/*
* This function calculates ct tag for two bams, it assumes they are from the same template and
* writes the tag to the first read in position terms.
*/
static void bam_template_cigar(bam1_t *b1, bam1_t *b2, kstring_t *str)
{
bam1_t *swap;
int i, end;
uint32_t *cigar;
str->l = 0;
if (b1->core.tid != b2->core.tid || b1->core.tid < 0 || b1->core.pos < 0 || b2->core.pos < 0 || b1->core.flag&BAM_FUNMAP || b2->core.flag&BAM_FUNMAP) return; // coordinateless or not on the same chr; skip
if (b1->core.pos > b2->core.pos) swap = b1, b1 = b2, b2 = swap; // make sure b1 has a smaller coordinate
kputc((b1->core.flag & BAM_FREAD1)? '1' : '2', str); // segment index
kputc((b1->core.flag & BAM_FREVERSE)? 'R' : 'F', str); // strand
for (i = 0, cigar = bam_get_cigar(b1); i < b1->core.n_cigar; ++i) {
kputw(bam_cigar_oplen(cigar[i]), str);
kputc(bam_cigar_opchr(cigar[i]), str);
}
end = bam_endpos(b1);
kputw(b2->core.pos - end, str);
kputc('T', str);
kputc((b2->core.flag & BAM_FREAD1)? '1' : '2', str); // segment index
kputc((b2->core.flag & BAM_FREVERSE)? 'R' : 'F', str); // strand
for (i = 0, cigar = bam_get_cigar(b2); i < b2->core.n_cigar; ++i) {
kputw(bam_cigar_oplen(cigar[i]), str);
kputc(bam_cigar_opchr(cigar[i]), str);
}
uint8_t* data;
if ((data = bam_aux_get(b1,"ct")) != NULL) bam_aux_del(b1, data);
if ((data = bam_aux_get(b2,"ct")) != NULL) bam_aux_del(b2, data);
bam_aux_append(b1, "ct", 'Z', str->l+1, (uint8_t*)str->s);
}
// Is it plausible that these reads are properly paired?
// Can't really give definitive answer without checking isize
static bool plausibly_properly_paired(bam1_t* a, bam1_t* b)
{
if ((a->core.flag & BAM_FUNMAP) || (b->core.flag & BAM_FUNMAP)) return false;
assert(a->core.tid >= 0); // This should never happen if FUNMAP is set correctly
if (a->core.tid != b->core.tid) return false;
bam1_t* first = a;
bam1_t* second = b;
int32_t a_pos = a->core.flag&BAM_FREVERSE ? bam_endpos(a) : a->core.pos;
int32_t b_pos = b->core.flag&BAM_FREVERSE ? bam_endpos(b) : b->core.pos;
if (a_pos > b_pos) {
first = b;
second = a;
} else {
first = a;
second = b;
}
if (!(first->core.flag&BAM_FREVERSE) && (second->core.flag&BAM_FREVERSE))
return true;
else
return false;
}
void SamWriterPrivate::Write(const BamRecord& record)
{
#if PBBAM_AUTOVALIDATE
Validator::Validate(record);
#endif
const auto rawRecord = internal::BamRecordMemory::GetRawData(record);
// store bin number
// min_shift=14 & n_lvls=5 are SAM/BAM "magic numbers"
rawRecord->core.bin = hts_reg2bin(rawRecord->core.pos,
bam_endpos(rawRecord.get()), 14, 5);
// write record to file
const int ret = sam_write1(file_.get(), header_.get(), rawRecord.get());
if (ret <= 0)
throw std::runtime_error("could not write record");
}
static int32_t unclipped_end(bam1_t *b) {
uint32_t *cigar = bam_get_cigar(b);
int32_t end_pos, clipped = 0;
int32_t i;
end_pos = bam_endpos(b);
// now get the clipped end bases (if any)
// if we get to the beginning of the cigar string
// without hitting a non-clip then the results are meaningless
for (i = b->core.n_cigar - 1; i >= 0; i--) {
char c = bam_cigar_opchr(cigar[i]);
if (c == 'S' || c == 'H') { // clips
clipped += bam_cigar_oplen(cigar[i]);
} else {
break;
}
}
return end_pos + clipped;
}
static int mplp_func(void *data, bam1_t *b)
{
char *ref;
mplp_aux_t *ma = (mplp_aux_t*)data;
int ret, ref_len;
while (1)
{
int has_ref;
ret = ma->iter? sam_itr_next(ma->fp, ma->iter, b) : sam_read1(ma->fp, ma->h, b);
if (ret < 0) break;
// The 'B' cigar operation is not part of the specification, considering as obsolete.
// bam_remove_B(b);
if (b->core.tid < 0 || (b->core.flag&BAM_FUNMAP)) continue; // exclude unmapped reads
if (ma->conf->rflag_require && !(ma->conf->rflag_require&b->core.flag)) continue;
if (ma->conf->rflag_filter && ma->conf->rflag_filter&b->core.flag) continue;
if (ma->conf->bed)
{
// test overlap
regitr_t *itr = ma->conf->bed_itr;
int beg = b->core.pos, end = bam_endpos(b)-1;
int overlap = regidx_overlap(ma->conf->bed, ma->h->target_name[b->core.tid],beg,end, itr);
if ( !ma->conf->bed_logic && !overlap )
{
// exclude only reads which are fully contained in the region
while ( regitr_overlap(itr) )
{
if ( beg < itr->beg ) { overlap = 1; break; }
if ( end > itr->end ) { overlap = 1; break; }
}
}
if ( !overlap ) continue;
}
if ( bam_smpl_get_sample_id(ma->conf->bsmpl,ma->bam_id,b)<0 ) continue;
if (ma->conf->flag & MPLP_ILLUMINA13) {
int i;
uint8_t *qual = bam_get_qual(b);
for (i = 0; i < b->core.l_qseq; ++i)
qual[i] = qual[i] > 31? qual[i] - 31 : 0;
}
if (ma->conf->fai && b->core.tid >= 0) {
has_ref = mplp_get_ref(ma, b->core.tid, &ref, &ref_len);
if (has_ref && ref_len <= b->core.pos) { // exclude reads outside of the reference sequence
fprintf(stderr,"[%s] Skipping because %d is outside of %d [ref:%d]\n",
__func__, b->core.pos, ref_len, b->core.tid);
continue;
}
} else {
has_ref = 0;
}
if (has_ref && (ma->conf->flag&MPLP_REALN)) sam_prob_realn(b, ref, ref_len, (ma->conf->flag & MPLP_REDO_BAQ)? 7 : 3);
if (has_ref && ma->conf->capQ_thres > 10) {
int q = sam_cap_mapq(b, ref, ref_len, ma->conf->capQ_thres);
if (q < 0) continue; // skip
else if (b->core.qual > q) b->core.qual = q;
}
if (b->core.qual < ma->conf->min_mq) continue;
else if ((ma->conf->flag&MPLP_NO_ORPHAN) && (b->core.flag&BAM_FPAIRED) && !(b->core.flag&BAM_FPROPER_PAIR)) continue;
return ret;
};
return ret;
}
// currently, this function ONLY works if each read has one hit
static void bam_mating_core(samFile* in, samFile* out, int remove_reads, int proper_pair_check, int add_ct)
{
bam_hdr_t *header;
bam1_t *b[2];
int curr, has_prev, pre_end = 0, cur_end = 0;
kstring_t str;
str.l = str.m = 0; str.s = 0;
header = sam_hdr_read(in);
if (header == NULL) {
fprintf(stderr, "[bam_mating_core] ERROR: Couldn't read header\n");
exit(1);
}
// Accept unknown, unsorted, or queryname sort order, but error on coordinate sorted.
if ((header->l_text > 3) && (strncmp(header->text, "@HD", 3) == 0)) {
char *p, *q;
p = strstr(header->text, "\tSO:coordinate");
q = strchr(header->text, '\n');
// Looking for SO:coordinate within the @HD line only
// (e.g. must ignore in a @CO comment line later in header)
if ((p != 0) && (p < q)) {
fprintf(stderr, "[bam_mating_core] ERROR: Coordinate sorted, require grouped/sorted by queryname.\n");
exit(1);
}
}
sam_hdr_write(out, header);
b[0] = bam_init1();
b[1] = bam_init1();
curr = 0; has_prev = 0;
while (sam_read1(in, header, b[curr]) >= 0) {
bam1_t *cur = b[curr], *pre = b[1-curr];
if (cur->core.flag & BAM_FSECONDARY)
{
if ( !remove_reads ) sam_write1(out, header, cur);
continue; // skip secondary alignments
}
if (cur->core.flag & BAM_FSUPPLEMENTARY)
{
sam_write1(out, header, cur);
continue; // pass supplementary alignments through unchanged (TODO:make them match read they came from)
}
if (cur->core.tid < 0 || cur->core.pos < 0) // If unmapped set the flag
{
cur->core.flag |= BAM_FUNMAP;
}
if ((cur->core.flag&BAM_FUNMAP) == 0) // If mapped calculate end
{
cur_end = bam_endpos(cur);
// Check cur_end isn't past the end of the contig we're on, if it is set the UNMAP'd flag
if (cur_end > (int)header->target_len[cur->core.tid]) cur->core.flag |= BAM_FUNMAP;
}
if (has_prev) { // do we have a pair of reads to examine?
if (strcmp(bam_get_qname(cur), bam_get_qname(pre)) == 0) { // identical pair name
pre->core.flag |= BAM_FPAIRED;
cur->core.flag |= BAM_FPAIRED;
sync_mate(pre, cur);
if (pre->core.tid == cur->core.tid && !(cur->core.flag&(BAM_FUNMAP|BAM_FMUNMAP))
&& !(pre->core.flag&(BAM_FUNMAP|BAM_FMUNMAP))) // if safe set TLEN/ISIZE
{
uint32_t cur5, pre5;
cur5 = (cur->core.flag&BAM_FREVERSE)? cur_end : cur->core.pos;
pre5 = (pre->core.flag&BAM_FREVERSE)? pre_end : pre->core.pos;
cur->core.isize = pre5 - cur5; pre->core.isize = cur5 - pre5;
} else cur->core.isize = pre->core.isize = 0;
if (add_ct) bam_template_cigar(pre, cur, &str);
// TODO: Add code to properly check if read is in a proper pair based on ISIZE distribution
if (proper_pair_check && !plausibly_properly_paired(pre,cur)) {
pre->core.flag &= ~BAM_FPROPER_PAIR;
cur->core.flag &= ~BAM_FPROPER_PAIR;
}
// Write out result
if ( !remove_reads ) {
sam_write1(out, header, pre);
sam_write1(out, header, cur);
} else {
// If we have to remove reads make sure we do it in a way that doesn't create orphans with bad flags
if(pre->core.flag&BAM_FUNMAP) cur->core.flag &= ~(BAM_FPAIRED|BAM_FMREVERSE|BAM_FPROPER_PAIR);
if(cur->core.flag&BAM_FUNMAP) pre->core.flag &= ~(BAM_FPAIRED|BAM_FMREVERSE|BAM_FPROPER_PAIR);
if(!(pre->core.flag&BAM_FUNMAP)) sam_write1(out, header, pre);
if(!(cur->core.flag&BAM_FUNMAP)) sam_write1(out, header, cur);
}
has_prev = 0;
} else { // unpaired? clear bad info and write it out
if (pre->core.tid < 0 || pre->core.pos < 0 || pre->core.flag&BAM_FUNMAP) { // If unmapped
pre->core.flag |= BAM_FUNMAP;
pre->core.tid = -1;
pre->core.pos = -1;
}
pre->core.mtid = -1; pre->core.mpos = -1; pre->core.isize = 0;
pre->core.flag &= ~(BAM_FPAIRED|BAM_FMREVERSE|BAM_FPROPER_PAIR);
if ( !remove_reads || !(pre->core.flag&BAM_FUNMAP) ) sam_write1(out, header, pre);
}
} else has_prev = 1;
curr = 1 - curr;
pre_end = cur_end;
}
if (has_prev && !remove_reads) { // If we still have a BAM in the buffer it must be unpaired
bam1_t *pre = b[1-curr];
if (pre->core.tid < 0 || pre->core.pos < 0 || pre->core.flag&BAM_FUNMAP) { // If unmapped
pre->core.flag |= BAM_FUNMAP;
pre->core.tid = -1;
pre->core.pos = -1;
}
pre->core.mtid = -1; pre->core.mpos = -1; pre->core.isize = 0;
pre->core.flag &= ~(BAM_FPAIRED|BAM_FMREVERSE|BAM_FPROPER_PAIR);
sam_write1(out, header, pre);
}
bam_hdr_destroy(header);
bam_destroy1(b[0]);
bam_destroy1(b[1]);
free(str.s);
}
// Returns 0 to indicate read should be output 1 otherwise
static int process_aln(const bam_hdr_t *h, bam1_t *b, samview_settings_t* settings)
{
if (settings->remove_B) bam_remove_B(b);
if (settings->min_qlen > 0) {
int k, qlen = 0;
uint32_t *cigar = bam_get_cigar(b);
for (k = 0; k < b->core.n_cigar; ++k)
if ((bam_cigar_type(bam_cigar_op(cigar[k]))&1) || bam_cigar_op(cigar[k]) == BAM_CHARD_CLIP)
qlen += bam_cigar_oplen(cigar[k]);
if (qlen < settings->min_qlen) return 1;
}
if (b->core.qual < settings->min_mapQ || ((b->core.flag & settings->flag_on) != settings->flag_on) || (b->core.flag & settings->flag_off))
return 1;
if (settings->bed && (b->core.tid < 0 || !bed_overlap(settings->bed, h->target_name[b->core.tid], b->core.pos, bam_endpos(b))))
return 1;
if (settings->subsam_frac > 0.) {
uint32_t k = __ac_Wang_hash(__ac_X31_hash_string(bam_get_qname(b)) ^ settings->subsam_seed);
if ((double)(k&0xffffff) / 0x1000000 >= settings->subsam_frac) return 1;
}
if (settings->rghash) {
uint8_t *s = bam_aux_get(b, "RG");
if (s) {
khint_t k = kh_get(rg, settings->rghash, (char*)(s + 1));
if (k == kh_end(settings->rghash)) return 1;
}
}
if (settings->library) {
const char *p = bam_get_library((bam_hdr_t*)h, b);
if (!p || strcmp(p, settings->library) != 0) return 1;
}
if (settings->remove_aux_len) {
size_t i;
for (i = 0; i < settings->remove_aux_len; ++i) {
uint8_t *s = bam_aux_get(b, settings->remove_aux[i]);
if (s) {
bam_aux_del(b, s);
}
}
}
return 0;
}
static int mplp_func(void *data, bam1_t *b)
{
extern int bam_realn(bam1_t *b, const char *ref);
extern int bam_prob_realn_core(bam1_t *b, const char *ref, int);
extern int bam_cap_mapQ(bam1_t *b, char *ref, int thres);
mplp_aux_t *ma = (mplp_aux_t*)data;
int ret, skip = 0;
do {
int has_ref;
ret = ma->iter? sam_itr_next(ma->fp, ma->iter, b) : sam_read1(ma->fp, ma->h, b);
if (ret < 0) break;
// The 'B' cigar operation is not part of the specification, considering as obsolete.
// bam_remove_B(b);
if (b->core.tid < 0 || (b->core.flag&BAM_FUNMAP)) { // exclude unmapped reads
skip = 1;
continue;
}
if (ma->conf->rflag_require && !(ma->conf->rflag_require&b->core.flag)) { skip = 1; continue; }
if (ma->conf->rflag_filter && ma->conf->rflag_filter&b->core.flag) { skip = 1; continue; }
if (ma->conf->bed) { // test overlap
skip = !bed_overlap(ma->conf->bed, ma->h->target_name[b->core.tid], b->core.pos, bam_endpos(b));
if (skip) continue;
}
if (ma->conf->rghash) { // exclude read groups
uint8_t *rg = bam_aux_get(b, "RG");
skip = (rg && khash_str2int_get(ma->conf->rghash, (const char*)(rg+1), NULL)==0);
if (skip) continue;
}
if (ma->conf->flag & MPLP_ILLUMINA13) {
int i;
uint8_t *qual = bam_get_qual(b);
for (i = 0; i < b->core.l_qseq; ++i)
qual[i] = qual[i] > 31? qual[i] - 31 : 0;
}
has_ref = (ma->ref && ma->ref_id == b->core.tid)? 1 : 0;
skip = 0;
if (has_ref && (ma->conf->flag&MPLP_REALN)) bam_prob_realn_core(b, ma->ref, (ma->conf->flag & MPLP_REDO_BAQ)? 7 : 3);
if (has_ref && ma->conf->capQ_thres > 10) {
int q = bam_cap_mapQ(b, ma->ref, ma->conf->capQ_thres);
if (q < 0) skip = 1;
else if (b->core.qual > q) b->core.qual = q;
}
if (b->core.qual < ma->conf->min_mq) skip = 1;
else if ((ma->conf->flag&MPLP_NO_ORPHAN) && (b->core.flag&BAM_FPAIRED) && !(b->core.flag&BAM_FPROPER_PAIR)) skip = 1;
} while (skip);
return ret;
}
int calcCoverage(char *fName, Slice *slice, htsFile *in, hts_idx_t *idx, int flags) {
int ref;
int begRange;
int endRange;
char region[1024];
char region_name[512];
if (Slice_getChrStart(slice) != 1) {
fprintf(stderr, "Currently only allow a slice start position of 1\n");
return 1;
}
if (flags & M_UCSC_NAMING) {
sprintf(region,"chr%s", Slice_getSeqRegionName(slice));
} else {
sprintf(region,"%s", Slice_getSeqRegionName(slice));
}
bam_hdr_t *header = bam_hdr_init();
header = bam_hdr_read(in->fp.bgzf);
ref = bam_name2id(header, region);
if (ref < 0) {
fprintf(stderr, "Invalid region %s\n", region);
exit(1);
}
sprintf(region,"%s:%ld-%ld", region_name,
Slice_getSeqRegionStart(slice),
Slice_getSeqRegionEnd(slice));
if (hts_parse_reg(region, &begRange, &endRange) == NULL) {
fprintf(stderr, "Could not parse %s\n", region);
exit(2);
}
bam_hdr_destroy(header);
hts_itr_t *iter = sam_itr_queryi(idx, ref, begRange, endRange);
bam1_t *b = bam_init1();
Coverage *coverage = calloc(Slice_getLength(slice),sizeof(Coverage));
long counter = 0;
long overlapping = 0;
long bad = 0;
int startIndex = 0;
while (bam_itr_next(in, iter, b) >= 0) {
if (b->core.flag & (BAM_FUNMAP | BAM_FSECONDARY | BAM_FQCFAIL | BAM_FDUP)) {
bad++;
continue;
}
int end;
//end = bam_calend(&b->core, bam1_cigar(b));
end = bam_endpos(b);
// There is a special case for reads which have zero length and start at begRange (so end at begRange ie. before the first base we're interested in).
// That is the reason for the || end == begRange test
if (end == begRange) {
continue;
}
counter++;
if (!(counter%1000000)) {
if (verbosity > 1) { printf("."); }
fflush(stdout);
}
// Remember: b->core.pos is zero based!
int cigInd;
int refPos;
int readPos;
uint32_t *cigar = bam_get_cigar(b);
for (cigInd = readPos = 0, refPos = b->core.pos; cigInd < b->core.n_cigar; ++cigInd) {
int k;
int lenCigBlock = cigar[cigInd]>>4;
int op = cigar[cigInd]&0xf;
if (op == BAM_CMATCH || op == BAM_CEQUAL || op == BAM_CDIFF) {
for (k = 0; k < lenCigBlock; ++k) {
//if (ref[refPos+k] == 0) break; // out of boundary
coverage[refPos+k].coverage++;
}
if (k < lenCigBlock) break;
refPos += lenCigBlock; readPos += lenCigBlock;
} else if (op == BAM_CDEL) {
for (k = 0; k < lenCigBlock; ++k) {
// if (ref[refPos+k] == 0) break;
coverage[refPos+k].coverage++;
}
if (k < lenCigBlock) break;
refPos += lenCigBlock;
} else if (op == BAM_CSOFT_CLIP) {
readPos += lenCigBlock;
} else if (op == BAM_CHARD_CLIP) {
} else if (op == BAM_CINS) {
readPos += lenCigBlock;
} else if (op == BAM_CREF_SKIP) {
refPos += lenCigBlock;
}
}
#ifdef DONE
int j;
int done = 0;
int hadOverlap = 0;
for (j=startIndex; j < Vector_getNumElement(genes) && !done; j++) {
Gene *gene = Vector_getElementAt(genes,j);
if (!gene) {
continue;
}
// Remember: b->core.pos is zero based!
if (b->core.pos < Gene_getEnd(gene) && end >= Gene_getStart(gene)) {
int k;
int doneGene = 0;
for (k=0; k<Gene_getTranscriptCount(gene) && !doneGene; k++) {
Transcript *trans = Gene_getTranscriptAt(gene,k);
if (b->core.pos < Transcript_getEnd(trans) && end >= Transcript_getStart(trans)) {
int m;
for (m=0; m<Transcript_getExonCount(trans) && !doneGene; m++) {
Exon *exon = Transcript_getExonAt(trans,m);
if (b->core.pos < Exon_getEnd(exon) && end >= Exon_getStart(exon)) {
// Only count as overlapping once (could be that a read overlaps more than one gene)
if (!hadOverlap) {
overlapping++;
hadOverlap = 1;
}
gs = IDHash_getValue(geneCountsHash, Gene_getDbID(gene));
gs->score++;
doneGene = 1;
}
}
}
}
} else if (Gene_getStart(gene) > end) {
done = 1;
} else if (Gene_getEnd(gene) < b->core.pos+1) {
gs = IDHash_getValue(geneCountsHash, Gene_getDbID(gene));
printf("Gene %s (%s) score %ld\n",Gene_getStableId(gene),
Gene_getDisplayXref(gene) ? DBEntry_getDisplayId(Gene_getDisplayXref(gene)) : "",
gs->score);
if (verbosity > 1) {
printf("Removing gene %s (index %d) with extent %d to %d\n",
Gene_getStableId(gene),
gs->index,
Gene_getStart(gene),
Gene_getEnd(gene));
}
Vector_setElementAt(genes,j,NULL);
// Magic (very important for speed) - move startIndex to first non null gene
int n;
startIndex = 0;
for (n=0;n<Vector_getNumElement(genes);n++) {
void *v = Vector_getElementAt(genes,n);
if (v != NULL) {
break;
}
startIndex++;
}
if (verbosity > 1) {
printf("startIndex now %d\n",startIndex);
}
}
}
#endif
}
if (verbosity > 1) { printf("\n"); }
#ifdef DONE
// Print out read counts for what ever's left in the genes array
int n;
for (n=0;n<Vector_getNumElement(genes);n++) {
Gene *gene = Vector_getElementAt(genes,n);
if (gene != NULL) {
gs = IDHash_getValue(geneCountsHash, Gene_getDbID(gene));
printf("Gene %s (%s) score %ld\n",Gene_getStableId(gene),
Gene_getDisplayXref(gene) ? DBEntry_getDisplayId(Gene_getDisplayXref(gene)) : "",
gs->score);
}
}
#endif
printf("Read %ld reads. Number of bad reads (unmapped, qc fail, secondary, dup) %ld\n", counter, bad);
long i;
for (i=0; i< Slice_getLength(slice); i++) {
printf("%ld %ld\n", i+1, coverage[i].coverage);
}
sam_itr_destroy(iter);
bam_destroy1(b);
return 1;
}
int bam_pad2unpad(samFile *in, samFile *out, bam_hdr_t *h, faidx_t *fai)
{
bam1_t *b = 0;
kstring_t r, q;
int r_tid = -1;
uint32_t *cigar2 = 0;
int ret = 0, n2 = 0, m2 = 0, *posmap = 0;
b = bam_init1();
r.l = r.m = q.l = q.m = 0; r.s = q.s = 0;
int read_ret;
while ((read_ret = sam_read1(in, h, b)) >= 0) { // read one alignment from `in'
// Cannot depad unmapped CRAM data
if (b->core.flag & BAM_FUNMAP)
goto next_seq;
uint32_t *cigar = bam_get_cigar(b);
n2 = 0;
if (b->core.pos == 0 && b->core.tid >= 0 && strcmp(bam_get_qname(b), h->target_name[b->core.tid]) == 0) {
// fprintf(stderr, "[depad] Found embedded reference '%s'\n", bam_get_qname(b));
r_tid = b->core.tid;
if (0!=unpad_seq(b, &r)) {
fprintf(stderr, "[depad] ERROR: Problem parsing SEQ and/or CIGAR in reference %s\n", bam_get_qname(b));
return -1;
};
if (h->target_len[r_tid] != r.l) {
fprintf(stderr, "[depad] ERROR: (Padded) length of '%s' is %u in BAM header, but %llu in embedded reference\n", bam_get_qname(b), h->target_len[r_tid], (unsigned long long)(r.l));
return -1;
}
if (fai) {
// Check the embedded reference matches the FASTA file
if (load_unpadded_ref(fai, h->target_name[b->core.tid], h->target_len[b->core.tid], &q)) {
fprintf(stderr, "[depad] ERROR: Failed to load embedded reference '%s' from FASTA\n", h->target_name[b->core.tid]);
return -1;
}
assert(r.l == q.l);
int i;
for (i = 0; i < r.l; ++i) {
if (r.s[i] != q.s[i]) {
// Show gaps as ASCII 45
fprintf(stderr, "[depad] ERROR: Embedded sequence and reference FASTA don't match for %s base %i, '%c' vs '%c'\n",
h->target_name[b->core.tid], i+1,
r.s[i] ? seq_nt16_str[(int)r.s[i]] : 45,
q.s[i] ? seq_nt16_str[(int)q.s[i]] : 45);
return -1;
}
}
}
write_cigar(cigar2, n2, m2, bam_cigar_gen(b->core.l_qseq, BAM_CMATCH));
replace_cigar(b, n2, cigar2);
posmap = update_posmap(posmap, r);
} else if (b->core.n_cigar > 0) {
int i, k, op;
if (b->core.tid < 0) {
fprintf(stderr, "[depad] ERROR: Read '%s' has CIGAR but no RNAME\n", bam_get_qname(b));
return -1;
} else if (b->core.tid == r_tid) {
; // good case, reference available
//fprintf(stderr, "[depad] Have ref '%s' for read '%s'\n", h->target_name[b->core.tid], bam_get_qname(b));
} else if (fai) {
if (load_unpadded_ref(fai, h->target_name[b->core.tid], h->target_len[b->core.tid], &r)) {
fprintf(stderr, "[depad] ERROR: Failed to load '%s' from reference FASTA\n", h->target_name[b->core.tid]);
return -1;
}
posmap = update_posmap(posmap, r);
r_tid = b->core.tid;
// fprintf(stderr, "[depad] Loaded %s from FASTA file\n", h->target_name[b->core.tid]);
} else {
fprintf(stderr, "[depad] ERROR: Missing %s embedded reference sequence (and no FASTA file)\n", h->target_name[b->core.tid]);
return -1;
}
if (0!=unpad_seq(b, &q)) {
fprintf(stderr, "[depad] ERROR: Problem parsing SEQ and/or CIGAR in read %s\n", bam_get_qname(b));
return -1;
};
if (bam_cigar_op(cigar[0]) == BAM_CSOFT_CLIP) {
write_cigar(cigar2, n2, m2, cigar[0]);
} else if (bam_cigar_op(cigar[0]) == BAM_CHARD_CLIP) {
write_cigar(cigar2, n2, m2, cigar[0]);
if (b->core.n_cigar > 2 && bam_cigar_op(cigar[1]) == BAM_CSOFT_CLIP) {
write_cigar(cigar2, n2, m2, cigar[1]);
}
}
/* Determine CIGAR operator for each base in the aligned read */
for (i = 0, k = b->core.pos; i < q.l; ++i, ++k)
q.s[i] = q.s[i]? (r.s[k]? BAM_CMATCH : BAM_CINS) : (r.s[k]? BAM_CDEL : BAM_CPAD);
/* Include any pads if starts with an insert */
if (q.s[0] == BAM_CINS) {
for (k = 0; k+1 < b->core.pos && !r.s[b->core.pos - k - 1]; ++k);
if (k) write_cigar(cigar2, n2, m2, bam_cigar_gen(k, BAM_CPAD));
k = 0;
} else if (q.s[0] == BAM_CPAD) {
// Join 'k' CPAD to our first cigar op CPAD too.
for (k = 0; k+1 < b->core.pos && !r.s[b->core.pos - k - 1]; ++k);
} else {
k = 0;
}
/* Count consecutive CIGAR operators to turn into a CIGAR string */
for (i = 1, k++, op = q.s[0]; i < q.l; ++i) {
if (op != q.s[i]) {
write_cigar(cigar2, n2, m2, bam_cigar_gen(k, op));
op = q.s[i]; k = 1;
} else ++k;
}
write_cigar(cigar2, n2, m2, bam_cigar_gen(k, op));
if (bam_cigar_op(cigar[b->core.n_cigar-1]) == BAM_CSOFT_CLIP) {
write_cigar(cigar2, n2, m2, cigar[b->core.n_cigar-1]);
} else if (bam_cigar_op(cigar[b->core.n_cigar-1]) == BAM_CHARD_CLIP) {
if (b->core.n_cigar > 2 && bam_cigar_op(cigar[b->core.n_cigar-2]) == BAM_CSOFT_CLIP) {
write_cigar(cigar2, n2, m2, cigar[b->core.n_cigar-2]);
}
write_cigar(cigar2, n2, m2, cigar[b->core.n_cigar-1]);
}
/* Remove redundant P operators between M/X/=/D operators, e.g. 5M2P10M -> 15M */
int pre_op, post_op;
for (i = 2; i < n2; ++i)
if (bam_cigar_op(cigar2[i-1]) == BAM_CPAD) {
pre_op = bam_cigar_op(cigar2[i-2]);
post_op = bam_cigar_op(cigar2[i]);
/* Note don't need to check for X/= as code above will use M only */
if ((pre_op == BAM_CMATCH || pre_op == BAM_CDEL) && (post_op == BAM_CMATCH || post_op == BAM_CDEL)) {
/* This is a redundant P operator */
cigar2[i-1] = 0; // i.e. 0M
/* If had same operator either side, combine them in post_op */
if (pre_op == post_op) {
/* If CIGAR M, could treat as simple integers since BAM_CMATCH is zero*/
cigar2[i] = bam_cigar_gen(bam_cigar_oplen(cigar2[i-2]) + bam_cigar_oplen(cigar2[i]), post_op);
cigar2[i-2] = 0; // i.e. 0M
}
}
}
/* Remove the zero'd operators (0M) */
for (i = k = 0; i < n2; ++i)
if (cigar2[i]) cigar2[k++] = cigar2[i];
n2 = k;
replace_cigar(b, n2, cigar2);
}
/* Even unmapped reads can have a POS value, e.g. if their mate was mapped */
if (b->core.pos != -1) b->core.pos = posmap[b->core.pos];
if (b->core.mtid < 0 || b->core.mpos < 0) {
/* Nice case, no mate to worry about*/
// fprintf(stderr, "[depad] Read '%s' mate not mapped\n", bam_get_qname(b));
/* TODO - Warning if FLAG says mate should be mapped? */
/* Clean up funny input where mate position is given but mate reference is missing: */
b->core.mtid = -1;
b->core.mpos = -1;
} else if (b->core.mtid == b->core.tid) {
/* Nice case, same reference */
// fprintf(stderr, "[depad] Read '%s' mate mapped to same ref\n", bam_get_qname(b));
b->core.mpos = posmap[b->core.mpos];
} else {
/* Nasty case, Must load alternative posmap */
// fprintf(stderr, "[depad] Loading reference '%s' temporarily\n", h->target_name[b->core.mtid]);
if (!fai) {
fprintf(stderr, "[depad] ERROR: Needed reference %s sequence for mate (and no FASTA file)\n", h->target_name[b->core.mtid]);
return -1;
}
/* Temporarily load the other reference sequence */
if (load_unpadded_ref(fai, h->target_name[b->core.mtid], h->target_len[b->core.mtid], &r)) {
fprintf(stderr, "[depad] ERROR: Failed to load '%s' from reference FASTA\n", h->target_name[b->core.mtid]);
return -1;
}
posmap = update_posmap(posmap, r);
b->core.mpos = posmap[b->core.mpos];
/* Restore the reference and posmap*/
if (load_unpadded_ref(fai, h->target_name[b->core.tid], h->target_len[b->core.tid], &r)) {
fprintf(stderr, "[depad] ERROR: Failed to load '%s' from reference FASTA\n", h->target_name[b->core.tid]);
return -1;
}
posmap = update_posmap(posmap, r);
}
/* Most reads will have been moved so safest to always recalculate the BIN value */
b->core.bin = bam_reg2bin(b->core.pos, bam_endpos(b));
next_seq:
sam_write1(out, h, b);
}
if (read_ret < -1) {
fprintf(stderr, "[depad] truncated file.\n");
ret = 1;
}
free(r.s); free(q.s); free(posmap);
bam_destroy1(b);
return ret;
}
int main(int argc, char **argv) {
dlib::BamHandle in = dlib::BamHandle("bed_test.bam");
dlib::ParsedBed bed = dlib::ParsedBed("bed_test.bed", in.header);
bam1_t *b = bam_init1();
size_t diffs = 0;
void *lh3bed = bed_read("bed_test.bed");
samFile *so = sam_open("disagreed.bam", "wb9");
sam_hdr_write(so, in.header);
size_t disagrees = 0, agrees = 0;
int dbr = 0, lh3r = 0;
while(in.read(b) != -1) {
if(b->core.flag & (BAM_FUNMAP)) continue;
if((dbr = bed.bam1_test(b)) != (lh3r = bed_overlap(lh3bed, in.header->target_name[b->core.tid], b->core.pos, bam_endpos(b)))) {
LOG_EXIT("dbr: %i. lh3r: %i. Contig: %s. Position: %i. endpos; %i\n", dbr, lh3r, in.header->target_name[b->core.tid], b->core.pos, bam_endpos(b));
if(++disagrees % 100 == 0) LOG_DEBUG("disagrees: %lu.\n", disagrees);
sam_write1(so, in.header, b);
} else {
if(++agrees % 500000 == 0) LOG_DEBUG("agrees: %lu.\n", agrees);
}
}
sam_close(so);
bam_destroy1(b);
bed_destroy(lh3bed);
return EXIT_SUCCESS;
}
static int mplp_func(void *data, bam1_t *b)
{
char *ref;
mplp_aux_t *ma = (mplp_aux_t*)data;
int ret, skip = 0, ref_len;
do {
int has_ref;
ret = ma->iter? sam_itr_next(ma->fp, ma->iter, b) : sam_read1(ma->fp, ma->h, b);
if (ret < 0) break;
// The 'B' cigar operation is not part of the specification, considering as obsolete.
// bam_remove_B(b);
if (b->core.tid < 0 || (b->core.flag&BAM_FUNMAP)) { // exclude unmapped reads
skip = 1;
continue;
}
if (ma->conf->rflag_require && !(ma->conf->rflag_require&b->core.flag)) { skip = 1; continue; }
if (ma->conf->rflag_filter && ma->conf->rflag_filter&b->core.flag) { skip = 1; continue; }
if (ma->conf->bed && ma->conf->all == 0) { // test overlap
skip = !bed_overlap(ma->conf->bed, ma->h->target_name[b->core.tid], b->core.pos, bam_endpos(b));
if (skip) continue;
}
if (ma->conf->rghash) { // exclude read groups
uint8_t *rg = bam_aux_get(b, "RG");
skip = (rg && khash_str2int_get(ma->conf->rghash, (const char*)(rg+1), NULL)==0);
if (skip) continue;
}
if (ma->conf->flag & MPLP_ILLUMINA13) {
int i;
uint8_t *qual = bam_get_qual(b);
for (i = 0; i < b->core.l_qseq; ++i)
qual[i] = qual[i] > 31? qual[i] - 31 : 0;
}
if (ma->conf->fai && b->core.tid >= 0) {
has_ref = mplp_get_ref(ma, b->core.tid, &ref, &ref_len);
if (has_ref && ref_len <= b->core.pos) { // exclude reads outside of the reference sequence
fprintf(stderr,"[%s] Skipping because %d is outside of %d [ref:%d]\n",
__func__, b->core.pos, ref_len, b->core.tid);
skip = 1;
continue;
}
} else {
has_ref = 0;
}
skip = 0;
if (has_ref && (ma->conf->flag&MPLP_REALN)) sam_prob_realn(b, ref, ref_len, (ma->conf->flag & MPLP_REDO_BAQ)? 7 : 3);
if (has_ref && ma->conf->capQ_thres > 10) {
int q = sam_cap_mapq(b, ref, ref_len, ma->conf->capQ_thres);
if (q < 0) skip = 1;
else if (b->core.qual > q) b->core.qual = q;
}
if (b->core.qual < ma->conf->min_mq) skip = 1;
else if ((ma->conf->flag&MPLP_NO_ORPHAN) && (b->core.flag&BAM_FPAIRED) && !(b->core.flag&BAM_FPROPER_PAIR)) skip = 1;
} while (skip);
return ret;
}
