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 bigWigCorrelate(char *aFileName, char *bFileName)
/* bigWigCorrelate - Correlate bigWig files, optionally only on target regions.. */
{
struct genomeRangeTree *targetGrt = NULL;
if (restrictFile)
targetGrt = grtFromBigBed(restrictFile);
struct bbiFile *aBbi = bigWigFileOpen(aFileName);
struct bbiFile *bBbi = bigWigFileOpen(bFileName);
struct correlate *c = correlateNew();
struct bbiChromInfo *chrom, *chromList = bbiChromList(aBbi);
struct bigWigValsOnChrom *aVals = bigWigValsOnChromNew();
struct bigWigValsOnChrom *bVals = bigWigValsOnChromNew();
for (chrom = chromList; chrom != NULL; chrom = chrom->next)
{
addBwCorrelations(chrom, targetGrt, aVals, bVals, aBbi, bBbi, threshold, threshold, c);
}
printf("%g\n", correlateResult(c));
}
void averageFetchingEachChrom(struct bbiFile *bbi, struct bed **pBedList, int fieldCount,
FILE *f, FILE *bedF)
/* Do the averaging by sorting bedList by chromosome, and then processing each chromosome
* at once. Faster for long bedLists. */
{
/* Sort by chromosome. */
slSort(pBedList, bedCmpChrom);
struct bigWigValsOnChrom *chromVals = bigWigValsOnChromNew();
struct bed *bed, *bedList, *nextChrom;
verbose(1, "processing chromosomes");
for (bedList = *pBedList; bedList != NULL; bedList = nextChrom)
{
/* Figure out which chromosome we're working on, and the last bed using it. */
char *chrom = bedList->chrom;
nextChrom = nextChromInList(bedList);
verbose(2, "Processing %s\n", chrom);
if (bigWigValsOnChromFetchData(chromVals, chrom, bbi))
{
double *valBuf = chromVals->valBuf;
Bits *covBuf = chromVals->covBuf;
/* Loop through beds doing sums and outputting. */
for (bed = bedList; bed != nextChrom; bed = bed->next)
{
int size = 0, coverage = 0;
double sum = 0.0;
if (sampleAroundCenter > 0)
{
int center = (bed->chromStart + bed->chromEnd)/2;
int left = center - (sampleAroundCenter/2);
addBufIntervalInfo(valBuf, covBuf, left, left+sampleAroundCenter,
&size, &coverage, &sum);
}
else
{
if (fieldCount < 12)
{
addBufIntervalInfo(valBuf, covBuf, bed->chromStart, bed->chromEnd,
&size, &coverage, &sum);
}
else
{
int i;
for (i=0; i<bed->blockCount; ++i)
{
int start = bed->chromStart + bed->chromStarts[i];
int end = start + bed->blockSizes[i];
addBufIntervalInfo(valBuf, covBuf, start, end, &size, &coverage, &sum);
}
}
}
/* Print out result, fudging mean to 0 if no coverage at all. */
double mean = 0;
if (coverage > 0)
mean = sum/coverage;
fprintf(f, "%s\t%d\t%d\t%g\t%g\t%g\n", bed->name, size, coverage, sum, sum/size, mean);
optionallyPrintBedPlus(bedF, bed, fieldCount, mean);
}
verboseDot();
}
else
{
/* If no bigWig data on this chromosome, just output as if coverage is 0 */
for (bed = bedList; bed != nextChrom; bed = bed->next)
{
fprintf(f, "%s\t%d\t0\t0\t0\t0\n", bed->name, bedTotalBlockSize(bed));
optionallyPrintBedPlus(bedF, bed, fieldCount, 0);
}
}
}
verbose(1, "\n");
}
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);
}
void regCompanionChia(char *inBedPairs, char *output)
/* regCompanionChia - Analyse chia pet data against promoters and enhancers.. */
{
inArray[0].fileName = inBedPairs;
checkInputOpenFiles(inArray, ArraySize(inArray));
verbose(1, "Opened all %d inputs successfully\n", (int)ArraySize(inArray));
struct bed *chiaList = bedLoadTwoBlocks(inArray[0].lf);
slSort(&chiaList, bedCmp);
int chiaCount = slCount(chiaList);
struct bigWigValsOnChrom *chromVals = bigWigValsOnChromNew();
int inIx;
for (inIx=1; inIx < ArraySize(inArray); ++inIx)
{
/* Allocate output arrays. */
struct inInfo *in = &inArray[inIx];
double *out1 = AllocArray(in->out[0], chiaCount);
double *out2 = AllocArray(in->out[1], chiaCount);
/* Process input depending on type. */
verbose(1, "Processing %s", in->fileName);
switch(in->type)
{
case itPromoterBed:
doItPromoterBed(in, chiaList, out1, out2);
break;
case itUnstrandedBed:
doItUnstrandedBed(in, chiaList, out1, out2);
break;
case itBigWig:
doItBigWig(in, chiaList, chromVals, out1, out2);
break;
default:
internalErr();
break;
}
verbose(1, "\n");
}
/* do output */
FILE *f = mustOpen(output, "w");
struct bed *chia;
int chiaIx = 0;
for (chia = chiaList; chia != NULL; chia = chia->next, ++chiaIx)
{
// fprintf(f, "%s\t%d\t%d\tchia%d\t%d", chia->chrom, chia->chromStart, chia->chromEnd, chiaIx+1, chia->score);
fprintf(f, "%s\t%d\t%d\t%s\t%d", chia->chrom, chia->chromStart, chia->chromEnd, chia->name, chia->score);
int blockIx;
for (blockIx=0; blockIx < 2; ++blockIx)
{
fprintf(f, "\tblock%d\t%d", blockIx+1, chia->blockSizes[blockIx]);
for (inIx=1; inIx < ArraySize(inArray); ++inIx)
{
struct inInfo *in = &inArray[inIx];
double *out = in->out[blockIx];
fprintf(f, "\t%g", out[chiaIx]);
}
}
fprintf(f, "\n");
}
carefulClose(&f);
}
