void doBigWigReplicate(struct sqlConnection *conn, struct edwAssembly *assembly,
struct edwFile *elderEf, struct edwValidFile *elderVf,
struct edwFile *youngerEf, struct edwValidFile *youngerVf)
/* Do correlation analysis between elder and younger and save result to
* a new edwQaPairCorrelation record. Do this for a format where we have a bigWig file. */
{
if (pairExists(conn, elderEf->id, youngerEf->id, "edwQaPairCorrelation"))
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 = edwPathForFileId(conn, elderEf->id);
char *youngerPath = edwPathForFileId(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 edwQaPairCorrelation *cor;
AllocVar(cor);
cor->elderFileId = elderVf->fileId;
cor->youngerFileId = youngerVf->fileId;
cor->pearsonOverall = correlateResult(c);
cor->pearsonInEnriched = correlateResult(cInEnriched);
cor->pearsonClipped = correlateResult(cClipped);
edwQaPairCorrelationSaveToDb(conn, cor, "edwQaPairCorrelation", 128);
bigWigValsOnChromFree(&bVals);
bigWigValsOnChromFree(&aVals);
genomeRangeTreeFree(&targetGrt);
freez(&cor);
correlateFree(&c);
bigWigFileClose(&youngerBbi);
bigWigFileClose(&elderBbi);
freez(&youngerPath);
freez(&elderPath);
}
}
void metaBigClose(struct metaBig** pMb)
/* close the file and free up everything. */
{
struct metaBig* mb = *pMb;
hashFree(&mb->chromSizeHash);
if (mb->rgList)
hashFree(&mb->rgList);
if (mb->sections)
bedFreeList(&mb->sections);
if (mb->originalFileName)
freeMem(mb->originalFileName);
if (mb->fileName)
freeMem(mb->fileName);
if (mb->baseFileName)
freeMem(mb->baseFileName);
if (mb->remoteSiteAndDir)
freeMem(mb->remoteSiteAndDir);
#ifdef USE_HTSLIB
if (mb->idx)
hts_idx_destroy(mb->idx);
#endif
if (mb->type == isaBigBed)
bigBedFileClose(&mb->big.bbi);
#ifdef USE_HTSLIB
else if (mb->type == isaBam)
sam_close(mb->big.bam);
#endif
else
bigWigFileClose(&mb->big.bbi);
#ifdef USE_HTSLIB
if (mb->header)
bam_hdr_destroy(mb->header);
#endif
freez(pMb);
}
static void asbwClose(struct annoStreamer **pVSelf)
/* Close bbi handle and free self. */
{
if (pVSelf == NULL)
return;
struct annoStreamBigWig *self = *(struct annoStreamBigWig **)pVSelf;
bigWigFileClose(&(self->bbi));
self->intervalList = NULL;
lmCleanup(&(self->intervalQueryLm));
annoStreamerFree(pVSelf);
}
char *printBigWigViewInfo(FILE *f, char *indent, struct view *view,
struct composite *comp, struct taggedFile *tfList)
/* Print out info for a bigWig view, including subtracks. */
{
/* Look at all tracks in this view and calculate overall limits. */
double sumOfSums = 0, sumOfSumSquares = 0;
bits64 sumOfN = 0;
struct taggedFile *tf;
for (tf = tfList; tf != NULL; tf = tf->next)
{
if (sameString(tf->manifest->outputType, view->name))
{
char *relativeName = tf->manifest->fileName;
char *path = relativeName;
struct bbiFile *bbi = bigWigFileOpen(path);
struct bbiSummaryElement sum = bbiTotalSummary(bbi);
sumOfSums += sum.sumData;
sumOfSumSquares += sum.sumSquares;
sumOfN = sum.validCount;
bigWigFileClose(&bbi);
}
}
double mean = sumOfSums/sumOfN;
double std = calcStdFromSums(sumOfSums, sumOfSumSquares, sumOfN);
double clipMax = mean + 6*std;
/* Output view stanza. */
char type[64];
safef(type, sizeof(type), "bigWig %g %g", 0.0, clipMax);
fprintf(f, "%stype %s\n", indent, type);
fprintf(f, "%sviewLimits 0:%g\n", indent, clipMax);
fprintf(f, "%sminLimit 0\n", indent);
fprintf(f, "%smaxLimit %g\n", indent, clipMax);
fprintf(f, "%sautoScale off\n", indent);
fprintf(f, "%smaxHeightPixels 100:32:16\n", indent);
fprintf(f, "%swindowingFunction mean+whiskers\n", indent);
return cloneString(type);
}
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);
}
