4c23ed8c2c9becf91f0c521dac0983c21ca0e20b angie Mon Jan 6 14:35:20 2014 -0800 Tolerate missing '=' in metadata line for VCF3 because as Jonathan pointed out, the spec didn't demand key=val in 3.3.refs #12412 diff --git src/lib/vcf.c src/lib/vcf.c index b2eaed7..8b43371 100644 --- src/lib/vcf.c +++ src/lib/vcf.c @@ -1,1166 +1,1167 @@ /* VCF: Variant Call Format, version 4.0 / 4.1 * http://www.1000genomes.org/wiki/Analysis/Variant%20Call%20Format/vcf-variant-call-format-version-40 * http://www.1000genomes.org/wiki/Analysis/Variant%20Call%20Format/vcf-variant-call-format-version-41 */ #include "common.h" #include "dnautil.h" #include "errabort.h" #include <limits.h> #include "localmem.h" #include "net.h" #include "regexHelper.h" #include "vcf.h" /* Reserved but optional INFO keys: */ const char *vcfInfoAncestralAllele = "AA"; const char *vcfInfoPerAlleleGtCount = "AC"; // allele count in genotypes, for each ALT allele, // in the same order as listed const char *vcfInfoAlleleFrequency = "AF"; // allele frequency for each ALT allele in the same // order as listed: use this when estimated from // primary data, not called genotypes const char *vcfInfoNumAlleles = "AN"; // total number of alleles in called genotypes const char *vcfInfoBaseQuality = "BQ"; // RMS base quality at this position const char *vcfInfoCigar = "CIGAR"; // cigar string describing how to align an // alternate allele to the reference allele const char *vcfInfoIsDbSnp = "DB"; // dbSNP membership const char *vcfInfoDepth = "DP"; // combined depth across samples, e.g. DP=154 const char *vcfInfoEnd = "END"; // end position of the variant described in this // record (esp. for CNVs) const char *vcfInfoIsHapMap2 = "H2"; // membership in hapmap2 const char *vcfInfoIsHapMap3 = "H3"; // membership in hapmap3 const char *vcfInfoIs1000Genomes = "1000G"; // membership in 1000 Genomes const char *vcfInfoMappingQuality = "MQ"; // RMS mapping quality, e.g. MQ=52 const char *vcfInfoMapQual0Count = "MQ0"; // number of MAPQ == 0 reads covering this record const char *vcfInfoNumSamples = "NS"; // Number of samples with data const char *vcfInfoStrandBias = "SB"; // strand bias at this position const char *vcfInfoIsSomatic = "SOMATIC"; // indicates that the record is a somatic mutation, // for cancer genomics const char *vcfInfoIsValidated = "VALIDATED"; // validated by follow-up experiment /* Reserved but optional per-genotype keys: */ const char *vcfGtGenotype = "GT"; // Integer allele indices separated by "/" (unphased) // or "|" (phased). Allele values are 0 for // reference allele, 1 for the first allele in ALT, // 2 for the second allele in ALT and so on. const char *vcfGtDepth = "DP"; // Read depth at this position for this sample const char *vcfGtFilter = "FT"; // Analogous to variant's FILTER field const char *vcfGtLikelihoods = "GL"; // Three floating point log10-scaled likelihoods for // AA,AB,BB genotypes where A=ref and B=alt; // not applicable if site is not biallelic. const char *vcfGtPhred = "PL"; // Phred-scaled genotype likelihoods rounded to closest int const char *vcfGtConditionalQual = "GQ";// Conditional genotype quality // i.e. phred quality -10log_10 P(genotype call is wrong, // conditioned on the site's being variant) const char *vcfGtHaplotypeQualities = "HQ"; // two phred qualities comma separated const char *vcfGtPhaseSet = "PS"; // Set of phased genotypes to which this genotype belongs const char *vcfGtPhasingQuality = "PQ"; // Phred-scaled P(alleles ordered wrongly in heterozygote) const char *vcfGtExpectedAltAlleleCount = "EC"; // Typically used in association analyses // Make definitions of reserved INFO and genotype keys, in case people take them for // granted and don't make ##INFO headers for them: static struct vcfInfoDef *vcfSpecInfoDefs = NULL; static struct vcfInfoDef *vcfSpecGtFormatDefs = NULL; static void addInfoDef(struct vcfInfoDef **pList, const char *key, int fieldCount, enum vcfInfoType type, char *description) /* Allocate and initialize an info def and add it to pList. */ { struct vcfInfoDef *def; AllocVar(def); def->key = (char *)key; def->fieldCount = fieldCount; def->type = type; def->description = description; slAddHead(pList, def); } static void initVcfSpecInfoDefs() /* Make linked list of INFO defs reserved in the spec. */ { if (vcfSpecInfoDefs != NULL) return; addInfoDef(&vcfSpecInfoDefs, vcfInfoAncestralAllele, 1, vcfInfoString, "Ancestral allele"); addInfoDef(&vcfSpecInfoDefs, vcfInfoPerAlleleGtCount, 1, vcfInfoInteger, "Allele count in genotypes, for each ALT allele, in the same order as listed"); addInfoDef(&vcfSpecInfoDefs, vcfInfoAlleleFrequency, -1, vcfInfoFloat, "Allele frequency for each ALT allele in the same order as listed: " "use this when estimated from primary data, not called genotypes"); addInfoDef(&vcfSpecInfoDefs, vcfInfoNumAlleles, 1, vcfInfoInteger, "Total number of alleles in called genotypes"); addInfoDef(&vcfSpecInfoDefs, vcfInfoBaseQuality, 1, vcfInfoFloat, "RMS base quality at this position"); addInfoDef(&vcfSpecInfoDefs, vcfInfoCigar, 1, vcfInfoString, "CIGAR string describing how to align an alternate allele to the reference allele"); addInfoDef(&vcfSpecInfoDefs, vcfInfoIsDbSnp, 0, vcfInfoFlag, "dbSNP membership"); addInfoDef(&vcfSpecInfoDefs, vcfInfoDepth, 1, vcfInfoString, "Combined depth across samples, e.g. DP=154"); addInfoDef(&vcfSpecInfoDefs, vcfInfoEnd, 1, vcfInfoInteger, "End position of the variant described in this record " "(especially for structural variants)"); addInfoDef(&vcfSpecInfoDefs, vcfInfoIsHapMap2, 1, vcfInfoFlag, "Membership in HapMap 2"); addInfoDef(&vcfSpecInfoDefs, vcfInfoIsHapMap3, 1, vcfInfoFlag, "Membership in HapMap 3"); addInfoDef(&vcfSpecInfoDefs, vcfInfoIs1000Genomes, 1, vcfInfoFlag, "Membership in 1000 Genomes"); addInfoDef(&vcfSpecInfoDefs, vcfInfoMappingQuality, 1, vcfInfoFloat, "RMS mapping quality, e.g. MQ=52"); addInfoDef(&vcfSpecInfoDefs, vcfInfoMapQual0Count, 1, vcfInfoInteger, "Number of MAPQ == 0 reads covering this record"); addInfoDef(&vcfSpecInfoDefs, vcfInfoNumSamples, 1, vcfInfoInteger, "Number of samples with data"); addInfoDef(&vcfSpecInfoDefs, vcfInfoStrandBias, 1, vcfInfoString, "Strand bias at this position"); addInfoDef(&vcfSpecInfoDefs, vcfInfoIsSomatic, 1, vcfInfoFlag, "Indicates that the record is a somatic mutation, for cancer genomics"); addInfoDef(&vcfSpecInfoDefs, vcfInfoIsValidated, 1, vcfInfoFlag, "Validated by follow-up experiment"); } static void initVcfSpecGtFormatDefs() /* Make linked list of genotype info defs reserved in spec. */ { if (vcfSpecGtFormatDefs != NULL) return; addInfoDef(&vcfSpecGtFormatDefs, vcfGtGenotype, 1, vcfInfoString, "Integer allele indices separated by \"/\" (unphased) " "or \"|\" (phased). Allele values are 0 for " "reference allele, 1 for the first allele in ALT, " "2 for the second allele in ALT and so on, or \".\" for unknown"); addInfoDef(&vcfSpecGtFormatDefs, vcfGtDepth, 1, vcfInfoInteger, "Read depth at this position for this sample"); addInfoDef(&vcfSpecGtFormatDefs, vcfGtFilter, 1, vcfInfoString, "PASS to indicate that all filters have been passed, " "a semi-colon separated list of codes for filters that fail, " "or \".\" to indicate that filters have not been applied"); addInfoDef(&vcfSpecGtFormatDefs, vcfGtLikelihoods, -1, vcfInfoFloat, "Genotype likelihoods comprised of comma separated floating point " "log10-scaled likelihoods for all possible genotypes given the set " "of alleles defined in the REF and ALT fields. "); addInfoDef(&vcfSpecGtFormatDefs, vcfGtPhred, -1, vcfInfoInteger, "Phred-scaled genotype likelihoods rounded to the closest integer " "(and otherwise defined precisely as the genotype likelihoods (GL) field)"); addInfoDef(&vcfSpecGtFormatDefs, vcfGtConditionalQual, -1, vcfInfoFloat, "phred-scaled genotype posterior probabilities " "(and otherwise defined precisely as the genotype likelihoods (GL) field)" "; intended to store imputed genotype probabilities"); addInfoDef(&vcfSpecGtFormatDefs, vcfGtHaplotypeQualities, 2, vcfInfoFloat, "Two comma-separated phred-scaled haplotype qualities"); addInfoDef(&vcfSpecGtFormatDefs, vcfGtPhaseSet, 1, vcfInfoFloat, "A set of phased genotypes to which this genotype belongs"); addInfoDef(&vcfSpecGtFormatDefs, vcfGtPhasingQuality, 1, vcfInfoFloat, "Phasing quality, the phred-scaled probability that alleles are ordered " "incorrectly in a heterozygote (against all other members in the phase set)"); addInfoDef(&vcfSpecGtFormatDefs, vcfGtExpectedAltAlleleCount, 1, vcfInfoFloat, "Expected alternate allele count (typically used in association analyses)"); } static bool vcfFileStopDueToErrors(struct vcfFile *vcff) /* determine if we should stop due to the number of errors */ { return vcff->errCnt > vcff->maxErr; } static void vcfFileErr(struct vcfFile *vcff, char *format, ...) #if defined(__GNUC__) __attribute__((format(printf, 2, 3))) #endif ; static void vcfFileErr(struct vcfFile *vcff, char *format, ...) /* Send error message to errabort stack's warn handler and abort */ { vcff->errCnt++; if (vcff->maxErr == VCF_IGNORE_ERRS) return; va_list args; va_start(args, format); char formatPlus[1024]; if (vcff->lf != NULL) sprintf(formatPlus, "%s:%d: %s", vcff->lf->fileName, vcff->lf->lineIx, format); else strcpy(formatPlus, format); vaWarn(formatPlus, args); va_end(args); if (vcfFileStopDueToErrors(vcff)) errAbort("VCF: %d parser errors, quitting", vcff->errCnt); } static void *vcfFileAlloc(struct vcfFile *vcff, size_t size) /* Use vcff's local mem to allocate memory. */ { return lmAlloc( vcfFileLm(vcff), size); } INLINE char *vcfFileCloneStrZ(struct vcfFile *vcff, char *str, size_t size) /* Use vcff's local mem to allocate memory for a string and copy it. */ { return lmCloneStringZ( vcfFileLm(vcff), str, size); } INLINE char *vcfFileCloneStr(struct vcfFile *vcff, char *str) /* Use vcff's local mem to allocate memory for a string and copy it. */ { return vcfFileCloneStrZ(vcff, str, strlen(str)); } INLINE char *vcfFileCloneSubstr(struct vcfFile *vcff, char *line, regmatch_t substr) /* Allocate memory for and copy a substring of line. */ { return vcfFileCloneStrZ(vcff, line+substr.rm_so, (substr.rm_eo - substr.rm_so)); } #define vcfFileCloneVar(vcff, var) lmCloneMem( vcfFileLm(vcff), var, sizeof(var)) char *vcfFilePooledStr(struct vcfFile *vcff, char *str) /* Allocate memory for a string from vcff's shared string pool. */ { return hashStoreName(vcff->pool, str); // Always stored in main pool, not reuse pool } static enum vcfInfoType vcfInfoTypeFromSubstr(struct vcfFile *vcff, char *line, regmatch_t substr) /* Translate substring of line into vcfInfoType or complain. */ { char typeWord[16]; int substrLen = substr.rm_eo - substr.rm_so; if (substrLen > sizeof(typeWord) - 1) { vcfFileErr(vcff, "substring passed to vcfInfoTypeFromSubstr is too long."); return vcfInfoNoType; } safencpy(typeWord, sizeof(typeWord), line + substr.rm_so, substrLen); if (sameString("Integer", typeWord)) return vcfInfoInteger; if (sameString("Float", typeWord)) return vcfInfoFloat; if (sameString("Flag", typeWord)) return vcfInfoFlag; if (sameString("Character", typeWord)) return vcfInfoCharacter; if (sameString("String", typeWord)) return vcfInfoString; vcfFileErr(vcff, "Unrecognized type word \"%s\" in metadata line \"%s\"", typeWord, line); return vcfInfoNoType; } // Regular expressions to check format and extract information from header lines: static const char *fileformatRegex = "^##(file)?format=VCFv([0-9]+)(\\.([0-9]+))?$"; static const char *infoOrFormatRegex = "^##(INFO|FORMAT)=" "<ID=([A-Za-z0-9_:-]+)," "Number=(\\.|A|G|[0-9-]+)," "Type=([A-Za-z]+)," "Description=\"?(.*)\"?>$"; static const char *filterOrAltRegex = "^##(FILTER|ALT)=" "<ID=([^,]+)," "(Description|Type)=\"?(.*)\"?>$"; // VCF version 3.3 was different enough to warrant separate regexes: static const char *infoOrFormatRegex3_3 = "^##(INFO|FORMAT)=" "([A-Za-z0-9_:-]+)," "(\\.|A|G|[0-9-]+)," "([A-Za-z]+)," "\"?(.*)\"?$"; static const char *filterRegex3_3 = "^##(FILTER)=" "([^,]+)," "()\"?(.*)\"?$"; INLINE void nonAsciiWorkaround(char *line) // Workaround for annoying 3-byte quote marks included in some 1000 Genomes files: { (void)strSwapStrs(line, strlen(line)+1, "\342\200\234", "\""); (void)strSwapStrs(line, strlen(line)+1, "\342\200\235", "\""); } static void parseMetadataLine(struct vcfFile *vcff, char *line) /* Parse a VCF header line beginning with "##" that defines a metadata. */ { char *ptr = line; if (ptr == NULL && !startsWith(ptr, "##")) errAbort("Bad line passed to parseMetadataLine"); ptr += 2; char *firstEq = strchr(ptr, '='); if (firstEq == NULL) { + if (vcff->majorVersion >= 4) vcfFileErr(vcff, "Metadata line lacks '=': \"%s\"", line); return; } regmatch_t substrs[8]; // Some of the metadata lines are crucial for parsing the rest of the file: if (startsWith("##fileformat=", line) || startsWith("##format", line)) { if (regexMatchSubstr(line, fileformatRegex, substrs, ArraySize(substrs))) { // substrs[2] is major version #, substrs[3] is set only if there is a minor version, // and substrs[4] is the minor version #. vcff->majorVersion = atoi(line + substrs[2].rm_so); if (substrs[3].rm_so != -1) vcff->minorVersion = atoi(line + substrs[4].rm_so); } else vcfFileErr(vcff, "##fileformat line does not match expected pattern /%s/: \"%s\"", fileformatRegex, line); } else if (startsWith("##INFO=", line) || startsWith("##FORMAT=", line)) { boolean isInfo = startsWith("##INFO=", line); nonAsciiWorkaround(line); if (regexMatchSubstr(line, infoOrFormatRegex, substrs, ArraySize(substrs)) || regexMatchSubstr(line, infoOrFormatRegex3_3, substrs, ArraySize(substrs))) // substrs[2] is ID/key, substrs[3] is Number, [4] is Type and [5] is Description. { struct vcfInfoDef *def = vcfFileAlloc(vcff, sizeof(struct vcfInfoDef)); def->key = vcfFileCloneSubstr(vcff, line, substrs[2]); char *number = vcfFileCloneSubstr(vcff, line, substrs[3]); if (sameString(number, ".") || sameString(number, "A") || sameString(number, "G")) // A is #alts which varies line-to-line; "G" is #genotypes which we haven't // yet seen. Why is there a G here -- shouldn't such attributes go in the // genotype columns? def->fieldCount = -1; else def->fieldCount = atoi(number); def->type = vcfInfoTypeFromSubstr(vcff, line, substrs[4]); // greedy regex pulls in end quote, trim if found: if (line[substrs[5].rm_eo-1] == '"') line[substrs[5].rm_eo-1] = '\0'; def->description = vcfFileCloneSubstr(vcff, line, substrs[5]); slAddHead((isInfo ? &(vcff->infoDefs) : &(vcff->gtFormatDefs)), def); } else vcfFileErr(vcff, "##%s line does not match expected pattern /%s/ or /%s/: \"%s\"", (isInfo ? "INFO" : "FORMAT"), infoOrFormatRegex, infoOrFormatRegex3_3, line); } else if (startsWith("##FILTER=", line) || startsWith("##ALT=", line)) { boolean isFilter = startsWith("##FILTER", line); if (regexMatchSubstr(line, filterOrAltRegex, substrs, ArraySize(substrs)) || regexMatchSubstr(line, filterRegex3_3, substrs, ArraySize(substrs))) { // substrs[2] is ID/key, substrs[4] is Description. struct vcfInfoDef *def = vcfFileAlloc(vcff, sizeof(struct vcfInfoDef)); def->key = vcfFileCloneSubstr(vcff, line, substrs[2]); def->description = vcfFileCloneSubstr(vcff, line, substrs[4]); slAddHead((isFilter ? &(vcff->filterDefs) : &(vcff->altDefs)), def); } else { if (isFilter) vcfFileErr(vcff, "##FILTER line does not match expected pattern /%s/ or /%s/: \"%s\"", filterOrAltRegex, filterRegex3_3, line); else vcfFileErr(vcff, "##ALT line does not match expected pattern /%s/: \"%s\"", filterOrAltRegex, line); } } } static void expectColumnName2(struct vcfFile *vcff, char *exp1, char *exp2, char *words[], int ix) /* Every file must include a header naming the columns, though most column names are * fixed; make sure the names of fixed columns are as expected. */ { if (! sameString(exp1, words[ix])) { if (exp2 == NULL) vcfFileErr(vcff, "Expected column %d's name in header to be \"%s\" but got \"%s\"", ix+1, exp1, words[ix]); else if (! sameString(exp2, words[ix])) vcfFileErr(vcff, "Expected column %d's name in header to be \"%s\" or \"%s\" " "but got \"%s\"", ix+1, exp1, exp2, words[ix]); } } #define expectColumnName(vcff, exp, words, ix) expectColumnName2(vcff, exp, NULL, words, ix) // There might be a whole lot of genotype columns... #define VCF_MAX_COLUMNS 16 * 1024 static void parseColumnHeaderRow(struct vcfFile *vcff, char *line) /* Make sure column names are as we expect, and store genotype sample IDs if any are given. */ { if (line[0] != '#') { vcfFileErr(vcff, "Expected to find # followed by column names (\"#CHROM POS ...\"), " "not \"%s\"", line); lineFileReuse(vcff->lf); return; } char *words[VCF_MAX_COLUMNS]; int wordCount = chopLine(line+1, words); if (wordCount >= VCF_MAX_COLUMNS) vcfFileErr(vcff, "header contains at least %d columns; " "VCF_MAX_COLUMNS may need to be increased in vcf.c!", VCF_MAX_COLUMNS); expectColumnName(vcff, "CHROM", words, 0); expectColumnName(vcff, "POS", words, 1); expectColumnName(vcff, "ID", words, 2); expectColumnName(vcff, "REF", words, 3); expectColumnName(vcff, "ALT", words, 4); expectColumnName2(vcff, "QUAL", "PROB", words, 5); expectColumnName(vcff, "FILTER", words, 6); expectColumnName(vcff, "INFO", words, 7); if (wordCount > 8) { expectColumnName(vcff, "FORMAT", words, 8); if (wordCount < 10) vcfFileErr(vcff, "FORMAT column is given, but no sample IDs for genotype columns...?"); vcff->genotypeCount = (wordCount - 9); vcff->genotypeIds = vcfFileAlloc(vcff, vcff->genotypeCount * sizeof(char *)); int i; for (i = 9; i < wordCount; i++) vcff->genotypeIds[i-9] = vcfFileCloneStr(vcff, words[i]); } } struct vcfFile *vcfFileNew() /* Return a new, empty vcfFile object. */ { struct vcfFile *vcff = NULL; AllocVar(vcff); vcff->pool = hashNew(0); vcff->reusePool = NULL; // Must explicitly request a separate record pool return vcff; } void vcfFileMakeReusePool(struct vcfFile *vcff, int initialSize) // Creates a separate memory pool for records. Establishing this pool allows // using vcfFileFlushRecords to abandon previously read records and free // the associated memory. Very useful when reading an entire file in batches. { assert(vcff->reusePool == NULL); // don't duplicate this vcff->reusePool = lmInit(initialSize); } void vcfFileFlushRecords(struct vcfFile *vcff) // Abandons all previously read vcff->records and flushes the reuse pool (if it exists). // USE WITH CAUTION. All previously allocated record pointers are now invalid. { if (vcff->reusePool != NULL) { size_t poolSize = lmSize(vcff->reusePool); //if (poolSize > (48 * 1024 * 1024)) // printf("\nReuse pool %ld of %ld unused\n",lmAvailable(vcff->reusePool),poolSize); lmCleanup(&vcff->reusePool); vcff->reusePool = lmInit(poolSize); } vcff->records = NULL; } static struct vcfFile *vcfFileHeaderFromLineFile(struct lineFile *lf, int maxErr) /* Parse a VCF file into a vcfFile object. If maxErr not zero, then * continue to parse until this number of error have been reached. A maxErr * less than zero does not stop and reports all errors. * Set maxErr to VCF_IGNORE_ERRS for silence */ { initVcfSpecInfoDefs(); initVcfSpecGtFormatDefs(); if (lf == NULL) return NULL; struct vcfFile *vcff = vcfFileNew(); vcff->lf = lf; vcff->fileOrUrl = vcfFileCloneStr(vcff, lf->fileName); vcff->maxErr = (maxErr < 0) ? INT_MAX : maxErr; struct dyString *dyHeader = dyStringNew(1024); char *line = NULL; // First, metadata lines beginning with "##": while (lineFileNext(lf, &line, NULL) && startsWith("##", line)) { dyStringAppend(dyHeader, line); dyStringAppendC(dyHeader, '\n'); parseMetadataLine(vcff, line); } slReverse(&(vcff->infoDefs)); slReverse(&(vcff->filterDefs)); slReverse(&(vcff->gtFormatDefs)); // Did we get the bare minimum VCF header with supported version? if (vcff->majorVersion == 0) vcfFileErr(vcff, "missing ##fileformat= header line? Assuming 4.1."); if ((vcff->majorVersion != 4 || (vcff->minorVersion != 0 && vcff->minorVersion != 1)) && (vcff->majorVersion != 3)) vcfFileErr(vcff, "VCFv%d.%d not supported -- only v3.*, v4.0 or v4.1", vcff->majorVersion, vcff->minorVersion); // Next, one header line beginning with single "#" that names the columns: if (line == NULL) // EOF after metadata return vcff; dyStringAppend(dyHeader, line); dyStringAppendC(dyHeader, '\n'); parseColumnHeaderRow(vcff, line); vcff->headerString = dyStringCannibalize(&dyHeader); return vcff; } #define VCF_MAX_INFO 512 static void parseRefAndAlt(struct vcfFile *vcff, struct vcfRecord *record, char *ref, char *alt) /* Make an array of alleles, ref first, from the REF and comma-sep'd ALT columns. * Use the length of the reference sequence to set record->chromEnd. * Note: this trashes the alt argument, since this is expected to be its last use. */ { char *altAlleles[VCF_MAX_INFO]; int altCount = chopCommas(alt, altAlleles); record->alleleCount = 1 + altCount; record->alleles = vcfFileAlloc(vcff, record->alleleCount * sizeof(record->alleles[0])); record->alleles[0] = vcfFilePooledStr(vcff, ref); int i; for (i = 0; i < altCount; i++) record->alleles[1+i] = vcfFilePooledStr(vcff, altAlleles[i]); int refLen = strlen(ref); if (refLen == dnaFilteredSize(ref)) record->chromEnd = record->chromStart + refLen; } static void parseFilterColumn(struct vcfFile *vcff, struct vcfRecord *record, char *filterStr) /* Transform ;-separated filter codes into count + string array. */ { // We don't want to modify something allocated with vcfFilePooledStr because that uses // hash element names for storage! So don't make a vcfFilePooledStr copy of filterStr and // chop that; instead, chop a temp string and pool the words separately. static struct dyString *tmp = NULL; if (tmp == NULL) tmp = dyStringNew(0); dyStringClear(tmp); dyStringAppend(tmp, filterStr); record->filterCount = countChars(filterStr, ';') + 1; record->filters = vcfFileAlloc(vcff, record->filterCount * sizeof(char **)); (void)chopByChar(tmp->string, ';', record->filters, record->filterCount); int i; for (i = 0; i < record->filterCount; i++) record->filters[i] = vcfFilePooledStr(vcff, record->filters[i]); } struct vcfInfoDef *vcfInfoDefForKey(struct vcfFile *vcff, const char *key) /* Return infoDef for key, or NULL if it wasn't specified in the header or VCF spec. */ { struct vcfInfoDef *def; // I expect there to be fairly few definitions (less than a dozen) so // I'm just doing a linear search not hash: for (def = vcff->infoDefs; def != NULL; def = def->next) { if (sameString(key, def->key)) return def; } for (def = vcfSpecInfoDefs; def != NULL; def = def->next) { if (sameString(key, def->key)) return def; } return NULL; } static enum vcfInfoType typeForInfoKey(struct vcfFile *vcff, const char *key) /* Look up the type of INFO component key, in the definitions from the header, * and failing that, from the keys reserved in the spec. */ { struct vcfInfoDef *def = vcfInfoDefForKey(vcff, key); if (def == NULL) { vcfFileErr(vcff, "There is no INFO header defining \"%s\"", key); // default to string so we can display value as-is: return vcfInfoString; } return def->type; } static int parseInfoValue(struct vcfRecord *record, char *infoKey, enum vcfInfoType type, char *valStr, union vcfDatum **pData, bool **pMissingData) /* Parse a comma-separated list of values into array of union vcfInfoDatum and return count. */ { char *valWords[VCF_MAX_INFO]; int count = chopCommas(valStr, valWords); struct vcfFile *vcff = record->file; union vcfDatum *data = vcfFileAlloc(vcff, count * sizeof(union vcfDatum)); bool *missingData = vcfFileAlloc(vcff, count * sizeof(*missingData)); int j; for (j = 0; j < count; j++) { if (type != vcfInfoString && type != vcfInfoCharacter && sameString(valWords[j], ".")) missingData[j] = TRUE; switch (type) { case vcfInfoInteger: data[j].datInt = atoi(valWords[j]); break; case vcfInfoFloat: data[j].datFloat = atof(valWords[j]); break; case vcfInfoFlag: // Flag key might have a value in older VCFs e.g. 3.2's DB=0, DB=1 data[j].datString = vcfFilePooledStr(vcff, valWords[j]); break; case vcfInfoCharacter: data[j].datChar = valWords[j][0]; break; case vcfInfoString: data[j].datString = vcfFilePooledStr(vcff, valWords[j]); break; default: errAbort("invalid vcfInfoType (uninitialized?) %d", type); break; } } // If END is given, use it as chromEnd: if (sameString(infoKey, vcfInfoEnd)) record->chromEnd = data[0].datInt; *pData = data; *pMissingData = missingData; return count; } static void parseInfoColumn(struct vcfFile *vcff, struct vcfRecord *record, char *string) /* Translate string into array of vcfInfoElement. */ { if (sameString(string, ".")) { record->infoCount = 0; return; } char *elWords[VCF_MAX_INFO]; record->infoCount = chopByChar(string, ';', elWords, ArraySize(elWords)); if (record->infoCount >= VCF_MAX_INFO) vcfFileErr(vcff, "INFO column contains at least %d elements; " "VCF_MAX_INFO may need to be increased in vcf.c!", VCF_MAX_INFO); record->infoElements = vcfFileAlloc(vcff, record->infoCount * sizeof(struct vcfInfoElement)); char *emptyString = vcfFilePooledStr(vcff, ""); int i; for (i = 0; i < record->infoCount; i++) { char *elStr = elWords[i]; char *eq = strchr(elStr, '='); struct vcfInfoElement *el = &(record->infoElements[i]); if (eq == NULL) { el->key = vcfFilePooledStr(vcff, elStr); enum vcfInfoType type = typeForInfoKey(vcff, el->key); if (type != vcfInfoFlag) { struct vcfInfoDef *def = vcfInfoDefForKey(vcff, el->key); // Complain only if we are expecting a particular number of values for this keyword: if (def != NULL && def->fieldCount >= 0) vcfFileErr(vcff, "Missing = after key in INFO element: \"%s\" (type=%d)", elStr, type); if (type == vcfInfoString) { el->values = vcfFileAlloc(vcff, sizeof(union vcfDatum)); el->values[0].datString = emptyString; } } continue; } *eq = '\0'; el->key = vcfFilePooledStr(vcff, elStr); enum vcfInfoType type = typeForInfoKey(vcff, el->key); char *valStr = eq+1; el->count = parseInfoValue(record, el->key, type, valStr, &(el->values), &(el->missingData)); if (el->count >= VCF_MAX_INFO) vcfFileErr(vcff, "A single element of the INFO column has at least %d values; " "VCF_MAX_INFO may need to be increased in vcf.c!", VCF_MAX_INFO); } } struct vcfRecord *vcfRecordFromRow(struct vcfFile *vcff, char **words) /* Parse words from a VCF data line into a VCF record structure. */ { struct vcfRecord *record = vcfFileAlloc(vcff, sizeof(struct vcfRecord)); record->file = vcff; record->chrom = vcfFilePooledStr(vcff, words[0]); record->chromStart = lineFileNeedNum(vcff->lf, words, 1) - 1; // chromEnd may be overwritten by parseRefAndAlt and parseInfoColumn. record->chromEnd = record->chromStart+1; record->name = vcfFilePooledStr(vcff, words[2]); parseRefAndAlt(vcff, record, words[3], words[4]); record->qual = vcfFilePooledStr(vcff, words[5]); parseFilterColumn(vcff, record, words[6]); parseInfoColumn(vcff, record, words[7]); if (vcff->genotypeCount > 0) { record->format = vcfFilePooledStr(vcff, words[8]); record->genotypeUnparsedStrings = vcfFileAlloc(vcff, vcff->genotypeCount * sizeof(char *)); int i; // Don't bother actually parsing all these until & unless we need the info: for (i = 0; i < vcff->genotypeCount; i++) record->genotypeUnparsedStrings[i] = vcfFileCloneStr(vcff, words[9+i]); } return record; } struct vcfRecord *vcfNextRecord(struct vcfFile *vcff) /* Parse the words in the next line from vcff into a vcfRecord. Return NULL at end of file. * Note: this does not store record in vcff->records! */ { char *words[VCF_MAX_COLUMNS]; int wordCount; if ((wordCount = lineFileChop(vcff->lf, words)) <= 0) return NULL; int expected = 8; if (vcff->genotypeCount > 0) expected = 9 + vcff->genotypeCount; lineFileExpectWords(vcff->lf, expected, wordCount); return vcfRecordFromRow(vcff, words); } static boolean allelesHavePaddingBase(char **alleles, int alleleCount) /* Examine alleles to see if they either a) all start with the same base or * b) include a symbolic or 0-length allele. In either of those cases, there * must be an initial padding base that we'll need to trim from non-symbolic * alleles. */ { boolean hasPaddingBase = TRUE; char firstBase = '\0'; if (isAllNt(alleles[0], strlen(alleles[0]))) firstBase = alleles[0][0]; int i; for (i = 1; i < alleleCount; i++) { if (isAllNt(alleles[i], strlen(alleles[i]))) { if (firstBase == '\0') firstBase = alleles[i][0]; if (alleles[i][0] != firstBase) // Different first base implies unpadded alleles. hasPaddingBase = FALSE; } else { // Symbolic ALT allele: REF must have the padding base. hasPaddingBase = TRUE; break; } } return hasPaddingBase; } unsigned int vcfRecordTrimIndelLeftBase(struct vcfRecord *rec) /* For indels, VCF includes the left neighboring base; for example, if the alleles are * AA/- following a G base, then the VCF record will start one base to the left and have * "GAA" and "G" as the alleles. Also, if any alt allele is symbolic (e.g. <DEL>) then * the ref allele must have a padding base. * That is not nice for display for two reasons: * 1. Indels appear one base wider than their dbSNP entries. * 2. In pgSnp display mode, the two alleles are always the same color. * However, for hgTracks' mapBox we need the correct chromStart for identifying the * record in hgc -- so return the original chromStart. */ { unsigned int chromStartOrig = rec->chromStart; struct vcfFile *vcff = rec->file; if (rec->alleleCount > 1) { boolean hasPaddingBase = allelesHavePaddingBase(rec->alleles, rec->alleleCount); if (hasPaddingBase) { rec->chromStart++; int i; for (i = 0; i < rec->alleleCount; i++) { if (rec->alleles[i][1] == '\0') rec->alleles[i] = vcfFilePooledStr(vcff, "-"); else if (isAllNt(rec->alleles[i], strlen(rec->alleles[i]))) rec->alleles[i] = vcfFilePooledStr(vcff, rec->alleles[i]+1); else // don't trim first character of symbolic allele rec->alleles[i] = vcfFilePooledStr(vcff, rec->alleles[i]); } } } return chromStartOrig; } static struct vcfRecord *vcfParseData(struct vcfFile *vcff, int maxRecords) // Given a vcfFile into which the header has been parsed, and whose // lineFile is positioned at the beginning of a data row, parse and // return all data rows from lineFile. { if (vcff == NULL) return NULL; int recCount = 0; struct vcfRecord *records = NULL; struct vcfRecord *record; while ((record = vcfNextRecord(vcff)) != NULL) { if (maxRecords >= 0 && recCount >= maxRecords) break; slAddHead(&records, record); recCount++; } slReverse(&records); return records; } struct vcfFile *vcfFileMayOpen(char *fileOrUrl, int maxErr, int maxRecords, boolean parseAll) /* Open fileOrUrl and parse VCF header; return NULL if unable. * If parseAll, then read in all lines, parse and store in * vcff->records; if maxErr >= zero, then continue to parse until * there are maxErr+1 errors. A maxErr less than zero does not stop * and reports all errors. Set maxErr to VCF_IGNORE_ERRS for silence */ { struct lineFile *lf = NULL; if (startsWith("http://", fileOrUrl) || startsWith("ftp://", fileOrUrl) || startsWith("https://", fileOrUrl)) lf = netLineFileOpen(fileOrUrl); else lf = lineFileMayOpen(fileOrUrl, TRUE); struct vcfFile *vcff = vcfFileHeaderFromLineFile(lf, maxErr); if (parseAll) { vcff->records = vcfParseData(vcff, maxRecords); lineFileClose(&(vcff->lf)); // Not sure why it is closed. Angie? } return vcff; } struct vcfFile *vcfTabixFileMayOpen(char *fileOrUrl, char *chrom, int start, int end, int maxErr, int maxRecords) /* Open a VCF file that has been compressed and indexed by tabix and * parse VCF header, or return NULL if unable. If chrom is non-NULL, * seek to the position range and parse all lines in range into * vcff->records. If maxErr >= zero, then continue to parse until * there are maxErr+1 errors. A maxErr less than zero does not stop * and reports all errors. Set maxErr to VCF_IGNORE_ERRS for silence */ { struct lineFile *lf = lineFileTabixMayOpen(fileOrUrl, TRUE); struct vcfFile *vcff = vcfFileHeaderFromLineFile(lf, maxErr); if (vcff == NULL) return NULL; if (isNotEmpty(chrom) && start != end) { if (lineFileSetTabixRegion(lf, chrom, start, end)) { vcff->records = vcfParseData(vcff, maxRecords); lineFileClose(&(vcff->lf)); // Not sure why it is closed. Angie? } } return vcff; } int vcfRecordCmp(const void *va, const void *vb) /* Compare to sort based on position. */ { const struct vcfRecord *a = *((struct vcfRecord **)va); const struct vcfRecord *b = *((struct vcfRecord **)vb); int dif; dif = strcmp(a->chrom, b->chrom); if (dif == 0) dif = a->chromStart - b->chromStart; if (dif == 0) dif = a->chromEnd - b->chromEnd; // shortest first if (dif == 0) dif = strcmp(a->name, b->name); // finally by name return dif; } int vcfTabixBatchRead(struct vcfFile *vcff, char *chrom, int start, int end, int maxErr, int maxRecords) // Reads a batch of records from an opened and indexed VCF file, adding them to // vcff->records and returning the count of new records added in this batch. // Note: vcff->records will continue to be sorted, even if batches are loaded // out of order. Additionally, resulting vcff->records will contain no duplicates // so returned count refects only the new records added, as opposed to all records // in range. If maxErr >= zero, then continue to parse until there are maxErr+1 // errors. A maxErr less than zero does not stop and reports all errors. Set // maxErr to VCF_IGNORE_ERRS for silence. { int oldCount = slCount(vcff->records); if (lineFileSetTabixRegion(vcff->lf, chrom, start, end)) { struct vcfRecord *records = vcfParseData(vcff, maxRecords); if (records) { struct vcfRecord *lastRec = vcff->records; if (lastRec == NULL) vcff->records = records; else { // Considered just asserting the batches were in order, but a problem may // result when non-overlapping location windows pick up the same long variant. slSortMergeUniq(&(vcff->records), records, vcfRecordCmp, NULL); } } } return slCount(vcff->records) - oldCount; } void vcfFileFree(struct vcfFile **pVcff) /* Free a vcfFile object. */ { if (pVcff == NULL || *pVcff == NULL) return; struct vcfFile *vcff = *pVcff; if (vcff->maxErr == VCF_IGNORE_ERRS && vcff->errCnt > 0) { vcff->maxErr++; // Never completely silent! Go ahead and report how many errs were detected vcfFileErr(vcff,"Closing with %d errors.",vcff->errCnt); } freez(&(vcff->headerString)); hashFree(&(vcff->pool)); if (vcff->reusePool) lmCleanup(&vcff->reusePool); hashFree(&(vcff->byName)); lineFileClose(&(vcff->lf)); freez(pVcff); } const struct vcfRecord *vcfFileFindVariant(struct vcfFile *vcff, char *variantId) /* Return all records with name=variantId, or NULL if not found. */ { struct vcfRecord *varList = NULL; if (vcff->byName == NULL) { vcff->byName = hashNew(0); struct vcfRecord *rec; for (rec = vcff->records; rec != NULL; rec = rec->next) { if (sameString(rec->name, variantId)) { // Make shallow copy of rec so we can alter ->next: struct vcfRecord *newRec = vcfFileCloneVar(vcff,rec); slAddHead(&varList, newRec); } hashAdd(vcff->byName, rec->name, rec); } slReverse(&varList); } else { struct hashEl *hel = hashLookup(vcff->byName, variantId); while (hel != NULL) { if (sameString(hel->name, variantId)) { struct vcfRecord *rec = hel->val; struct vcfRecord *newRec = vcfFileCloneVar(vcff,rec); slAddHead(&varList, newRec); } hel = hel->next; } // Don't reverse varList -- hash element list was already reversed } return varList; } const struct vcfInfoElement *vcfRecordFindInfo(const struct vcfRecord *record, char *key) /* Find an INFO element, or NULL. */ { int i; for (i = 0; i < record->infoCount; i++) { if (sameString(key, record->infoElements[i].key)) return &(record->infoElements[i]); } return NULL; } struct vcfInfoDef *vcfInfoDefForGtKey(struct vcfFile *vcff, const char *key) /* Look up the type of genotype FORMAT component key, in the definitions from the header, * and failing that, from the keys reserved in the spec. */ { struct vcfInfoDef *def; // I expect there to be fairly few definitions (less than a dozen) so // I'm just doing a linear search not hash: for (def = vcff->gtFormatDefs; def != NULL; def = def->next) { if (sameString(key, def->key)) return def; } for (def = vcfSpecGtFormatDefs; def != NULL; def = def->next) { if (sameString(key, def->key)) return def; } return NULL; } static enum vcfInfoType typeForGtFormat(struct vcfFile *vcff, const char *key) /* Look up the type of FORMAT component key, in the definitions from the header, * and failing that, from the keys reserved in the spec. */ { struct vcfInfoDef *def = vcfInfoDefForGtKey(vcff, key); if (def == NULL) { vcfFileErr(vcff, "There is no FORMAT header defining \"%s\"", key); // default to string so we can display value as-is: return vcfInfoString; } return def->type; } #define VCF_MAX_FORMAT VCF_MAX_INFO #define VCF_MAX_FORMAT_LEN (VCF_MAX_FORMAT * 4) void vcfParseGenotypes(struct vcfRecord *record) /* Translate record->genotypesUnparsedStrings[] into proper struct vcfGenotype[]. * This destroys genotypesUnparsedStrings. */ { if (record->genotypeUnparsedStrings == NULL) return; struct vcfFile *vcff = record->file; record->genotypes = vcfFileAlloc(vcff, vcff->genotypeCount * sizeof(struct vcfGenotype)); char format[VCF_MAX_FORMAT_LEN]; safecpy(format, sizeof(format), record->format); char *formatWords[VCF_MAX_FORMAT]; int formatWordCount = chopByChar(format, ':', formatWords, ArraySize(formatWords)); if (formatWordCount >= VCF_MAX_FORMAT) { vcfFileErr(vcff, "The FORMAT column has at least %d words; " "VCF_MAX_FORMAT may need to be increased in vcf.c!", VCF_MAX_FORMAT); formatWordCount = VCF_MAX_FORMAT; } if (differentString(formatWords[0], vcfGtGenotype)) vcfFileErr(vcff, "FORMAT column should begin with \"%s\" but begins with \"%s\"", vcfGtGenotype, formatWords[0]); int i; // Store the pooled format word pointers and associated types for use in inner loop below. enum vcfInfoType formatTypes[VCF_MAX_FORMAT]; for (i = 0; i < formatWordCount; i++) { formatTypes[i] = typeForGtFormat(vcff, formatWords[i]); formatWords[i] = vcfFilePooledStr(vcff, formatWords[i]); } for (i = 0; i < vcff->genotypeCount; i++) { char *string = record->genotypeUnparsedStrings[i]; struct vcfGenotype *gt = &(record->genotypes[i]); // Each genotype can have multiple :-separated info elements: char *gtWords[VCF_MAX_FORMAT]; int gtWordCount = chopByChar(string, ':', gtWords, ArraySize(gtWords)); if (gtWordCount != formatWordCount) vcfFileErr(vcff, "The FORMAT column has %d words but the genotype column for %s " "has %d words", formatWordCount, vcff->genotypeIds[i], gtWordCount); if (gtWordCount > formatWordCount) gtWordCount = formatWordCount; gt->id = vcff->genotypeIds[i]; gt->infoCount = gtWordCount; gt->infoElements = vcfFileAlloc(vcff, gtWordCount * sizeof(struct vcfInfoElement)); int j; for (j = 0; j < gtWordCount; j++) { // Special parsing of genotype: if (sameString(formatWords[j], vcfGtGenotype)) { char *genotype = gtWords[j]; char *sep = strchr(genotype, '|'); if (sep != NULL) gt->isPhased = TRUE; else sep = strchr(genotype, '/'); if (genotype[0] == '.') gt->hapIxA = -1; else gt->hapIxA = atoi(genotype); if (sep == NULL) gt->isHaploid = TRUE; else if (sep[1] == '.') gt->hapIxB = -1; else gt->hapIxB = atoi(sep+1); } struct vcfInfoElement *el = &(gt->infoElements[j]); el->key = formatWords[j]; el->count = parseInfoValue(record, formatWords[j], formatTypes[j], gtWords[j], &(el->values), &(el->missingData)); if (el->count >= VCF_MAX_INFO) vcfFileErr(vcff, "A single element of the genotype column for \"%s\" " "has at least %d values; " "VCF_MAX_INFO may need to be increased in vcf.c!", gt->id, VCF_MAX_INFO); } } record->genotypeUnparsedStrings = NULL; } const struct vcfGenotype *vcfRecordFindGenotype(struct vcfRecord *record, char *sampleId) /* Find the genotype and associated info for the individual, or return NULL. * This calls vcfParseGenotypes if it has not already been called. */ { struct vcfFile *vcff = record->file; if (sampleId == NULL || vcff->genotypeCount == 0) return NULL; vcfParseGenotypes(record); int ix = stringArrayIx(sampleId, vcff->genotypeIds, vcff->genotypeCount); if (ix >= 0) return &(record->genotypes[ix]); return NULL; } static char *vcfDataLineAutoSqlString = "table vcfDataLine" "\"The fields of a Variant Call Format data line\"" " (" " string chrom; \"An identifier from the reference genome\"" " uint pos; \"The reference position, with the 1st base having position 1\"" " string id; \"Semi-colon separated list of unique identifiers where available\"" " string ref; \"Reference base(s)\"" " string alt; \"Comma separated list of alternate non-reference alleles " "called on at least one of the samples\"" " string qual; \"Phred-scaled quality score for the assertion made in ALT. i.e. " "give -10log_10 prob(call in ALT is wrong)\"" " string filter; \"PASS if this position has passed all filters. Otherwise, a " "semicolon-separated list of codes for filters that fail\"" " string info; \"Additional information encoded as a semicolon-separated series " "of short keys with optional comma-separated values\"" " string format; \"If genotype columns are specified in header, a " "semicolon-separated list of of short keys starting with GT\"" " string genotypes; \"If genotype columns are specified in header, a tab-separated " "set of genotype column values; each value is a colon-separated " "list of values corresponding to keys in the format column\"" " )"; struct asObject *vcfAsObj() // Return asObject describing fields of VCF { return asParseText(vcfDataLineAutoSqlString); } char *vcfGetSlashSepAllelesFromWords(char **words, struct dyString *dy) /* Overwrite dy with a /-separated allele string from VCF words, * skipping the extra initial base that VCF requires for indel alleles if necessary. * Return dy->string for convenience. */ { dyStringClear(dy); // VCF reference allele gets its own column: char *refAllele = words[3]; char *altAlleles = words[4]; // Make a vcfRecord-like allele array (ref in [0], alts after) so we can check for padding base: int alCount = 1 + countChars(altAlleles, ',') + 1; char *alleles[alCount]; alleles[0] = refAllele; dyStringAppend(dy, altAlleles); chopByChar(dy->string, ',', &(alleles[1]), alCount-1); boolean hasPaddingBase = allelesHavePaddingBase(alleles, alCount); int offset = hasPaddingBase ? 1 : 0; // Build a /-separated allele string, trimming first base if offset is 1: dyStringClear(dy); if (refAllele[offset] == '\0') dyStringAppendC(dy, '-'); else dyStringAppend(dy, refAllele+offset); if (isNotEmpty(altAlleles) && differentString(altAlleles, ".")) { // Read alleles from altAlleles because the contents of alleles[] are trashed // by our reuse of dy. char *p; while ((p = strchr(altAlleles, ',')) != NULL) { dyStringAppendC(dy, '/'); int len = p - altAlleles; if (len-offset == 0) dyStringAppendC(dy, '-'); else if (isAllNt(altAlleles, len)) dyStringAppendN(dy, altAlleles+offset, len-offset); else dyStringAppendN(dy, altAlleles, len); // symbolic altAlleles = p+1; } dyStringAppendC(dy, '/'); int len = strlen(altAlleles); if (len-offset == 0) dyStringAppendC(dy, '-'); else if (isAllNt(altAlleles, len)) dyStringAppendN(dy, altAlleles+offset, len-offset); else dyStringAppendN(dy, altAlleles, len); // symbolic } return dy->string; }