);
}
if (FLAG_siteDepthMax > 0) {
ge.setSiteDepthMax(FLAG_siteDepthMax);
logger->info("Set site depth maximum to %d", FLAG_siteDepthMax);
}
if (FLAG_siteMACMin > 0) {
ge.setSiteMACMin(FLAG_siteMACMin);
logger->info("Set site minimum MAC to %d", FLAG_siteDepthMin);
}
if (FLAG_annoType != "") {
ge.setAnnoType(FLAG_annoType.c_str());
logger->info("Set annotype type filter to %s", FLAG_annoType.c_str());
}
std::vector<std::string> vcfSampleNames;
ge.getPeopleName(&vcfSampleNames);
logger->info("Loaded [ %zu ] samples from VCF files", vcfSampleNames.size());
DataLoader dataLoader;
dataLoader.setPhenotypeImputation(FLAG_imputePheno);
dataLoader.setCovariateImputation(FLAG_imputeCov);
if (FLAG_multiplePheno.empty()) {
dataLoader.loadPhenotype(FLAG_pheno, FLAG_mpheno, FLAG_phenoName);
// // load phenotypes
// std::map<std::string, double> phenotype;
// if (FLAG_pheno.empty()) {
// logger->error("Cannot do association when phenotype is missing!");
// return -1;
// }
// // check if alternative phenotype columns are used
// if (!FLAG_mpheno.empty() && !FLAG_phenoName.empty()) {
// logger->error("Please specify either --mpheno or --pheno-name");
// return -1;
// }
// if (!FLAG_mpheno.empty()) {
// int col = atoi(FLAG_mpheno);
// int ret = loadPedPhenotypeByColumn(FLAG_pheno.c_str(), &phenotype,
// col);
// if (ret < 0) {
// logger->error("Loading phenotype failed!");
// return -1;
// }
// } else if (!FLAG_phenoName.empty()) {
// int ret = loadPedPhenotypeByHeader(FLAG_pheno.c_str(), &phenotype,
// FLAG_phenoName.c_str());
// if (ret < 0) {
// logger->error("Loading phenotype failed!");
// return -1;
// }
// } else {
// int col = 1; // default use the first phenotype
// int ret = loadPedPhenotypeByColumn(FLAG_pheno.c_str(), &phenotype,
// col);
// if (ret < 0) {
// logger->error("Loading phenotype failed!");
// return -1;
// }
// }
// logger->info("Loaded [ %zu ] sample pheontypes.", phenotype.size());
// rearrange phenotypes
// drop samples from phenotype or vcf
matchPhenotypeAndVCF("missing phenotype", &dataLoader, &ge);
// // phenotype names (vcf sample names) arranged in the same order as in
// VCF
// std::vector<std::string> phenotypeNameInOrder;
// std::vector<double>
// phenotypeInOrder; // phenotype arranged in the same order as in VCF
// rearrange(phenotype, vcfSampleNames, &vcfSampleToDrop,
// &phenotypeNameInOrder,
// &phenotypeInOrder, FLAG_imputePheno);
// if (vcfSampleToDrop.size()) {
// // exclude this sample from parsing VCF
// ge.excludePeople(vcfSampleToDrop);
// // output dropped samples
// for (size_t i = 0; i < vcfSampleToDrop.size(); ++i) {
// if (i == 0)
// logger->warn(
// "Total [ %zu ] samples are dropped from VCF file due to missing
// "
// "phenotype",
// vcfSampleToDrop.size());
// if (i >= 10) {
// logger->warn(
// "Skip outputting additional [ %d ] samples with missing "
// "phenotypes.",
// ((int)vcfSampleToDrop.size() - 10));
// break;
// }
// logger->warn("Drop sample [ %s ] from VCF file due to missing
// phenotype",
// (vcfSampleToDrop)[i].c_str());
// }
// // logger->warn("Drop %zu sample from VCF file since we don't have
// their
// // phenotypes", vcfSampleToDrop.size());
// }
// if (phenotypeInOrder.size() != phenotype.size()) {
// logger->warn(
// "Drop [ %d ] samples from phenotype file due to missing genotypes
// from "
// "VCF files",
// (int)(phenotype.size() - phenotypeInOrder.size()));
// // We may output these samples by comparing keys of phenotype and
// // phenotypeNameInOrder
// }
dataLoader.loadCovariate(FLAG_cov, FLAG_covName);
matchCovariateAndVCF("missing covariate", &dataLoader, &ge);
// // load covariate
// Matrix covariate;
// HandleMissingCov handleMissingCov = COVARIATE_DROP;
// if (FLAG_imputeCov) {
// handleMissingCov = COVARIATE_IMPUTE;
// }
// if (FLAG_cov.empty() && !FLAG_covName.empty()) {
// logger->info("Use phenotype file as covariate file [ %s ]",
// FLAG_pheno.c_str());
// FLAG_cov = FLAG_pheno;
// }
// if (!FLAG_cov.empty()) {
// logger->info("Begin to read covariate file.");
// std::vector<std::string> columnNamesInCovariate;
// std::set<std::string> sampleToDropInCovariate;
// int ret = loadCovariate(FLAG_cov.c_str(), phenotypeNameInOrder,
// FLAG_covName.c_str(), handleMissingCov,
// &covariate,
// &columnNamesInCovariate,
// &sampleToDropInCovariate);
// if (ret < 0) {
// logger->error("Load covariate file failed !");
// exit(1);
// }
// // drop phenotype samples
// if (!sampleToDropInCovariate.empty()) {
// int idx = 0;
// int n = phenotypeNameInOrder.size();
// for (int i = 0; i < n; ++i) {
// if (sampleToDropInCovariate.count(phenotypeNameInOrder[i]) !=
// 0) { // need to drop
// continue;
// }
// phenotypeNameInOrder[idx] = phenotypeNameInOrder[i];
// phenotypeInOrder[idx] = phenotypeInOrder[i];
// idx++;
// }
// phenotypeNameInOrder.resize(idx);
// phenotypeInOrder.resize(idx);
// logger->warn(
// "[ %zu ] sample phenotypes are dropped due to lacking
// covariates.",
// sampleToDropInCovariate.size());
// }
// // drop vcf samples;
// for (std::set<std::string>::const_iterator iter =
// sampleToDropInCovariate.begin();
// iter != sampleToDropInCovariate.end(); ++iter) {
// ge.excludePeople(iter->c_str());
// }
// }
} else {
dataLoader.loadMultiplePhenotype(FLAG_multiplePheno, FLAG_pheno, FLAG_cov);
matchPhenotypeAndVCF("missing phenotype", &dataLoader, &ge);
matchCovariateAndVCF("missing covariate", &dataLoader, &ge);
}
dataLoader.loadSex();
if (FLAG_sex) {
dataLoader.useSexAsCovariate();
matchCovariateAndVCF("missing sex", &dataLoader, &ge);
}
// // load sex
// std::vector<int> sex;
// if (loadSex(FLAG_pheno, phenotypeNameInOrder, &sex)) {
// logger->error("Cannot load sex of samples from phenotype file");
// exit(1);
// }
// if (FLAG_sex) { // append sex in covariate
// std::vector<int> index; // mark missing samples
// int numMissing = findMissingSex(sex, &index);
// logger->info("Futher exclude %d samples with missing sex", numMissing);
// removeByIndex(index, &sex);
// excludeSamplesByIndex(index, &ge, &phenotypeNameInOrder,
// &phenotypeInOrder,
// &covariate);
// appendToMatrix("Sex", sex, &covariate);
// }
if (!FLAG_condition.empty()) {
dataLoader.loadMarkerAsCovariate(FLAG_inVcf, FLAG_condition);
matchCovariateAndVCF("missing in conditioned marker(s)", &dataLoader, &ge);
}
// // load conditional markers
// if (!FLAG_condition.empty()) {
// Matrix geno;
// std::vector<std::string> rowLabel;
// if (loadMarkerFromVCF(FLAG_inVcf, FLAG_condition, &rowLabel, &geno) < 0)
// {
// logger->error("Load conditional markers [ %s ] from [ %s ] failed.",
// FLAG_condition.c_str(), FLAG_inVcf.c_str());
// exit(1);
// }
// if (appendGenotype(&covariate, phenotypeNameInOrder, geno, rowLabel) < 0)
// {
// logger->error(
// "Failed to combine conditional markers [ %s ] from [ %s ] failed.",
// FLAG_condition.c_str(), FLAG_inVcf.c_str());
// exit(1);
// }
// }
dataLoader.checkConstantCovariate();
// // check if some covariates are constant for all samples
// // e.g. user may include covariate "1" in addition to intercept
// // in such case, we will give a fatal error
// for (int i = 0; i < covariate.cols; ++i) {
// std::set<double> s;
// s.clear();
// for (int j = 0; j < covariate.rows; ++j) {
// s.insert(covariate[j][i]);
// }
// if (s.size() == 1) {
// logger->error(
// "Covariate [ %s ] equals [ %g ] for all samples, cannot fit "
// "model...\n",
// covariate.GetColumnLabel(i), *s.begin());
// exit(1);
// }
// }
g_SummaryHeader = new SummaryHeader;
g_SummaryHeader->recordCovariate(dataLoader.getCovariate());
// record raw phenotype
g_SummaryHeader->recordPhenotype("Trait",
dataLoader.getPhenotype().extractCol(0));
// adjust phenotype
// bool binaryPhenotype;
if (FLAG_qtl) {
// binaryPhenotype = false;
dataLoader.setTraitType(DataLoader::PHENOTYPE_QTL);
logger->info("-- Force quantitative trait mode -- ");
} else {
if (dataLoader.detectPhenotypeType() == DataLoader::PHENOTYPE_BINARY) {
logger->warn("-- Enabling binary phenotype mode -- ");
dataLoader.setTraitType(DataLoader::PHENOTYPE_BINARY);
} else {
dataLoader.setTraitType(DataLoader::PHENOTYPE_QTL);
}
// binaryPhenotype = isBinaryPhenotype(phenotypeInOrder);
// if (binaryPhenotype) {
// logger->warn("-- Enabling binary phenotype mode -- ");
// convertBinaryPhenotype(&phenotypeInOrder);
// }
}
if (FLAG_useResidualAsPhenotype) {
dataLoader.useResidualAsPhenotype();
g_SummaryHeader->recordEstimation(dataLoader.getEstimation());
}
// // use residual as phenotype
// if (FLAG_useResidualAsPhenotype) {
// if (binaryPhenotype) {
// logger->warn(
// "WARNING: Skip transforming binary phenotype, although you want to
// "
// "use residual as phenotype!");
// } else {
// if (covariate.cols > 0) {
// LinearRegression lr;
// Vector pheno;
// Matrix covAndInt;
// copy(phenotypeInOrder, &pheno);
// copyCovariateAndIntercept(covariate.rows, covariate, &covAndInt);
// if (!lr.FitLinearModel(covAndInt, pheno)) {
// logger->error(
// "Cannot fit model: [ phenotype ~ 1 + covariates ], now use the
// "
// "original phenotype");
// } else {
// const int n = lr.GetResiduals().Length();
// for (int i = 0; i < n; ++i) {
// phenotypeInOrder[i] = lr.GetResiduals()[i];
// }
// covariate.Dimension(0, 0);
// logger->info(
// "DONE: Fit model [ phenotype ~ 1 + covariates ] and model "
// "residuals will be used as responses.");
// }
// } else { // no covaraites
// centerVector(&phenotypeInOrder);
// logger->info("DONE: Use residual as phenotype by centerng it");
// }
// }
// }
if (FLAG_inverseNormal) {
dataLoader.inverseNormalizePhenotype();
g_SummaryHeader->setInverseNormalize(FLAG_inverseNormal);
}
// // phenotype transformation
// if (FLAG_inverseNormal) {
// if (binaryPhenotype) {
// logger->warn(
// "WARNING: Skip transforming binary phenotype, although you required
// "
// "inverse normalization!");
// } else {
// logger->info("Now applying inverse normalize transformation.");
// inverseNormalizeLikeMerlin(&phenotypeInOrder);
// g_SummaryHeader->setInverseNormalize(FLAG_inverseNormal);
// logger->info("DONE: inverse normal transformation finished.");
// }
// }
g_SummaryHeader->recordPhenotype("AnalyzedTrait",
dataLoader.getPhenotype().extractCol(0));
if (dataLoader.getPhenotype().nrow() == 0) {
logger->fatal("There are 0 samples with valid phenotypes, quitting...");
exit(1);
}
// if (phenotypeInOrder.empty()) {
// logger->fatal("There are 0 samples with valid phenotypes, quitting...");
// exit(1);
// }
logger->info("Analysis begins with [ %d ] samples...",
dataLoader.getPhenotype().nrow());
//////////////////////////////////////////////////////////////////////////////
// prepare each model
bool singleVariantMode = FLAG_modelSingle.size() || FLAG_modelMeta.size();
bool groupVariantMode = (FLAG_modelBurden.size() || FLAG_modelVT.size() ||
FLAG_modelKernel.size());
if (singleVariantMode && groupVariantMode) {
logger->error("Cannot support both single variant and region based tests");
exit(1);
}
ModelManager modelManager(FLAG_outPrefix);
// set up models in qtl/binary modes
if (dataLoader.isBinaryPhenotype()) {
modelManager.setBinaryOutcome();
matchPhenotypeAndVCF("missing phenotype (not case/control)", &dataLoader,
&ge);
} else {
modelManager.setQuantitativeOutcome();
}
// create models
modelManager.create("single", FLAG_modelSingle);
modelManager.create("burden", FLAG_modelBurden);
modelManager.create("vt", FLAG_modelVT);
modelManager.create("kernel", FLAG_modelKernel);
modelManager.create("meta", FLAG_modelMeta);
if (FLAG_outputRaw) {
modelManager.create("outputRaw", "dump");
}
const std::vector<ModelFitter*>& model = modelManager.getModel();
const std::vector<FileWriter*>& fOuts = modelManager.getResultFile();
const size_t numModel = model.size();
// TODO: optimize this by avoidding data copying
Matrix phenotypeMatrix;
Matrix covariate;
toMatrix(dataLoader.getPhenotype(), &phenotypeMatrix);
toMatrix(dataLoader.getCovariate(), &covariate);
// determine VCF file reading pattern
// current support:
// * line by line ( including range selection)
// * gene by gene
// * range by range
std::string rangeMode = "Single";
if (FLAG_geneFile.size() && (FLAG_setFile.size() || FLAG_setList.size())) {
logger->error("Cannot specify both gene file and set file.");
exit(1);
}
if (!FLAG_gene.empty() && FLAG_geneFile.empty()) {
logger->error("Please provide gene file for gene bases analysis.");
exit(1);
}
OrderedMap<std::string, RangeList> geneRange;
if (FLAG_geneFile.size()) {
rangeMode = "Gene";
int ret =
loadGeneFile(FLAG_geneFile.c_str(), FLAG_gene.c_str(), &geneRange);
if (ret < 0 || geneRange.size() == 0) {
logger->error("Error loading gene file or gene list is empty!");
return -1;
} else {
logger->info("Loaded [ %zu ] genes.", geneRange.size());
}
}
if (!FLAG_set.empty() && FLAG_setFile.empty()) {
logger->error("Please provide set file for set bases analysis.");
exit(1);
}
if (FLAG_setFile.size()) {
rangeMode = "Range";
int ret = loadRangeFile(FLAG_setFile.c_str(), FLAG_set.c_str(), &geneRange);
if (ret < 0 || geneRange.size() == 0) {
logger->error("Error loading set file or set list is empty!");
return -1;
} else {
logger->info("Loaded [ %zu ] set to tests.", geneRange.size());
}
}
if (FLAG_setList.size()) {
rangeMode = "Range";
int ret = appendListToRange(FLAG_setList, &geneRange);
if (ret < 0) {
logger->error("Error loading set list or set list is empty!");
return -1;
}
}
DataConsolidator dc;
dc.setSex(&dataLoader.getSex());
dc.setFormula(&dataLoader.getFormula());
dc.setGenotypeCounter(ge.getGenotypeCounter());
// load kinshp if needed by family models
if (modelManager.hasFamilyModel() ||
(!FLAG_modelMeta.empty() && !FLAG_kinship.empty())) {
logger->info("Family-based model specified. Loading kinship file...");
// process auto kinship
if (dc.setKinshipSample(dataLoader.getPhenotype().getRowName()) ||
dc.setKinshipFile(DataConsolidator::KINSHIP_AUTO, FLAG_kinship) ||
dc.setKinshipEigenFile(DataConsolidator::KINSHIP_AUTO,
FLAG_kinshipEigen) ||
dc.loadKinship(DataConsolidator::KINSHIP_AUTO)) {
logger->error(
"Failed to load autosomal kinship (you may use vcf2kinship to "
"generate one).");
exit(1);
}
if (dc.setKinshipFile(DataConsolidator::KINSHIP_X, FLAG_xHemiKinship) ||
dc.setKinshipEigenFile(DataConsolidator::KINSHIP_X,
FLAG_xHemiKinshipEigen) ||
dc.loadKinship(DataConsolidator::KINSHIP_X)) {
logger->warn(
"Autosomal kinship loaded, but no hemizygote region kinship "
"provided, some sex chromosome tests will be skipped.");
// keep the program going
}
} else if (!FLAG_kinship.empty() && FLAG_modelMeta.empty()) {
logger->info(
"Family-based model not specified. Options related to kinship will be "
"ignored here.");
}
// set imputation method
if (FLAG_impute.empty()) {
logger->info("Impute missing genotype to mean (by default)");
dc.setStrategy(DataConsolidator::IMPUTE_MEAN);
} else if (FLAG_impute == "mean") {
logger->info("Impute missing genotype to mean");
dc.setStrategy(DataConsolidator::IMPUTE_MEAN);
} else if (FLAG_impute == "hwe") {
logger->info("Impute missing genotype by HWE");
dc.setStrategy(DataConsolidator::IMPUTE_HWE);
} else if (FLAG_impute == "drop") {
logger->info("Drop missing genotypes");
dc.setStrategy(DataConsolidator::DROP);
}
dc.setPhenotypeName(dataLoader.getPhenotype().getRowName());
// set up par region
ParRegion parRegion(FLAG_xLabel, FLAG_xParRegion);
dc.setParRegion(&parRegion);
// genotype will be extracted and stored
Matrix& genotype = dc.getOriginalGenotype();
if (FLAG_freqUpper > 0) {
ge.setSiteFreqMax(FLAG_freqUpper);
logger->info("Set upper minor allele frequency limit to %g",
FLAG_freqUpper);
}
if (FLAG_freqLower > 0) {
ge.setSiteFreqMin(FLAG_freqLower);
logger->info("Set lower minor allele frequency limit to %g",
FLAG_freqLower);
}
// handle sex chromosome
ge.setParRegion(&parRegion);
ge.setSex(&dataLoader.getSex());
// use dosage instead GT
if (!FLAG_dosageTag.empty()) {
ge.setDosageTag(FLAG_dosageTag);
logger->info("Use dosage genotype from VCF flag %s.",
FLAG_dosageTag.c_str());
}
// genotype QC options
if (FLAG_indvDepthMin > 0) {
ge.setGDmin(FLAG_indvDepthMin);
logger->info("Minimum GD set to %d (or marked as missing genotype).",
FLAG_indvDepthMin);
}
if (FLAG_indvDepthMax > 0) {
ge.setGDmax(FLAG_indvDepthMax);
logger->info("Maximum GD set to %d (or marked as missing genotype).",
FLAG_indvDepthMax);
}
if (FLAG_indvQualMin > 0) {
ge.setGQmin(FLAG_indvQualMin);
logger->info("Minimum GQ set to %d (or marked as missing genotype).",
FLAG_indvQualMin);
}
dc.preRegressionCheck(phenotypeMatrix, covariate);
logger->info("Analysis started");
Result& buf = dc.getResult();
// we have three modes:
// * single variant reading, single variant test
// * range variant reading, single variant test
// * range variant reading, group variant test
if (rangeMode == "Single" && singleVariantMode) { // use line by line mode
buf.addHeader("CHROM");
buf.addHeader("POS");
buf.addHeader("REF");
buf.addHeader("ALT");
buf.addHeader("N_INFORMATIVE");
// output headers
for (size_t m = 0; m < model.size(); m++) {
model[m]->writeHeader(fOuts[m], buf);
}
int variantProcessed = 0;
while (true) {
buf.clearValue();
int ret = ge.extractSingleGenotype(&genotype, &buf);
if (ret == GenotypeExtractor::FILE_END) { // reach file end
break;
}
if (ret == GenotypeExtractor::FAIL_FILTER) {
continue;
}
if (ret != GenotypeExtractor::SUCCEED) {
logger->error("Extract genotype failed at site: %s:%s!",
buf["CHROM"].c_str(), buf["POS"].c_str());
continue;
}
if (genotype.cols == 0) {
logger->warn("Extract [ %s:%s ] has 0 variants, skipping",
buf["CHROM"].c_str(), buf["POS"].c_str());
continue;
}
++variantProcessed;
dc.consolidate(phenotypeMatrix, covariate, genotype);
buf.updateValue("N_INFORMATIVE", toString(genotype.rows));
// fit each model
for (size_t m = 0; m != numModel; m++) {
model[m]->reset();
model[m]->fit(&dc);
model[m]->writeOutput(fOuts[m], buf);
}
}
logger->info("Analyzed [ %d ] variants", variantProcessed);
} else if (rangeMode != "Single" &&
singleVariantMode) { // read by gene/range model, single variant
// test
buf.addHeader(rangeMode);
buf.addHeader("CHROM");
buf.addHeader("POS");
buf.addHeader("REF");
buf.addHeader("ALT");
buf.addHeader("N_INFORMATIVE");
// output headers
for (size_t m = 0; m < numModel; m++) {
model[m]->writeHeader(fOuts[m], buf);
}
std::string geneName;
RangeList rangeList;
int variantProcessed = 0;
for (size_t i = 0; i < geneRange.size(); ++i) {
geneRange.at(i, &geneName, &rangeList);
ge.setRange(rangeList);
while (true) {
buf.clearValue();
int ret = ge.extractSingleGenotype(&genotype, &buf);
if (ret == GenotypeExtractor::FILE_END) { // reach end of this region
break;
}
if (ret == GenotypeExtractor::FAIL_FILTER) {
continue;
}
if (ret != GenotypeExtractor::SUCCEED) {
logger->error("Extract genotype failed for gene %s!",
geneName.c_str());
continue;
}
if (genotype.cols == 0) {
logger->warn("Gene %s has 0 variants, skipping", geneName.c_str());
continue;
}
++variantProcessed;
dc.consolidate(phenotypeMatrix, covariate, genotype);
buf.updateValue(rangeMode, geneName);
buf.updateValue("N_INFORMATIVE", genotype.rows);
// #pragma omp parallel for
for (size_t m = 0; m != numModel; m++) {
model[m]->reset();
model[m]->fit(&dc);
model[m]->writeOutput(fOuts[m], buf);
}
}
}
logger->info("Analyzed [ %d ] variants from [ %d ] genes/regions",
variantProcessed, (int)geneRange.size());
} else if (rangeMode != "Single" &&
groupVariantMode) { // read by gene/range mode, group variant
// test
buf.addHeader(rangeMode);
buf.addHeader("RANGE");
buf.addHeader("N_INFORMATIVE");
buf.addHeader("NumVar");
buf.addHeader("NumPolyVar");
// output headers
for (size_t m = 0; m < numModel; m++) {
model[m]->writeHeader(fOuts[m], buf);
}
std::string geneName;
RangeList rangeList;
int variantProcessed = 0;
ge.enableAutoMerge();
for (size_t i = 0; i < geneRange.size(); ++i) {
geneRange.at(i, &geneName, &rangeList);
ge.setRange(rangeList);
buf.clearValue();
int ret = ge.extractMultipleGenotype(&genotype);
if (ret != GenotypeExtractor::SUCCEED) {
logger->error("Extract genotype failed for gene %s!", geneName.c_str());
continue;
}
if (genotype.cols == 0) {
logger->info("Gene %s has 0 variants, skipping", geneName.c_str());
continue;
}
variantProcessed += genotype.cols; // genotype is people by marker
dc.consolidate(phenotypeMatrix, covariate, genotype);
buf.updateValue(rangeMode, geneName);
buf.updateValue("RANGE", rangeList.toString());
buf.updateValue("N_INFORMATIVE", genotype.rows);
buf.updateValue("NumVar", genotype.cols);
buf.updateValue("NumPolyVar",
dc.getFlippedToMinorPolymorphicGenotype().cols);
// #ifdef _OPENMP
// #pragma omp parallel for
// #endif
for (size_t m = 0; m != numModel; m++) {
model[m]->reset();
model[m]->fit(&dc);
model[m]->writeOutput(fOuts[m], buf);
}
}
logger->info("Analyzed [ %d ] variants from [ %d ] genes/regions",
variantProcessed, (int)geneRange.size());
} else {
logger->error(
"Unsupported reading mode and test modes! (need more parameters?)");
exit(1);
}
// Resource cleaning up
modelManager.close();
delete g_SummaryHeader;
time_t endTime = time(0);
logger->info("Analysis ends at: %s", currentTime().c_str());
int elapsedSecond = (int)(endTime - startTime);
logger->info("Analysis took %d seconds", elapsedSecond);
return 0;
}