struct bed *findCompatible(struct bed *newBed, struct hash *oldHash, struct hash *usedHash)
/* Try and find an old bed compatible with new bed. */
{
struct binKeeper *bk = hashFindVal(oldHash, newBed->chrom);
int bestDiff = BIGNUM;
struct bed *bestBed = NULL;
if (bk == NULL)
return NULL;
struct binElement *bin, *binList = binKeeperFind(bk, newBed->chromStart, newBed->chromEnd);
for (bin = binList; bin != NULL; bin = bin->next)
{
struct bed *oldBed = bin->val;
if (oldBed->strand[0] == newBed->strand[0])
{
if (!hashLookup(usedHash, oldBed->name))
{
if (bedCompatibleExtension(oldBed, newBed) || endUtrChangeOnly(oldBed, newBed))
{
int diff = bedTotalBlockSize(oldBed) - bedTotalBlockSize(newBed);
if (diff < 0) diff = -diff;
if (diff < bestDiff)
{
bestDiff = diff;
bestBed = oldBed;
}
}
}
}
}
slFreeList(&binList);
return bestBed;
}
double bedOverlapRatio(struct bed *a, struct bed *b)
/* Return TRUE if a overlaps b by at least given ratio. */
{
int aSize = bedTotalBlockSize(a);
int bSize = bedTotalBlockSize(b);
int overlapSize = bedSameStrandOverlap(a,b);
if (aSize == bSize && aSize == overlapSize)
return 1.0; /* Avoid rounding here. */
double x = overlapSize;
double aCov = x / aSize;
double bCov = x / bSize;
return min(aCov, bCov);
}
struct dnaSeq *twoBitAndBedToSeq(struct twoBitFile *tbf, struct bed *bed)
/* Get sequence defined by bed. Exclude introns. */
{
struct dnaSeq *seq;
if (bed->blockCount <= 1)
{
seq = twoBitReadSeqFrag(tbf, bed->chrom, bed->chromStart, bed->chromEnd);
freeMem(seq->name);
seq->name = cloneString(bed->name);
}
else
{
int totalBlockSize = bedTotalBlockSize(bed);
AllocVar(seq);
seq->name = cloneString(bed->name);
seq->dna = needMem(totalBlockSize+1);
seq->size = totalBlockSize;
int i;
int seqOffset = 0;
for (i=0; i<bed->blockCount; ++i)
{
int exonSize = bed->blockSizes[i];
int exonStart = bed->chromStart + bed->chromStarts[i];
struct dnaSeq *exon = twoBitReadSeqFrag(tbf, bed->chrom, exonStart, exonStart+exonSize);
memcpy(seq->dna + seqOffset, exon->dna, exonSize);
seqOffset += exonSize;
dnaSeqFree(&exon);
}
}
if (bed->strand[0] == '-')
reverseComplement(seq->dna, seq->size);
return seq;
}
int findLastIntronPos(struct hash *bedHash, char *name)
/* Find last intron position in RNA coordinates if we have
* a bed for this mRNA. Otherwise (or if it's single exon)
* return 0. */
{
struct bed *bed = hashFindVal(bedHash, name);
if (bed == NULL)
return 0;
if (bed->blockCount < 2)
return 0;
int rnaSize = bedTotalBlockSize(bed);
if (bed->strand[0] == '+')
return rnaSize - bed->blockSizes[bed->blockCount-1];
else
return rnaSize - bed->blockSizes[0];
}
static struct psl *bedToPsl(struct bed *bed, struct hash *chromSizes)
/* Convert a single bed to a PSL. */
{
int qSize = bedTotalBlockSize(bed);
struct psl *psl;
if (keepQuery)
psl = pslNew(bed->chrom, hashIntVal(chromSizes, bed->chrom), bed->chromStart, bed->chromEnd,
bed->chrom, hashIntVal(chromSizes, bed->chrom), bed->chromStart, bed->chromEnd,
((bed->strand[0] == '\0') ? "+" : bed->strand), (bed->blockCount == 0) ? 1 : bed->blockCount, 0);
else
psl = pslNew(bed->name, qSize, 0, qSize,
bed->chrom, hashIntVal(chromSizes, bed->chrom), bed->chromStart, bed->chromEnd,
((bed->strand[0] == '\0') ? "+" : bed->strand), (bed->blockCount == 0) ? 1 : bed->blockCount, 0);
psl->match = psl->qSize;
if (bed->blockCount == 0)
bedToPsl4(bed, psl);
else
bedToPsl12(bed, psl);
return psl;
}
void txInfoAssemble(char *txBedFile, char *cdsEvFile, char *txCdsPredictFile, char *altSpliceFile,
char *exceptionFile, char *sizePolyAFile, char *pslFile, char *flipFile, char *outFile)
/* txInfoAssemble - Assemble information from various sources into txInfo table.. */
{
/* Build up hash of evidence keyed by transcript name. */
struct hash *cdsEvHash = hashNew(18);
struct cdsEvidence *cdsEv, *cdsEvList = cdsEvidenceLoadAll(cdsEvFile);
for (cdsEv = cdsEvList; cdsEv != NULL; cdsEv = cdsEv->next)
hashAddUnique(cdsEvHash, cdsEv->name, cdsEv);
verbose(2, "Loaded %d elements from %s\n", cdsEvHash->elCount, cdsEvFile);
/* Build up hash of bestorf structures keyed by transcript name */
struct hash *predictHash = hashNew(18);
struct cdsEvidence *predict, *predictList = cdsEvidenceLoadAll(txCdsPredictFile);
for (predict = predictList; predict != NULL; predict = predict->next)
hashAddUnique(predictHash, predict->name, predict);
verbose(2, "Loaded %d predicts from %s\n", predictHash->elCount, txCdsPredictFile);
/* Build up structure for random access of retained introns */
struct bed *altSpliceList = bedLoadNAll(altSpliceFile, 6);
verbose(2, "Loaded %d alts from %s\n", slCount(altSpliceList), altSpliceFile);
struct hash *altSpliceHash = bedsIntoHashOfKeepers(altSpliceList);
/* Read in exception info. */
struct hash *selenocysteineHash, *altStartHash;
genbankExceptionsHash(exceptionFile, &selenocysteineHash, &altStartHash);
/* Read in polyA sizes */
struct hash *sizePolyAHash = hashNameIntFile(sizePolyAFile);
verbose(2, "Loaded %d from %s\n", sizePolyAHash->elCount, sizePolyAFile);
/* Read in psls */
struct hash *pslHash = hashNew(20);
struct psl *psl, *pslList = pslLoadAll(pslFile);
for (psl = pslList; psl != NULL; psl = psl->next)
hashAdd(pslHash, psl->qName, psl);
verbose(2, "Loaded %d from %s\n", pslHash->elCount, pslFile);
/* Read in accessions that we flipped for better splice sites. */
struct hash *flipHash = hashWordsInFile(flipFile, 0);
/* Open primary gene input and output. */
struct lineFile *lf = lineFileOpen(txBedFile, TRUE);
FILE *f = mustOpen(outFile, "w");
/* Main loop - process each gene */
char *row[12];
while (lineFileRow(lf, row))
{
struct bed *bed = bedLoad12(row);
verbose(3, "Processing %s\n", bed->name);
/* Initialize info to zero */
struct txInfo info;
ZeroVar(&info);
/* Figure out name, sourceAcc, and isRefSeq from bed->name */
info.name = bed->name;
info.category = "n/a";
if (isRfam(bed->name) || stringIn("tRNA", bed->name) != NULL)
{
info.sourceAcc = cloneString(bed->name);
}
else
{
info.sourceAcc = txAccFromTempName(bed->name);
}
info.isRefSeq = startsWith("NM_", info.sourceAcc);
if (startsWith("antibody.", info.sourceAcc)
|| startsWith("CCDS", info.sourceAcc) || isRfam(info.sourceAcc)
|| stringIn("tRNA", info.sourceAcc) != NULL)
{
/* Fake up some things for antibody frag and CCDS that don't have alignments. */
info.sourceSize = bedTotalBlockSize(bed);
info.aliCoverage = 1.0;
info.aliIdRatio = 1.0;
info. genoMapCount = 1;
}
else
{
/* Loop through all psl's associated with our RNA. Figure out
* our overlap with each, and pick best one. */
struct hashEl *hel, *firstPslHel = hashLookup(pslHash, info.sourceAcc);
if (firstPslHel == NULL)
errAbort("%s is not in %s", info.sourceAcc, pslFile);
int mapCount = 0;
struct psl *psl, *bestPsl = NULL;
int coverage, bestCoverage = 0;
boolean isFlipped = (hashLookup(flipHash, info.sourceAcc) != NULL);
for (hel = firstPslHel; hel != NULL; hel = hashLookupNext(hel))
{
psl = hel->val;
mapCount += 1;
coverage = pslBedOverlap(psl, bed);
if (coverage > bestCoverage)
{
bestCoverage = coverage;
bestPsl = psl;
}
/* If we flipped it, try it on the opposite strand too. */
if (isFlipped)
{
psl->strand[0] = (psl->strand[0] == '+' ? '-' : '+');
coverage = pslBedOverlap(psl, bed);
if (coverage > bestCoverage)
{
bestCoverage = coverage;
bestPsl = psl;
}
psl->strand[0] = (psl->strand[0] == '+' ? '-' : '+');
}
}
if (bestPsl == NULL)
errAbort("%s has no overlapping alignments with %s in %s",
bed->name, info.sourceAcc, pslFile);
/* Figure out and save alignment statistics. */
int polyA = hashIntValDefault(sizePolyAHash, bed->name, 0);
info.sourceSize = bestPsl->qSize - polyA;
info.aliCoverage = (double)bestCoverage / info.sourceSize;
info.aliIdRatio = (double)(bestPsl->match + bestPsl->repMatch)/
(bestPsl->match + bestPsl->misMatch + bestPsl->repMatch);
info. genoMapCount = mapCount;
}
/* Get orf size and start/end complete from cdsEv. */
if (bed->thickStart < bed->thickEnd)
{
cdsEv = hashFindVal(cdsEvHash, bed->name);
if (cdsEv != NULL)
{
info.orfSize = cdsEv->end - cdsEv->start;
info.startComplete = cdsEv->startComplete;
info.endComplete = cdsEv->endComplete;
}
}
/* Get score from prediction. */
predict = hashFindVal(predictHash, bed->name);
if (predict != NULL)
info.cdsScore = predict->score;
/* Figure out nonsense-mediated-decay from bed itself. */
info.nonsenseMediatedDecay = isNonsenseMediatedDecayTarget(bed);
/* Figure out if retained intron from bed and alt-splice keeper hash */
info.retainedIntron = hasRetainedIntron(bed, altSpliceHash);
info.strangeSplice = countStrangeSplices(bed, altSpliceHash);
info.atacIntrons = countAtacIntrons(bed, altSpliceHash);
info.bleedIntoIntron = addIntronBleed(bed, altSpliceHash);
/* Look up selenocysteine info. */
info.selenocysteine = (hashLookup(selenocysteineHash, bed->name) != NULL);
/* Loop through bed looking for small gaps indicative of frame shift/stop */
int i, lastBlock = bed->blockCount-1;
int exonCount = 1;
for (i=0; i < lastBlock; ++i)
{
int gapStart = bed->chromStarts[i] + bed->blockSizes[i];
int gapEnd = bed->chromStarts[i+1];
int gapSize = gapEnd - gapStart;
switch (gapSize)
{
case 1:
case 2:
info.genomicFrameShift = TRUE;
break;
case 3:
info.genomicStop = TRUE;
break;
default:
exonCount += 1;
break;
}
}
info.exonCount = exonCount;
/* Write info, free bed. */
txInfoTabOut(&info, f);
bedFree(&bed);
}
/* Clean up and go home. */
carefulClose(&f);
}
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");
}
int scoreNoncodingBed(struct bed *bed)
/* Score noncoding bed weighing number of exons and total size. */
{
return bed->blockCount*100 + bedTotalBlockSize(bed);
}
