void doBedReplicate(struct sqlConnection *conn, char *format, struct cdwAssembly *assembly,
struct cdwFile *elderEf, struct cdwValidFile *elderVf,
struct cdwFile *youngerEf, struct cdwValidFile *youngerVf)
/* Do correlation analysis between elder and younger and save result to
* a new cdwQaPairCorrelation record. Do this for a format where we have a bigBed file. */
{
/* If got both pairs, work is done already */
if (pairExists(conn, elderEf->id, youngerEf->id, "cdwQaPairSampleOverlap"))
return;
/* Get files for both younger and older. */
char *elderPath = cdwPathForFileId(conn, elderEf->id);
char *youngerPath = cdwPathForFileId(conn, youngerEf->id);
/* Do replicate calcs on bed3 lists from files. */
struct bed3 *elderBedList = bed3LoadAll(elderPath);
struct bed3 *youngerBedList = bed3LoadAll(youngerPath);
doBed3Replicate(conn, format, assembly, elderEf, elderVf, elderBedList,
youngerEf, youngerVf, youngerBedList);
/* Clean up. */
bed3FreeList(&elderBedList);
bed3FreeList(&youngerBedList);
freez(&youngerPath);
freez(&elderPath);
}
void doBigWigReplicate(struct sqlConnection *conn, struct cdwAssembly *assembly,
struct cdwFile *elderEf, struct cdwValidFile *elderVf,
struct cdwFile *youngerEf, struct cdwValidFile *youngerVf)
/* Do correlation analysis between elder and younger and save result to
* a new cdwQaPairCorrelation record. Do this for a format where we have a bigWig file. */
{
if (pairExists(conn, elderEf->id, youngerEf->id, "cdwQaPairCorrelation"))
return;
char *enrichedIn = elderVf->enrichedIn;
if (!isEmpty(enrichedIn) && !sameString(enrichedIn, "unknown"))
{
struct genomeRangeTree *targetGrt = genomeRangeTreeForTarget(conn, assembly, enrichedIn);
/* Get open big wig files for both younger and older. */
char *elderPath = cdwPathForFileId(conn, elderEf->id);
char *youngerPath = cdwPathForFileId(conn, youngerEf->id);
struct bbiFile *elderBbi = bigWigFileOpen(elderPath);
struct bbiFile *youngerBbi = bigWigFileOpen(youngerPath);
/* Figure out thresholds */
double elderThreshold = twoStdsOverMean(elderBbi);
double youngerThreshold = twoStdsOverMean(youngerBbi);
/* Loop through a chromosome at a time adding to correlation, and at the end save result in r.*/
struct correlate *c = correlateNew(), *cInEnriched = correlateNew(), *cClipped = correlateNew();
struct bbiChromInfo *chrom, *chromList = bbiChromList(elderBbi);
struct bigWigValsOnChrom *aVals = bigWigValsOnChromNew();
struct bigWigValsOnChrom *bVals = bigWigValsOnChromNew();
for (chrom = chromList; chrom != NULL; chrom = chrom->next)
{
addBwCorrelations(chrom, targetGrt, aVals, bVals, elderBbi, youngerBbi,
elderThreshold, youngerThreshold, c, cInEnriched, cClipped);
}
/* Make up correlation structure . */
struct cdwQaPairCorrelation *cor;
AllocVar(cor);
cor->elderFileId = elderVf->fileId;
cor->youngerFileId = youngerVf->fileId;
cor->pearsonOverall = correlateResult(c);
cor->pearsonInEnriched = correlateResult(cInEnriched);
cor->pearsonClipped = correlateResult(cClipped);
cdwQaPairCorrelationSaveToDb(conn, cor, "cdwQaPairCorrelation", 128);
bigWigValsOnChromFree(&bVals);
bigWigValsOnChromFree(&aVals);
genomeRangeTreeFree(&targetGrt);
freez(&cor);
correlateFree(&c);
bigWigFileClose(&youngerBbi);
bigWigFileClose(&elderBbi);
freez(&youngerPath);
freez(&elderPath);
}
}
void sendFileByAcc(struct sqlConnection *conn, char* acc)
/* send file identified by acc (=cdwValidFile.licensePlate), suggests a canonical filename of the format
* <licensePlate>.<originalExtension>
* Example URL: http://hgwdev.soe.ucsc.edu/cgi-bin/cdwGetFile?acc=SCH000FSW */
{
struct cdwValidFile *vf = cdwValidFileFromLicensePlate(conn, acc);
if (vf==NULL)
errExit("%s is not a valid accession in the CDW.", acc);
struct cdwFile *ef = cdwFileFromId(conn, vf->fileId);
char* filePath = cdwPathForFileId(conn, vf->fileId);
mustHaveAccess(conn, ef);
// use the license plate as the basename of the downloaded file.
// Take the extension from the submitted filename, as cdwFile.format is not the same as the extension
// e.g. format=fasta -> fa.gz
char *submitExt = skipBeyondDelimit(basename(ef->submitFileName), '.');
char suggestName[8000];
safef(suggestName, sizeof(suggestName), "%s.%s", vf->licensePlate, submitExt);
apacheSendX(vf->format, filePath, suggestName);
}
void sendFileByPath(struct sqlConnection *conn, char *path)
/* send file identified by a submission pathname (cdwFile.submitFileName),
* suggests the original filename. */
/* path can be a suffix that matches a filename, so the initial '/hive' or
* '/data' can be omitted. *
* Example URL for testing:
* http://hgwdev.soe.ucsc.edu/cgi-bin/cdwGetFile/hive/groups/cirm/pilot/labs/quake/130625_M00361_0080_000000000-A43D1/Sample_1_L13_C31_IL3541-701-506/1_L13_C31_IL3541-701-506_TAAGGCGA-ACTGCATA_L001_R1_001.fastq.gz */
{
int fileId = cdwFileIdFromPathSuffix(conn, path);
if (fileId == 0)
errExit("A file with suffix %s does not exist in the database", path);
char *localPath = cdwPathForFileId(conn, fileId);
if (localPath == NULL)
errExit("A local file with suffix %s was not found in the database. This is an internal error.", path);
struct cdwFile *ef = cdwFileFromId(conn, fileId);
if (ef == NULL)
errExit("Could not find cdwFile for path %s", path);
mustHaveAccess(conn, ef);
apacheSendX(NULL, localPath, basename(ef->submitFileName));
}
void doEnrichmentsFromBed(struct sqlConnection *conn,
struct cdwFile *ef, struct cdwValidFile *vf,
struct cdwAssembly *assembly, struct target *targetList)
/* Figure out enrichments from a bed file. */
{
char *bedPath = cdwPathForFileId(conn, ef->id);
struct bed3 *sampleList = bed3LoadAll(bedPath);
doEnrichmentsFromBed3Sample(sampleList, conn, ef, vf, assembly, targetList);
bed3FreeList(&sampleList);
freez(&bedPath);
}
void maybeRemoveFile(struct sqlConnection *conn, long long fileId, boolean really)
/* Remove references to file, and file itself from database. If really is FALSE just print out
* what we would do. */
{
char query[256];
/* Delete from all the auxiliarry tables - tables are alphabetical to help update. */
cdwReally
sqlSafef(query, sizeof(query), "delete from cdwBamFile where fileId=%lld", fileId);
maybeDoUpdate(conn, query, really);
sqlSafef(query, sizeof(query), "delete from cdwFastqFile where fileId=%lld", fileId);
maybeDoUpdate(conn, query, really);
sqlSafef(query, sizeof(query), "delete from cdwVcfFile where fileId=%lld", fileId);
maybeDoUpdate(conn, query, really);
sqlSafef(query, sizeof(query), "delete from cdwTrackViz where fileId=%lld", fileId);
maybeDoUpdate(conn, query, really);
sqlSafef(query, sizeof(query), "delete from cdwQaContam where fileId=%lld", fileId);
maybeDoUpdate(conn, query, really);
sqlSafef(query, sizeof(query), "delete from cdwQaEnrich where fileId=%lld", fileId);
maybeDoUpdate(conn, query, really);
sqlSafef(query, sizeof(query), "delete from cdwQaFail where fileId=%lld", fileId);
maybeDoUpdate(conn, query, really);
sqlSafef(query, sizeof(query),
"delete from cdwQaPairCorrelation where elderFileId=%lld or youngerFileId=%lld",
fileId, fileId);
maybeDoUpdate(conn, query, really);
sqlSafef(query, sizeof(query),
"delete from cdwQaPairSampleOverlap where elderFileId=%lld or youngerFileId=%lld",
fileId, fileId);
maybeDoUpdate(conn, query, really);
sqlSafef(query, sizeof(query),
"delete from cdwQaPairedEndFastq where fileId1=%lld or fileId2=%lld",
fileId, fileId);
maybeDoUpdate(conn, query, really);
sqlSafef(query, sizeof(query), "delete from cdwQaRepeat where fileId=%lld", fileId);
maybeDoUpdate(conn, query, really);
sqlSafef(query, sizeof(query), "delete from cdwValidFile where fileId=%lld", fileId);
maybeDoUpdate(conn, query, really);
/* Get file name */
char *path = cdwPathForFileId(conn, fileId);
/* Delete from cdwFileTable */
sqlSafef(query, sizeof(query), "delete from cdwFile where id=%lld", fileId);
maybeDoUpdate(conn, query, really);
/* Delete file */
if (really)
remove(path);
else
printf("remove: %s\n", path);
}
static struct genomeRangeTree *genomeRangeTreeForTarget(struct sqlConnection *conn,
struct cdwAssembly *assembly, char *enrichedIn)
/* Return genome range tree filled with enrichment target for assembly */
{
char query[256];
sqlSafef(query, sizeof(query), "select * from cdwQaEnrichTarget where assemblyId=%d and name='%s'",
assembly->id, enrichedIn);
struct cdwQaEnrichTarget *target = cdwQaEnrichTargetLoadByQuery(conn, query);
if (target == NULL)
errAbort("Can't find %s enrichment target for assembly %s", enrichedIn, assembly->name);
char *targetPath = cdwPathForFileId(conn, target->fileId);
struct genomeRangeTree *targetGrt = cdwGrtFromBigBed(targetPath);
cdwQaEnrichTargetFree(&target);
freez(&targetPath);
return targetGrt;
}
struct target *targetsForAssembly(struct sqlConnection *conn, struct cdwAssembly *assembly)
/* Get list of enrichment targets for given assembly */
{
char query[128];
sqlSafef(query, sizeof(query), "select * from cdwQaEnrichTarget where assemblyId=%d", assembly->id);
struct cdwQaEnrichTarget *et, *etList = cdwQaEnrichTargetLoadByQuery(conn, query);
/* Wrap a new structure around the enrichment targets where we'll store summary info. */
struct target *target, *targetList = NULL, **targetTail = &targetList;
for (et = etList; et != NULL; et = et->next)
{
char *targetBed = cdwPathForFileId(conn, et->fileId);
struct genomeRangeTree *grt = cdwGrtFromBigBed(targetBed);
target = targetNew(et, grt);
*targetTail = target;
targetTail = &target->next;
freez(&targetBed);
}
return targetList;
}
void doEnrichmentsFromBigWig(struct sqlConnection *conn,
struct cdwFile *ef, struct cdwValidFile *vf,
struct cdwAssembly *assembly, struct target *targetList)
/* Figure out enrichments from a bigBed file. */
{
/* Get path to bigBed, open it, and read all chromosomes. */
char *bigWigPath = cdwPathForFileId(conn, ef->id);
struct bbiFile *bbi = bigWigFileOpen(bigWigPath);
struct bbiChromInfo *chrom, *chromList = bbiChromList(bbi);
struct bigWigValsOnChrom *valsOnChrom = bigWigValsOnChromNew();
/* This takes a while, so let's figure out what parts take the time. */
long totalBigQueryTime = 0;
long totalOverlapTime = 0;
/* Do a pretty complex loop that just aims to set target->overlapBases and ->uniqOverlapBases
* for all targets. This is complicated by just wanting to keep one chromosome worth of
* bigWig data in memory. Also just for performance we do a lookup of target range tree to
* get chromosome specific one to use, which avoids a hash lookup in the inner loop. */
for (chrom = chromList; chrom != NULL; chrom = chrom->next)
{
long startBigQueryTime = clock1000();
boolean gotData = bigWigValsOnChromFetchData(valsOnChrom, chrom->name, bbi);
long endBigQueryTime = clock1000();
totalBigQueryTime += endBigQueryTime - startBigQueryTime;
if (gotData)
{
double *valBuf = valsOnChrom->valBuf;
Bits *covBuf = valsOnChrom->covBuf;
/* Loop through all targets adding overlaps from ivList */
long startOverlapTime = clock1000();
struct target *target;
for (target = targetList; target != NULL; target = target->next)
{
if (target->skip)
continue;
struct genomeRangeTree *grt = target->grt;
struct rbTree *targetTree = genomeRangeTreeFindRangeTree(grt, chrom->name);
if (targetTree != NULL)
{
struct range *range, *rangeList = rangeTreeList(targetTree);
for (range = rangeList; range != NULL; range = range->next)
{
int s = range->start, e = range->end, i;
for (i=s; i<=e; ++i)
{
if (bitReadOne(covBuf, i))
{
double x = valBuf[i];
target->uniqOverlapBases += 1;
target->overlapBases += x;
}
}
}
}
}
long endOverlapTime = clock1000();
totalOverlapTime += endOverlapTime - startOverlapTime;
}
}
verbose(1, "totalBig %0.3f, totalOverlap %0.3f\n", 0.001*totalBigQueryTime, 0.001*totalOverlapTime);
/* Now loop through targets and save enrichment info to database */
struct target *target;
for (target = targetList; target != NULL; target = target->next)
{
if (target->skip)
continue;
struct cdwQaEnrich *enrich = enrichFromOverlaps(ef, vf, assembly, target,
target->overlapBases, target->uniqOverlapBases);
cdwQaEnrichSaveToDb(conn, enrich, "cdwQaEnrich", 128);
cdwQaEnrichFree(&enrich);
}
bigWigValsOnChromFree(&valsOnChrom);
bbiChromInfoFreeList(&chromList);
bigWigFileClose(&bbi);
freez(&bigWigPath);
}
/* This old way is ~3 times as slow */
void doEnrichmentsFromBigWig(struct sqlConnection *conn,
struct cdwFile *ef, struct cdwValidFile *vf,
struct cdwAssembly *assembly, struct target *targetList)
/* Figure out enrichments from a bigBed file. */
{
/* Get path to bigBed, open it, and read all chromosomes. */
char *bigWigPath = cdwPathForFileId(conn, ef->id);
struct bbiFile *bbi = bigWigFileOpen(bigWigPath);
struct bbiChromInfo *chrom, *chromList = bbiChromList(bbi);
/* This takes a while, so let's figure out what parts take the time. */
long totalBigQueryTime = 0;
long totalOverlapTime = 0;
/* Do a pretty complex loop that just aims to set target->overlapBases and ->uniqOverlapBases
* for all targets. This is complicated by just wanting to keep one chromosome worth of
* bigWig data in memory. Also just for performance we do a lookup of target range tree to
* get chromosome specific one to use, which avoids a hash lookup in the inner loop. */
for (chrom = chromList; chrom != NULL; chrom = chrom->next)
{
/* Get list of intervals in bigWig for this chromosome, and feed it to a rangeTree. */
struct lm *lm = lmInit(0);
long startBigQueryTime = clock1000();
struct bbiInterval *ivList = bigWigIntervalQuery(bbi, chrom->name, 0, chrom->size, lm);
long endBigQueryTime = clock1000();
totalBigQueryTime += endBigQueryTime - startBigQueryTime;
struct bbiInterval *iv;
/* Loop through all targets adding overlaps from ivList */
long startOverlapTime = clock1000();
struct target *target;
for (target = targetList; target != NULL; target = target->next)
{
struct genomeRangeTree *grt = target->grt;
struct rbTree *targetTree = genomeRangeTreeFindRangeTree(grt, chrom->name);
if (targetTree != NULL)
{
for (iv = ivList; iv != NULL; iv = iv->next)
{
int overlap = rangeTreeOverlapSize(targetTree, iv->start, iv->end);
target->uniqOverlapBases += overlap;
target->overlapBases += overlap * iv->val;
}
}
}
long endOverlapTime = clock1000();
totalOverlapTime += endOverlapTime - startOverlapTime;
lmCleanup(&lm);
}
verbose(1, "totalBig %0.3f, totalOverlap %0.3f\n", 0.001*totalBigQueryTime, 0.001*totalOverlapTime);
/* Now loop through targets and save enrichment info to database */
struct target *target;
for (target = targetList; target != NULL; target = target->next)
{
struct cdwQaEnrich *enrich = enrichFromOverlaps(ef, vf, assembly, target,
target->overlapBases, target->uniqOverlapBases);
cdwQaEnrichSaveToDb(conn, enrich, "cdwQaEnrich", 128);
cdwQaEnrichFree(&enrich);
}
bbiChromInfoFreeList(&chromList);
bigWigFileClose(&bbi);
freez(&bigWigPath);
}
void doEnrichmentsFromBigBed(struct sqlConnection *conn,
struct cdwFile *ef, struct cdwValidFile *vf,
struct cdwAssembly *assembly, struct target *targetList)
/* Figure out enrichments from a bigBed file. */
{
/* Get path to bigBed, open it, and read all chromosomes. */
char *bigBedPath = cdwPathForFileId(conn, ef->id);
struct bbiFile *bbi = bigBedFileOpen(bigBedPath);
struct bbiChromInfo *chrom, *chromList = bbiChromList(bbi);
/* Do a pretty complex loop that just aims to set target->overlapBases and ->uniqOverlapBases
* for all targets. This is complicated by just wanting to keep one chromosome worth of
* bigBed data in memory. */
for (chrom = chromList; chrom != NULL; chrom = chrom->next)
{
/* Get list of intervals in bigBed for this chromosome, and feed it to a rangeTree. */
struct lm *lm = lmInit(0);
struct bigBedInterval *ivList = bigBedIntervalQuery(bbi, chrom->name, 0, chrom->size, 0, lm);
struct bigBedInterval *iv;
struct rbTree *bbTree = rangeTreeNew();
for (iv = ivList; iv != NULL; iv = iv->next)
rangeTreeAdd(bbTree, iv->start, iv->end);
struct range *bbRange, *bbRangeList = rangeTreeList(bbTree);
/* Loop through all targets adding overlaps from ivList and unique overlaps from bbRangeList */
struct target *target;
for (target = targetList; target != NULL; target = target->next)
{
if (target->skip)
continue;
struct genomeRangeTree *grt = target->grt;
struct rbTree *targetTree = genomeRangeTreeFindRangeTree(grt, chrom->name);
if (targetTree != NULL)
{
struct bigBedInterval *iv;
for (iv = ivList; iv != NULL; iv = iv->next)
{
int overlap = rangeTreeOverlapSize(targetTree, iv->start, iv->end);
target->overlapBases += overlap;
}
for (bbRange = bbRangeList; bbRange != NULL; bbRange = bbRange->next)
{
int overlap = rangeTreeOverlapSize(targetTree, bbRange->start, bbRange->end);
target->uniqOverlapBases += overlap;
}
}
}
rangeTreeFree(&bbTree);
lmCleanup(&lm);
}
/* Now loop through targets and save enrichment info to database */
struct target *target;
for (target = targetList; target != NULL; target = target->next)
{
if (target->skip)
continue;
struct cdwQaEnrich *enrich = enrichFromOverlaps(ef, vf, assembly, target,
target->overlapBases, target->uniqOverlapBases);
cdwQaEnrichSaveToDb(conn, enrich, "cdwQaEnrich", 128);
cdwQaEnrichFree(&enrich);
}
bbiChromInfoFreeList(&chromList);
bigBedFileClose(&bbi);
freez(&bigBedPath);
}
void doBigBedReplicate(struct sqlConnection *conn, char *format, struct cdwAssembly *assembly,
struct cdwFile *elderEf, struct cdwValidFile *elderVf,
struct cdwFile *youngerEf, struct cdwValidFile *youngerVf)
/* Do correlation analysis between elder and younger and save result to
* a new cdwQaPairCorrelation record. Do this for a format where we have a bigBed file. */
{
/* If got both pairs, work is done already */
if (pairExists(conn, elderEf->id, youngerEf->id, "cdwQaPairSampleOverlap")
&& pairExists(conn, elderEf->id, youngerEf->id, "cdwQaPairCorrelation"))
return;
int numColIx = 0;
if (sameString(format, "narrowPeak") || sameString(format, "broadPeak"))
numColIx = 6; // signalVal
else
numColIx = 4; // score
numColIx -= 3; // Subtract off chrom/start/end
char *enrichedIn = elderVf->enrichedIn;
struct genomeRangeTree *targetGrt = NULL;
if (!isEmpty(enrichedIn) && !sameString(enrichedIn, "unknown"))
targetGrt = genomeRangeTreeForTarget(conn, assembly, enrichedIn);
/* Get open big bed files for both younger and older. */
char *elderPath = cdwPathForFileId(conn, elderEf->id);
char *youngerPath = cdwPathForFileId(conn, youngerEf->id);
struct bbiFile *elderBbi = bigBedFileOpen(elderPath);
struct bbiFile *youngerBbi = bigBedFileOpen(youngerPath);
/* Loop through a chromosome at a time adding to correlation, and at the end save result in r.*/
struct correlate *c = correlateNew(), *cInEnriched = correlateNew();
struct bbiChromInfo *chrom, *chromList = bbiChromList(elderBbi);
long long elderTotalSpan = 0, youngerTotalSpan = 0, overlapTotalSpan = 0;
for (chrom = chromList; chrom != NULL; chrom = chrom->next)
{
addBbCorrelations(chrom, targetGrt, elderBbi, youngerBbi, numColIx, c, cInEnriched,
&elderTotalSpan, &youngerTotalSpan, &overlapTotalSpan);
}
/* Make up correlation structure and save. */
if (!pairExists(conn, elderEf->id, youngerEf->id, "cdwQaPairCorrelation"))
{
struct cdwQaPairCorrelation *cor;
AllocVar(cor);
cor->elderFileId = elderVf->fileId;
cor->youngerFileId = youngerVf->fileId;
cor->pearsonOverall = correlateResult(c);
cor->pearsonInEnriched = correlateResult(cInEnriched);
cdwQaPairCorrelationSaveToDb(conn, cor, "cdwQaPairCorrelation", 128);
freez(&cor);
}
/* Also make up sample structure and save. */
if (!pairExists(conn, elderEf->id, youngerEf->id, "cdwQaPairSampleOverlap"))
{
struct cdwQaPairSampleOverlap *sam;
AllocVar(sam);
sam->elderFileId = elderVf->fileId;
sam->youngerFileId = youngerVf->fileId;
sam->elderSampleBases = elderTotalSpan;
sam->youngerSampleBases = youngerTotalSpan;
sam->sampleOverlapBases = overlapTotalSpan;
setSampleSampleEnrichment(sam, format, assembly, elderVf, youngerVf);
cdwQaPairSampleOverlapSaveToDb(conn, sam, "cdwQaPairSampleOverlap", 128);
freez(&sam);
}
genomeRangeTreeFree(&targetGrt);
correlateFree(&c);
bigBedFileClose(&youngerBbi);
bigBedFileClose(&elderBbi);
freez(&youngerPath);
freez(&elderPath);
}
