double calcNormScoreFactor(char *fileName, int scoreCol)
/* Figure out what to multiply things by to get a nice browser score (0-1000) */
{
struct lineFile *lf = lineFileOpen(fileName, TRUE);
char *row[scoreCol+1];
double sum = 0, sumSquares = 0;
int n = 0;
double minVal=0, maxVal=0;
int fieldCount;
while ((fieldCount = lineFileChop(lf, row)) != 0)
{
lineFileExpectAtLeast(lf, scoreCol+1, fieldCount);
double x = sqlDouble(row[scoreCol]);
if (n == 0)
minVal = maxVal = x;
if (x < minVal) minVal = x;
if (x > maxVal) maxVal = x;
sum += x;
sumSquares += x*x;
n += 1;
}
lineFileClose(&lf);
double std = calcStdFromSums(sum, sumSquares, n);
double mean = sum/n;
double highEnd = mean + std;
if (highEnd > maxVal) highEnd = maxVal;
return 1000.0/highEnd;
}
double twoStdsOverMean(struct bbiFile *bbi)
/* Figure out what is two standard deviations over mean for a bigWig file. This is
* an often useful threshold. */
{
struct bbiSummaryElement sum = bbiTotalSummary(bbi);
double mean = sum.sumData/sum.validCount;
double std = calcStdFromSums(sum.sumData, sum.sumSquares, sum.validCount);
return mean + 2*std;
}
double bigWigStd(struct bbiFile* bw)
/* return the mean value of a bigWig */
{
double na = NANUM;
struct bbiSummaryElement bbs = bbiTotalSummary(bw);
if (bbs.validCount == 0)
return na;
else
return calcStdFromSums(bbs.sumData, bbs.sumSquares, bbs.validCount);
}
boolean bbiSummaryArray(struct bbiFile *bbi, char *chrom, bits32 start, bits32 end,
BbiFetchIntervals fetchIntervals,
enum bbiSummaryType summaryType, int summarySize, double *summaryValues)
/* Fill in summaryValues with data from indicated chromosome range in bigWig file.
* Be sure to initialize summaryValues to a default value, which will not be touched
* for regions without data in file. (Generally you want the default value to either
* be 0.0 or nan("") depending on the application.) Returns FALSE if no data
* at that position. */
{
struct bbiSummaryElement *elements;
AllocArray(elements, summarySize);
boolean ret = bbiSummaryArrayExtended(bbi, chrom, start, end,
fetchIntervals, summarySize, elements);
if (ret)
{
int i;
double covFactor = (double)summarySize/(end - start);
for (i=0; i<summarySize; ++i)
{
struct bbiSummaryElement *el = &elements[i];
if (el->validCount > 0)
{
double val;
switch (summaryType)
{
case bbiSumMean:
val = el->sumData/el->validCount;
break;
case bbiSumMax:
val = el->maxVal;
break;
case bbiSumMin:
val = el->minVal;
break;
case bbiSumCoverage:
val = covFactor*el->validCount;
break;
case bbiSumStandardDeviation:
val = calcStdFromSums(el->sumData, el->sumSquares, el->validCount);
break;
default:
internalErr();
val = 0.0;
break;
}
summaryValues[i] = val;
}
}
}
freeMem(elements);
return ret;
}
void bbiIntervalStatsReport(struct bbiInterval *bbList, char *table,
char *chrom, bits32 start, bits32 end)
/* Write out little statistical report in HTML */
{
/* Loop through list and calculate some stats. */
bits64 iCount = 0;
bits64 iTotalSize = 0;
bits32 biggestSize = 0, smallestSize = BIGNUM;
struct bbiInterval *bb;
double sum = 0.0, sumSquares = 0.0;
double minVal = bbList->val, maxVal = bbList->val;
for (bb = bbList; bb != NULL; bb = bb->next)
{
iCount += 1;
bits32 size = bb->end - bb->start;
iTotalSize += size;
if (biggestSize < size)
biggestSize = size;
if (smallestSize > size)
smallestSize = size;
double val = bb->val;
sum += val;
sumSquares += val * val;
if (minVal > val)
minVal = val;
if (maxVal < val)
maxVal = val;
}
char num1Buf[64], num2Buf[64]; /* big enough for 2^64 (and then some) */
sprintLongWithCommas(num1Buf, iCount);
sprintLongWithCommas(num2Buf, iTotalSize);
bits32 winSize = end-start;
printf("<B>Statistics on:</B> %s <B>items covering</B> %s bases (%4.2f%% coverage)<BR>\n",
num1Buf, num2Buf, 100.0*iTotalSize/winSize);
printf("<B>Average item spans</B> %4.2f <B>bases.</B> ", (double)iTotalSize/iCount);
if (biggestSize != smallestSize)
{
sprintLongWithCommas(num1Buf, smallestSize);
sprintLongWithCommas(num2Buf, biggestSize);
printf("<B>Minimum span</B> %s <B>maximum span</B> %s", num1Buf, num2Buf);
}
printf("<BR>\n");
printf("<B>Average value</B> %g <B>min</B> %g <B>max</B> %g <B> standard deviation </B> %g<BR>\n",
sum/iCount, minVal, maxVal, calcStdFromSums(sum, sumSquares, iCount));
}
void printStats(FILE *f, struct slDouble *list)
/* Print out stats on list: ave +-std min 1/4 median 3/4 max */
{
int count = 0;
struct slDouble *el;
double sum=0, sumSquared=0;
for (el = list; el != NULL; el = el->next)
{
sum += el->val;
sumSquared += el->val * el->val;
count += 1;
}
double ave = sum/count;
double std = calcStdFromSums(sum, sumSquared, count);
double minVal, q1, median, q3, maxVal;
slDoubleBoxWhiskerCalc(list, &minVal, &q1, &median, &q3, &maxVal);
fprintf(f, "\t%g+-%g [%g %g %g %g %g]", ave, std, minVal, q1, median, q3, maxVal);
}
void aveNoQuartiles(char *fileName)
/* aveNoQuartiles - Compute only min,max,mean,stdDev no quartiles */
{
bits64 count = 0;
struct lineFile *lf = lineFileOpen(fileName, TRUE);
char *words[128], *word;
int wordCount;
int wordIx = col-1;
double sumData = 0.0, sumSquares = 0.0;
double minVal = DBL_MAX, maxVal = -DBL_MAX;
while ((wordCount = lineFileChop(lf, words)) > 0)
{
word = words[wordIx];
if (word[0] == '-' || isdigit(word[0]))
{
double val = sqlDouble(word);
if (minVal > val) minVal = val;
if (maxVal < val) maxVal = val;
sumData += val;
sumSquares += val * val;
++count;
}
}
if (count == 0)
errAbort("No numerical data column %d of %s", col, fileName);
double average = sumData/count;
double stdDev = calcStdFromSums(sumData, sumSquares, count);
if (tableOut)
{
printf("# min max mean N sum stddev\n");
printf("%g %g %g %llu %g %g\n",
minVal, maxVal, average, count, sumData, stdDev);
}
else
{
printf("average %f\n", average);
printf("min %f\n", minVal);
printf("max %f\n", maxVal);
printf("count %llu\n", count);
printf("total %f\n", sumData);
printf("standard deviation %f\n", stdDev);
}
}
void bigBedDrawDense(struct track *tg, int seqStart, int seqEnd,
struct hvGfx *hvg, int xOff, int yOff, int width,
MgFont *font, Color color)
/* Use big-bed summary data to quickly draw bigBed. */
{
struct bbiSummaryElement *summary = tg->summary;
if (summary)
{
char *denseCoverage = trackDbSettingClosestToHome(tg->tdb, "denseCoverage");
if (denseCoverage != NULL)
{
double startVal = 0, endVal = atof(denseCoverage);
if (endVal <= 0)
{
struct bbiSummaryElement sumAll = *tg->sumAll;
double mean = sumAll.sumData/sumAll.validCount;
double std = calcStdFromSums(sumAll.sumData, sumAll.sumSquares, sumAll.validCount);
rangeFromMinMaxMeanStd(0, sumAll.maxVal, mean, std, &startVal, &endVal);
}
int x;
for (x=0; x<width; ++x)
{
if (summary[x].validCount > 0)
{
Color color = shadesOfGray[grayInRange(summary[x].maxVal, startVal, endVal)];
hvGfxBox(hvg, x+xOff, yOff, 1, tg->heightPer, color);
}
}
}
else
{
int x;
for (x=0; x<width; ++x)
{
if (summary[x].validCount > 0)
{
hvGfxBox(hvg, x+xOff, yOff, 1, tg->heightPer, color);
}
}
}
}
freez(&tg->summary);
}
void bigWigInfo(char *fileName)
/* bigWigInfo - Print out information about bigWig file.. */
{
struct bbiFile *bwf = bigWigFileOpen(fileName);
if (optionExists("minMax"))
{
struct bbiSummaryElement sum = bbiTotalSummary(bwf);
printf("%f %f\n", sum.minVal, sum.maxVal);
return;
}
printf("version: %d\n", bwf->version);
printf("isCompressed: %s\n", (bwf->uncompressBufSize > 0 ? "yes" : "no"));
printf("isSwapped: %d\n", bwf->isSwapped);
printLabelAndLongNumber("primaryDataSize", bwf->unzoomedIndexOffset - bwf->unzoomedDataOffset);
if (bwf->levelList != NULL)
{
long long indexEnd = bwf->levelList->dataOffset;
printLabelAndLongNumber("primaryIndexSize", indexEnd - bwf->unzoomedIndexOffset);
}
printf("zoomLevels: %d\n", bwf->zoomLevels);
if (optionExists("zooms"))
{
struct bbiZoomLevel *zoom;
for (zoom = bwf->levelList; zoom != NULL; zoom = zoom->next)
printf("\t%d\t%d\n", zoom->reductionLevel, (int)(zoom->indexOffset - zoom->dataOffset));
}
struct bbiChromInfo *chrom, *chromList = bbiChromList(bwf);
printf("chromCount: %d\n", slCount(chromList));
if (optionExists("chroms"))
for (chrom=chromList; chrom != NULL; chrom = chrom->next)
printf("\t%s %d %d\n", chrom->name, chrom->id, chrom->size);
struct bbiSummaryElement sum = bbiTotalSummary(bwf);
printLabelAndLongNumber("basesCovered", sum.validCount);
printf("mean: %f\n", sum.sumData/sum.validCount);
printf("min: %f\n", sum.minVal);
printf("max: %f\n", sum.maxVal);
printf("std: %f\n", calcStdFromSums(sum.sumData, sum.sumSquares, sum.validCount));
}
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 doSummaryStatsBigWig(struct sqlConnection *conn)
/* Put up page showing summary stats for bigWig track. */
{
struct trackDb *track = curTrack;
char *table = curTable;
char *shortLabel = (track == NULL ? table : track->shortLabel);
char *fileName = bigWigFileName(table, conn);
long startTime = clock1000();
htmlOpen("%s (%s) Big Wig Summary Statistics", shortLabel, table);
if (anySubtrackMerge(database, curTable))
hPrintf("<P><EM><B>Note:</B> subtrack merge is currently ignored on this "
"page (not implemented yet). Statistics shown here are only for "
"the primary table %s (%s).</EM>", shortLabel, table);
struct bbiFile *bwf = bigWigFileOpen(fileName);
struct region *region, *regionList = getRegions();
double sumData = 0, sumSquares = 0, minVal = 0, maxVal = 0;
bits64 validCount = 0;
if (!anyFilter() && !anyIntersection())
{
for (region = regionList; region != NULL; region = region->next)
{
struct bbiSummaryElement sum;
if (bbiSummaryArrayExtended(bwf, region->chrom, region->start, region->end,
bigWigIntervalQuery, 1, &sum))
{
if (validCount == 0)
{
minVal = sum.minVal;
maxVal = sum.maxVal;
}
else
{
if (sum.minVal < minVal)
minVal = sum.minVal;
if (sum.maxVal > maxVal)
maxVal = sum.maxVal;
}
sumData += sum.sumData;
sumSquares += sum.sumSquares;
validCount += sum.validCount;
}
}
}
else
{
double ll, ul;
enum wigCompare cmp;
getWigFilter(database, curTable, &cmp, &ll, &ul);
for (region = regionList; region != NULL; region = region->next)
{
struct lm *lm = lmInit(0);
struct bbiInterval *iv, *ivList;
ivList = intersectedFilteredBbiIntervalsOnRegion(conn, bwf, region, cmp, ll, ul, lm);
for (iv = ivList; iv != NULL; iv = iv->next)
{
double val = iv->val;
double size = iv->end - iv->start;
if (validCount == 0)
minVal = maxVal = val;
else
{
if (val < minVal)
minVal = val;
if (val > maxVal)
maxVal = val;
}
sumData += size*val;
sumSquares += size*val*val;
validCount += size;
}
lmCleanup(&lm);
}
}
hTableStart();
floatStatRow("mean", sumData/validCount);
floatStatRow("min", minVal);
floatStatRow("max", maxVal);
floatStatRow("standard deviation", calcStdFromSums(sumData, sumSquares, validCount));
numberStatRow("bases with data", validCount);
long long regionSize = basesInRegion(regionList,0);
long long gapTotal = gapsInRegion(conn, regionList,0);
numberStatRow("bases with sequence", regionSize - gapTotal);
numberStatRow("bases in region", regionSize);
wigFilterStatRow(conn);
stringStatRow("intersection", cartUsualString(cart, hgtaIntersectTable, "off"));
long wigFetchTime = clock1000() - startTime;
floatStatRow("load and calc time", 0.001*wigFetchTime);
hTableEnd();
bbiFileClose(&bwf);
htmlClose();
}
