minimap2 0.2.27.0 → 0.2.29.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: f618028eabf476209264541d9037f68961548eb02dad4a22338bacdfe383fce7
-  data.tar.gz: f97eb69e9b1e78357cd738ba2a63ce36034e0fbd7c253c5a89a14b23ade19b01
+  metadata.gz: 15977381155ba5ee7d3352f9669db9dcf074e2073ac54961bdff2f6c9af77fb1
+  data.tar.gz: 67abab72034a636e796cc316e9891c81e049dec9ce8a4cd5b68f975384ba152d
 SHA512:
-  metadata.gz: 1ab742822a921c06f31671b0a555a0220cc9b2d0dc2e9d6ef7b72ff90fbc33de4d2f9285819b5064a28e3789065290e21f7e4dedb162c9927aefd8c860ceea35
-  data.tar.gz: 8d0b005004a1ac625a61d8a68073b31c08b4d156308ebfa01e18508e2ed520d948fa5d2a2e4804978a0846abc90bdc142183ae2d7d2c466cb2b00f87afff4d71
+  metadata.gz: b6df8d975c534b9c71715919867abc905e1bdf5712d3c1ba63cd6d15ea657367678153b448187186b128ee65abfc190f7415dc5924bb009cc61e4debdb2801fa
+  data.tar.gz: 87c2bf817379daf71b410e39ed3e5e5599329f0aa44f1db59449acb42e8033d58ec46fc74e94136e60595d101eaef94b4ba7561d7f4b9fe82d7a5cc6c62d77fa

data/README.md

@@ -65,6 +65,7 @@ pp hits
 * Minimap2 module
   - fastx_read                  Read fasta/fastq file.
   - revcomp                     Reverse complement sequence.
+  - execute                     Calls the main function of Minimap2 with arguments. `Minimap2.execute("--version")`
 
   * Aligner class
     * attributes

data/ext/cmappy/cmappy.c

@@ -50,13 +50,13 @@ void mm_reset_timer(void)
 	mm_realtime0 = realtime();
 }
 
-mm_reg1_t *mm_map_aux(const mm_idx_t *mi, const char *seq1, const char *seq2, int *n_regs, mm_tbuf_t *b, const mm_mapopt_t *opt)
+mm_reg1_t *mm_map_aux(const mm_idx_t *mi, const char* seqname, const char *seq1, const char *seq2, int *n_regs, mm_tbuf_t *b, const mm_mapopt_t *opt)
 {
 	mm_reg1_t *r;
 
 	// Py_BEGIN_ALLOW_THREADS
 	if (seq2 == 0) {
-		r = mm_map(mi, strlen(seq1), seq1, n_regs, b, opt, NULL);
+		r = mm_map(mi, strlen(seq1), seq1, n_regs, b, opt, seqname);
 	} else {
 		int _n_regs[2];
 		mm_reg1_t *regs[2];
@@ -73,7 +73,7 @@ mm_reg1_t *mm_map_aux(const mm_idx_t *mi, const char *seq1, const char *seq2, in
 			seq[1][i] = seq_comp_table[t];
 		}
 		if (len[1]&1) seq[1][len[1]>>1] = seq_comp_table[(uint8_t)seq[1][len[1]>>1]];
-		mm_map_frag(mi, 2, len, (const char**)seq, _n_regs, regs, b, opt, NULL);
+		mm_map_frag(mi, 2, len, (const char**)seq, _n_regs, regs, b, opt, seqname);
 		for (i = 0; i < _n_regs[1]; ++i)
 			regs[1][i].rev = !regs[1][i].rev;
 		*n_regs = _n_regs[0] + _n_regs[1];

data/ext/cmappy/cmappy.h

@@ -33,7 +33,7 @@ int mm_verbose_level(int v);
 void mm_reset_timer(void);
 
 extern unsigned char seq_comp_table[256];
-mm_reg1_t *mm_map_aux(const mm_idx_t *mi, const char *seq1, const char *seq2, int *n_regs, mm_tbuf_t *b, const mm_mapopt_t *opt);
+mm_reg1_t *mm_map_aux(const mm_idx_t *mi, const char* seqname, const char *seq1, const char *seq2, int *n_regs, mm_tbuf_t *b, const mm_mapopt_t *opt);
 
 char *mappy_revcomp(int len, const uint8_t *seq);
 

data/ext/minimap2/FAQ.md

@@ -2,7 +2,7 @@
 
 Without `-a`, `-c` or `--cs`, minimap2 only finds *approximate* mapping
 locations without detailed base alignment. In particular, the start and end
-positions of the alignment are impricise. With one of those options, minimap2
+positions of the alignment are imprecise. With one of those options, minimap2
 will perform base alignment, which is generally more accurate but is much
 slower.
 

data/ext/minimap2/Makefile

@@ -2,7 +2,7 @@ CFLAGS=		-g -Wall -O2 -Wc++-compat #-Wextra
 CPPFLAGS=	-DHAVE_KALLOC
 INCLUDES=
 OBJS=		kthread.o kalloc.o misc.o bseq.o sketch.o sdust.o options.o index.o \
-			lchain.o align.o hit.o seed.o map.o format.o pe.o esterr.o splitidx.o \
+			lchain.o align.o hit.o seed.o jump.o map.o format.o pe.o esterr.o splitidx.o \
 			ksw2_ll_sse.o
 PROG=		minimap2
 PROG_EXTRA=	sdust minimap2-lite
@@ -115,8 +115,9 @@ esterr.o: mmpriv.h minimap.h bseq.h kseq.h
 example.o: minimap.h kseq.h
 format.o: kalloc.h mmpriv.h minimap.h bseq.h kseq.h
 hit.o: mmpriv.h minimap.h bseq.h kseq.h kalloc.h khash.h
-index.o: kthread.h bseq.h minimap.h mmpriv.h kseq.h kvec.h kalloc.h khash.h
-index.o: ksort.h
+index.o: kthread.h bseq.h minimap.h mmpriv.h kseq.h ksw2.h kalloc.h kvec.h
+index.o: khash.h ksort.h
+jump.o: mmpriv.h minimap.h bseq.h kseq.h
 kalloc.o: kalloc.h
 ksw2_extd2_sse.o: ksw2.h kalloc.h
 ksw2_exts2_sse.o: ksw2.h kalloc.h

data/ext/minimap2/NEWS.md

@@ -1,3 +1,71 @@
+Release 2.29-r1283 (18 April 2025)
+----------------------------------
+
+Notable changes to minimap2:
+
+ * New feature: added the `splice:sr` preset for short RNA-seq read alignment.
+   Users may use `-j` to specify known gene annotation to improve spliced
+   alignment close to the ends of short reads. Also added `--write-junc` and
+   `--pass1` for 2-pass short-read RNA-seq alignment.
+
+ * Experimental feature: read splice scores from a file specified by `--spsc`
+   and consider the scores during base alignment. The feature makes it possible
+   to apply advanced splice models and to improve spliced alignment.
+
+ * Change: adjusted the mapping quality calculation for spliced alignment.
+
+ * Bugfixes: a) missing overlap alignment when base alignment is requested
+   (#969); b) incorrect summary information for long genomes (#1192); c)
+   missing parameter check for `--score-N` (#1226).
+
+ * Improvement: a) warn about absent junction files (#1229); b) report an error
+   if a wrong preset prefixed with "splice" is specified (#589).
+
+Notable changes to mappy:
+
+ * Improvement: allow passing read name (#1260)
+
+ * Improvement: exposed score for ambiguous bases (#1240)
+
+Minimap2 now supports short/long genomic/RNA-seq read alignment along with
+contig alignment and all-vs-all read overlapping. It produces identical genomic
+long-read or contig alignment to v2.27. Short genomic read alignment and the
+mapping quality of long RNA-seq read alignment may slightly differ in very rare
+cases.
+
+(2.29: 18 April 2025, r1283)
+
+
+
+Release 2.28-r1209 (27 March 2024)
+----------------------------------
+
+Notable changes to minimap2:
+
+ * Bugfix: `--MD` was not working properly due to the addition of `--ds` in the
+   last release (#1181 and #1182).
+
+ * New feature: added an experimental preset `lq:hqae` for aligning accurate
+   long reads back to their assembly. It has been observed that `map-hifi` and
+   `lr:hq` may produce many wrong alignments around centromeres when accurate
+   long reads (PacBio HiFi or Nanopore duplex/Q20+) are mapped to a diploid
+   assembly constructed from them. This new preset produces much more accurate
+   alignment. It is still experimental and may be subjective to changes in
+   future.
+
+ * Change: reduced the default `--cap-kalloc` to 500m to lower the peak
+   memory consumption (#855).
+
+Notable changes to mappy:
+
+ * Bugfix: mappy option struct was out of sync with minimap2 (#1177).
+
+Minimap2 should output identical alignments to v2.27.
+
+(2.28: 27 March 2024, r1209)
+
+
+
 Release 2.27-r1193 (12 March 2024)
 ----------------------------------
 

data/ext/minimap2/README.md

@@ -14,13 +14,15 @@ cd minimap2 && make
 # use presets (no test data)
 ./minimap2 -ax map-pb ref.fa pacbio.fq.gz > aln.sam       # PacBio CLR genomic reads
 ./minimap2 -ax map-ont ref.fa ont.fq.gz > aln.sam         # Oxford Nanopore genomic reads
-./minimap2 -ax map-hifi ref.fa pacbio-ccs.fq.gz > aln.sam # PacBio HiFi/CCS genomic reads (v2.19 or later)
-./minimap2 -ax lr:hq ref.fa ont-Q20.fq.gz > aln.sam       # Nanopore Q20 genomic reads (v2.27 or later)
+./minimap2 -ax map-hifi ref.fa pacbio-ccs.fq.gz > aln.sam # PacBio HiFi/CCS genomic reads (v2.19+)
+./minimap2 -ax lr:hq ref.fa ont-Q20.fq.gz > aln.sam       # Nanopore Q20 genomic reads (v2.27+)
 ./minimap2 -ax sr ref.fa read1.fa read2.fa > aln.sam      # short genomic paired-end reads
 ./minimap2 -ax splice ref.fa rna-reads.fa > aln.sam       # spliced long reads (strand unknown)
-./minimap2 -ax splice -uf -k14 ref.fa reads.fa > aln.sam  # noisy Nanopore Direct RNA-seq
-./minimap2 -ax splice:hq -uf ref.fa query.fa > aln.sam    # Final PacBio Iso-seq or traditional cDNA
-./minimap2 -ax splice --junc-bed anno.bed12 ref.fa query.fa > aln.sam  # prioritize on annotated junctions
+./minimap2 -ax splice -uf -k14 ref.fa reads.fa > aln.sam  # noisy Nanopore direct RNA-seq
+./minimap2 -ax splice:hq -uf ref.fa query.fa > aln.sam    # PacBio Kinnex/Iso-seq (RNA-seq)
+./minimap2 -ax splice --junc-bed=anno.bed12 ref.fa query.fa > aln.sam  # use annotated junctions
+./minimap2 -ax splice:sr ref.fa r1.fq r2.fq > aln.sam     # short-read RNA-seq (v2.29+)
+./minimap2 -ax splice:sr -j anno.bed12 ref.fa r1.fq r2.fq > aln.sam
 ./minimap2 -cx asm5 asm1.fa asm2.fa > aln.paf             # intra-species asm-to-asm alignment
 ./minimap2 -x ava-pb reads.fa reads.fa > overlaps.paf     # PacBio read overlap
 ./minimap2 -x ava-ont reads.fa reads.fa > overlaps.paf    # Nanopore read overlap
@@ -38,7 +40,8 @@ man ./minimap2.1
     - [Map long noisy genomic reads](#map-long-genomic)
     - [Map long mRNA/cDNA reads](#map-long-splice)
     - [Find overlaps between long reads](#long-overlap)
-    - [Map short accurate genomic reads](#short-genomic)
+    - [Map short genomic reads](#short-genomic)
+    - [Map short RNA-seq reads](#short-rna-seq)
     - [Full genome/assembly alignment](#full-genome)
   - [Advanced features](#advanced)
     - [Working with >65535 CIGAR operations](#long-cigar)
@@ -74,8 +77,8 @@ Detailed evaluations are available from the [minimap2 paper][doi] or the
 Minimap2 is optimized for x86-64 CPUs. You can acquire precompiled binaries from
 the [release page][release] with:
 ```sh
-curl -L https://github.com/lh3/minimap2/releases/download/v2.27/minimap2-2.27_x64-linux.tar.bz2 | tar -jxvf -
-./minimap2-2.27_x64-linux/minimap2
+curl -L https://github.com/lh3/minimap2/releases/download/v2.29/minimap2-2.29_x64-linux.tar.bz2 | tar -jxvf -
+./minimap2-2.29_x64-linux/minimap2
 ```
 If you want to compile from the source, you need to have a C compiler, GNU make
 and zlib development files installed. Then type `make` in the source code
@@ -171,9 +174,8 @@ or the last exons.
 
 Minimap2 rates an alignment by the score of the max-scoring sub-segment,
 *excluding* introns, and marks the best alignment as primary in SAM. When a
-spliced gene also has unspliced pseudogenes, minimap2 does not intentionally
-prefer spliced alignment, though in practice it more often marks the spliced
-alignment as the primary. By default, minimap2 outputs up to five secondary
+spliced gene also has unspliced pseudogenes, minimap2 slightly prefers
+the spliced alignment. By default, minimap2 outputs up to five secondary
 alignments (i.e. likely pseudogenes in the context of RNA-seq mapping). This
 can be tuned with option **-N**.
 
@@ -204,6 +206,10 @@ bonus score (tuned by `--junc-bonus`) if an aligned junction matches a junction
 in the annotation. Option `--junc-bed` also takes 5-column BED, including the
 strand field. In this case, each line indicates an oriented junction.
 
+**Note:** `--junc-bed` is intended for long noisy RNA-seq reads only.
+Applying the option to short RNA-seq reads would increase run time with little
+improvement to junction accuracy.
+
 #### <a name="long-overlap"></a>Find overlaps between long reads
 
 ```sh
@@ -216,7 +222,7 @@ the overlapping mode because it is slow and may produce false positive
 overlaps. However, if performance is not a concern, you may try to add `-a` or
 `-c` anyway.
 
-#### <a name="short-genomic"></a>Map short accurate genomic reads
+#### <a name="short-genomic"></a>Map short genomic reads
 
 ```sh
 minimap2 -ax sr ref.fa reads-se.fq > aln.sam           # single-end alignment
@@ -229,8 +235,18 @@ be paired if they are adjacent in the input stream and have the same name (with
 the `/[0-9]` suffix trimmed if present). Single- and paired-end reads can be
 mixed.
 
-Minimap2 does not work well with short spliced reads. There are many capable
-RNA-seq mappers for short reads.
+#### <a name="short-rna-seq"></a>Map short RNA-seq reads
+
+```sh
+minimap2 -ax splice:sr ref.fa reads-se.fq.gz > aln.sam           # single-end
+minimap2 -ax splice:sr ref.fa r1.fq.gz r2.fq.gz > aln.sam        # paired-end
+minimap2 -ax splice:sr -j anno.bed ref.fa r1.fq r2.fq > aln.sam  # use annotation
+# 2-pass alignment
+minimap2 -x splice:sr -j anno.bed --write-junc ref.fa r1.fq r2.fq > junc.bed
+minimap2 -ax splice:sr -j anno.bed --pass1=junc.bed ref.fa r1.fq r2.fq > aln.sam
+```
+The new preset `splice:sr` was added in v2.29. It functions similarly to `sr`
+except that it performs spliced alignment.
 
 #### <a name="full-genome"></a>Full genome/assembly alignment
 

data/ext/minimap2/align.c

@@ -6,6 +6,8 @@
 #include "mmpriv.h"
 #include "ksw2.h"
 
+#define MM_MAX_QLEN_FLANK 100
+
 static void ksw_gen_simple_mat(int m, int8_t *mat, int8_t a, int8_t b, int8_t sc_ambi)
 {
 	int i, j;
@@ -258,7 +260,7 @@ static void mm_update_extra(mm_reg1_t *r, const uint8_t *qseq, const uint8_t *ts
 	if (p == 0) return;
 	mm_fix_cigar(r, qseq, tseq, &qshift, &tshift);
 	qseq += qshift, tseq += tshift; // qseq and tseq may be shifted due to the removal of leading I/D
-	r->blen = r->mlen = 0;
+	r->blen = r->mlen = 0, r->is_spliced = 0;
 	for (k = 0; k < p->n_cigar; ++k) {
 		uint32_t op = p->cigar[k]&0xf, len = p->cigar[k]>>4;
 		if (op == MM_CIGAR_MATCH) {
@@ -292,7 +294,7 @@ static void mm_update_extra(mm_reg1_t *r, const uint8_t *qseq, const uint8_t *ts
 			if (s < 0) s = 0;
 			toff += len;
 		} else if (op == MM_CIGAR_N_SKIP) {
-			toff += len;
+			r->is_spliced = 1, toff += len;
 		}
 	}
 	p->dp_max = p->dp_max0 = (int32_t)(max + .499);
@@ -300,9 +302,8 @@ static void mm_update_extra(mm_reg1_t *r, const uint8_t *qseq, const uint8_t *ts
 	if (is_eqx) mm_update_cigar_eqx(r, qseq, tseq); // NB: it has to be called here as changes to qseq and tseq are not returned
 }
 
-static void mm_append_cigar(mm_reg1_t *r, uint32_t n_cigar, uint32_t *cigar) // TODO: this calls the libc realloc()
+void mm_enlarge_cigar(mm_reg1_t *r, uint32_t n_cigar) // TODO: this calls the libc realloc()
 {
-	mm_extra_t *p;
 	if (n_cigar == 0) return;
 	if (r->p == 0) {
 		uint32_t capacity = n_cigar + sizeof(mm_extra_t)/4;
@@ -314,6 +315,13 @@ static void mm_append_cigar(mm_reg1_t *r, uint32_t n_cigar, uint32_t *cigar) //
 		kroundup32(r->p->capacity);
 		r->p = (mm_extra_t*)realloc(r->p, r->p->capacity * 4);
 	}
+}
+
+static void mm_append_cigar(mm_reg1_t *r, uint32_t n_cigar, const uint32_t *cigar)
+{
+	mm_extra_t *p;
+	if (n_cigar == 0) return;
+	mm_enlarge_cigar(r, n_cigar);
 	p = r->p;
 	if (p->n_cigar > 0 && (p->cigar[p->n_cigar-1]&0xf) == (cigar[0]&0xf)) { // same CIGAR op at the boundary
 		p->cigar[p->n_cigar-1] += cigar[0]>>4<<4;
@@ -325,29 +333,31 @@ static void mm_append_cigar(mm_reg1_t *r, uint32_t n_cigar, uint32_t *cigar) //
 	}
 }
 
-static void mm_align_pair(void *km, const mm_mapopt_t *opt, int qlen, const uint8_t *qseq, int tlen, const uint8_t *tseq, const uint8_t *junc, const int8_t *mat, int w, int end_bonus, int zdrop, int flag, ksw_extz_t *ez)
+static void mm_align_pair(void *km, const mm_mapopt_t *opt, int qlen, const uint8_t *qseq, int tlen, const uint8_t *tseq, const uint8_t *junc,
+						  const int8_t *mat, int w, int end_bonus, int zdrop, int ksw_flag, ksw_extz_t *ez)
 {
 	if (mm_dbg_flag & MM_DBG_PRINT_ALN_SEQ) {
 		int i;
-		fprintf(stderr, "===> q=(%d,%d), e=(%d,%d), bw=%d, flag=%d, zdrop=%d <===\n", opt->q, opt->q2, opt->e, opt->e2, w, flag, opt->zdrop);
+		fprintf(stderr, "===> q=(%d,%d), e=(%d,%d), bw=%d, ksw_flag=%d, zdrop=%d, end_bonus=%d <===\n", opt->q, opt->q2, opt->e, opt->e2, w, ksw_flag, opt->zdrop, end_bonus);
 		for (i = 0; i < tlen; ++i) fputc("ACGTN"[tseq[i]], stderr);
 		fputc('\n', stderr);
 		for (i = 0; i < qlen; ++i) fputc("ACGTN"[qseq[i]], stderr);
 		fputc('\n', stderr);
 	}
 	if (opt->transition != 0 && opt->b != opt->transition)
-		flag |= KSW_EZ_GENERIC_SC;
-	if (opt->max_sw_mat > 0 && (int64_t)tlen * qlen > opt->max_sw_mat) {
+		ksw_flag |= KSW_EZ_GENERIC_SC;
+	if (opt->max_sw_mat > 0 && (int64_t)tlen * qlen > opt->max_sw_mat) { // too much memory; skip alignment
 		ksw_reset_extz(ez);
 		ez->zdropped = 1;
-	} else if (opt->flag & MM_F_SPLICE) {
-		int flag_tmp = flag;
-		if (!(opt->flag & MM_F_SPLICE_OLD)) flag_tmp |= KSW_EZ_SPLICE_CMPLX;
-		ksw_exts2_sse(km, qlen, qseq, tlen, tseq, 5, mat, opt->q, opt->e, opt->q2, opt->noncan, zdrop, opt->junc_bonus, flag_tmp, junc, ez);
-	} else if (opt->q == opt->q2 && opt->e == opt->e2)
-		ksw_extz2_sse(km, qlen, qseq, tlen, tseq, 5, mat, opt->q, opt->e, w, zdrop, end_bonus, flag, ez);
-	else
-		ksw_extd2_sse(km, qlen, qseq, tlen, tseq, 5, mat, opt->q, opt->e, opt->q2, opt->e2, w, zdrop, end_bonus, flag, ez);
+	} else if (opt->flag & MM_F_SPLICE) { // spliced alignment
+		assert((ksw_flag & KSW_EZ_SPLICE_FOR) == 0 || (ksw_flag & KSW_EZ_SPLICE_REV) == 0);
+		if (!(opt->flag & MM_F_SPLICE_OLD)) ksw_flag |= KSW_EZ_SPLICE_CMPLX;
+		ksw_exts2_sse(km, qlen, qseq, tlen, tseq, 5, mat, opt->q, opt->e, opt->q2, opt->noncan, zdrop, end_bonus, opt->junc_bonus, opt->junc_pen, ksw_flag, junc, ez);
+	} else if (opt->q == opt->q2 && opt->e == opt->e2) { // affine gap
+		ksw_extz2_sse(km, qlen, qseq, tlen, tseq, 5, mat, opt->q, opt->e, w, zdrop, end_bonus, ksw_flag, ez);
+	} else { // dual affine gap
+		ksw_extd2_sse(km, qlen, qseq, tlen, tseq, 5, mat, opt->q, opt->e, opt->q2, opt->e2, w, zdrop, end_bonus, ksw_flag, ez);
+	}
 	if (mm_dbg_flag & MM_DBG_PRINT_ALN_SEQ) {
 		int i;
 		fprintf(stderr, "score=%d, cigar=", ez->score);
@@ -357,6 +367,45 @@ static void mm_align_pair(void *km, const mm_mapopt_t *opt, int qlen, const uint
 	}
 }
 
+static int mm_align_sr_rna(void *km, const mm_mapopt_t *opt, int qlen, const uint8_t *qseq, int tlen, const uint8_t *tseq, const uint8_t *junc, uint8_t *tseq2, uint8_t *junc2,
+							const int8_t *mat, int w, int end_bonus, int zdrop, int ksw_flag, ksw_extz_t *ez)
+{
+	int32_t ilen = opt->q2 * 2, tlen2 = qlen * 2 + ilen;
+	int32_t i, ll = 0, lr = 0, nn = 0, n_ins = 0;
+	if (!(opt->flag & MM_F_SPLICE)) return 0; // only for spliced alignment
+	if (qlen > MM_MAX_QLEN_FLANK || qlen * 2 + ilen > tlen) return 0; // the query sequence can't be too long and the target sequence must be long enough
+	for (i = 0; i < qlen; ++i) // exact match length from the left
+		if (qseq[i] == tseq[i] && qseq[i] < 4)
+			++ll;
+	for (i = 0; i < qlen; ++i) // exact match length from the right
+		if (qseq[qlen - 1 - i] == tseq[tlen - 1 - i] && qseq[qlen - 1 - i] < 4)
+			++lr;
+	if (qlen - (ll + lr) > 9) return 0; // qlen may be smaller than ll+lr
+	memcpy(tseq2, tseq, qlen);
+	memset(&tseq2[qlen], 4, ilen);
+	memcpy(&tseq2[qlen + ilen], &tseq[tlen - qlen], qlen);
+	if (junc) {
+		memcpy(junc2, junc, qlen);
+		memset(&junc2[qlen], 0, ilen);
+		memcpy(&junc2[qlen + ilen], &junc[tlen - qlen], qlen);
+	}
+	if (!(opt->flag & MM_F_SPLICE_OLD)) ksw_flag |= KSW_EZ_SPLICE_CMPLX;
+	ksw_exts2_sse(km, qlen, qseq, tlen2, tseq2, 5, mat, opt->q, opt->e, opt->q2, opt->noncan, zdrop, end_bonus, opt->junc_bonus, opt->junc_pen, ksw_flag, junc2, ez);
+	if (ez->zdropped) return 0;
+	if ((ez->cigar[0]&0xf) != KSW_CIGAR_MATCH || (ez->cigar[ez->n_cigar-1]&0xf) != KSW_CIGAR_MATCH) return 0;
+	for (i = 0; i < ez->n_cigar; ++i) { // count the number of introns in the alignment
+		if ((ez->cigar[i]&0xf) == KSW_CIGAR_N_SKIP)
+			++nn;
+		else if ((ez->cigar[i]&0xf) == KSW_CIGAR_INS)
+			++n_ins;
+	}
+	if (nn != 1 || n_ins > 0) return 0; // the heuristic only works when there is exactly one intron
+	for (i = 0; i < ez->n_cigar; ++i)
+		if ((ez->cigar[i]&0xf) == KSW_CIGAR_N_SKIP)
+			ez->cigar[i] += (tlen - tlen2) << 4;
+	return 1;
+}
+
 static inline int mm_get_hplen_back(const mm_idx_t *mi, uint32_t rid, uint32_t x)
 {
 	int64_t i, off0 = mi->seq[rid].offset, off = off0 + x;
@@ -586,12 +635,19 @@ static void mm_fix_bad_ends_splice(void *km, const mm_mapopt_t *opt, const mm_id
 	}
 }
 
+static inline void mm_get_junc(const mm_idx_t *mi, int32_t ctg, int32_t st, int32_t en, int32_t rev, uint8_t *junc)
+{
+	if (mi->spsc) mm_idx_spsc_get(mi, ctg, st, en, rev, junc);
+	else if (mi->I) mm_idx_bed_junc(mi, ctg, st, en, junc);
+	else memset(junc, 0, en - st);
+}
+
 static void mm_align1(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, int qlen, uint8_t *qseq0[2], mm_reg1_t *r, mm_reg1_t *r2, int n_a, mm128_t *a, ksw_extz_t *ez, int splice_flag)
 {
-	int is_sr = !!(opt->flag & MM_F_SR), is_splice = !!(opt->flag & MM_F_SPLICE);
+	int is_sr = !!(opt->flag & MM_F_SR), is_splice = !!(opt->flag & MM_F_SPLICE), is_sr_rna = (!!(opt->flag & MM_F_SR_RNA) && is_splice);
 	int32_t rid = a[r->as].x<<1>>33, rev = a[r->as].x>>63, as1, cnt1;
-	uint8_t *tseq, *qseq, *junc;
-	int32_t i, l, bw, bw_long, dropped = 0, extra_flag = 0, rs0, re0, qs0, qe0;
+	uint8_t *tseq, *qseq, *junc, *tseq2 = 0, *junc2 = 0;
+	int32_t i, l, bw, bw_long, dropped = 0, ksw_flag = 0, rs0, re0, qs0, qe0;
 	int32_t rs, re, qs, qe;
 	int32_t rs1, qs1, re1, qe1;
 	int8_t mat[25];
@@ -626,9 +682,10 @@ static void mm_align1(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, int
 	assert(cnt1 > 0);
 
 	if (is_splice) {
-		if (splice_flag & MM_F_SPLICE_FOR) extra_flag |= rev? KSW_EZ_SPLICE_REV : KSW_EZ_SPLICE_FOR;
-		if (splice_flag & MM_F_SPLICE_REV) extra_flag |= rev? KSW_EZ_SPLICE_FOR : KSW_EZ_SPLICE_REV;
-		if (opt->flag & MM_F_SPLICE_FLANK) extra_flag |= KSW_EZ_SPLICE_FLANK;
+		if (splice_flag & MM_F_SPLICE_FOR) ksw_flag |= rev? KSW_EZ_SPLICE_REV : KSW_EZ_SPLICE_FOR;
+		if (splice_flag & MM_F_SPLICE_REV) ksw_flag |= rev? KSW_EZ_SPLICE_FOR : KSW_EZ_SPLICE_REV;
+		if (opt->flag & MM_F_SPLICE_FLANK) ksw_flag |= KSW_EZ_SPLICE_FLANK;
+		if (mi->spsc) ksw_flag |= KSW_EZ_SPLICE_SCORE;
 	}
 
 	/* Look for the start and end of regions to perform DP. This sounds easy
@@ -713,6 +770,12 @@ static void mm_align1(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, int
 	tseq = (uint8_t*)kmalloc(km, re0 - rs0);
 	junc = (uint8_t*)kmalloc(km, re0 - rs0);
 
+	if (is_sr_rna) {
+		int32_t max_tlen2 = MM_MAX_QLEN_FLANK * 2 + opt->q2 * 2;
+		tseq2 = Kmalloc(km, uint8_t, max_tlen2 * 2);
+		junc2 = tseq2 + max_tlen2;
+	}
+
 	if (qs > 0 && rs > 0) { // left extension; probably the condition can be changed to "qs > qs0 && rs > rs0"
 		if (opt->flag & MM_F_QSTRAND) {
 			qseq = &qseq0[0][qs0];
@@ -721,11 +784,11 @@ static void mm_align1(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, int
 			qseq = &qseq0[rev][qs0];
 			mm_idx_getseq(mi, rid, rs0, rs, tseq);
 		}
-		mm_idx_bed_junc(mi, rid, rs0, rs, junc);
+		mm_get_junc(mi, rid, rs0, rs, !!(ksw_flag&KSW_EZ_SPLICE_REV), junc);
 		mm_seq_rev(qs - qs0, qseq);
 		mm_seq_rev(rs - rs0, tseq);
 		mm_seq_rev(rs - rs0, junc);
-		mm_align_pair(km, opt, qs - qs0, qseq, rs - rs0, tseq, junc, mat, bw, opt->end_bonus, r->split_inv? opt->zdrop_inv : opt->zdrop, extra_flag|KSW_EZ_EXTZ_ONLY|KSW_EZ_RIGHT|KSW_EZ_REV_CIGAR, ez);
+		mm_align_pair(km, opt, qs - qs0, qseq, rs - rs0, tseq, junc, mat, bw, opt->end_bonus, r->split_inv? opt->zdrop_inv : opt->zdrop, ksw_flag|KSW_EZ_EXTZ_ONLY|KSW_EZ_RIGHT|KSW_EZ_REV_CIGAR, ez);
 		if (ez->n_cigar > 0) {
 			mm_append_cigar(r, ez->n_cigar, ez->cigar);
 			r->p->dp_score += ez->max;
@@ -737,14 +800,14 @@ static void mm_align1(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, int
 	re1 = rs, qe1 = qs;
 	assert(qs1 >= 0 && rs1 >= 0);
 
-	for (i = is_sr? cnt1 - 1 : 1; i < cnt1; ++i) { // gap filling
+	for (i = is_sr? cnt1 - 1 : 1; i < cnt1; ++i) { // gap filling; for short genomic reads, fill from the first seed to the last
 		if ((a[as1+i].y & (MM_SEED_IGNORE|MM_SEED_TANDEM)) && i != cnt1 - 1) continue;
 		if (is_sr && !(mi->flag & MM_I_HPC)) {
 			re = (int32_t)a[as1 + i].x + 1;
 			qe = (int32_t)a[as1 + i].y + 1;
 		} else mm_adjust_minier(mi, qseq0, &a[as1 + i], &re, &qe);
 		re1 = re, qe1 = qe;
-		if (i == cnt1 - 1 || (a[as1+i].y&MM_SEED_LONG_JOIN) || (qe - qs >= opt->min_ksw_len && re - rs >= opt->min_ksw_len)) {
+		if (i == cnt1 - 1 || (a[as1+i].y&MM_SEED_LONG_JOIN) || (qe - qs >= opt->min_ksw_len && re - rs >= opt->min_ksw_len)) { // gap filling
 			int j, bw1 = bw_long, zdrop_code;
 			if (a[as1+i].y & MM_SEED_LONG_JOIN)
 				bw1 = qe - qs > re - rs? qe - qs : re - rs;
@@ -756,21 +819,29 @@ static void mm_align1(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, int
 				qseq = &qseq0[rev][qs];
 				mm_idx_getseq(mi, rid, rs, re, tseq);
 			}
-			mm_idx_bed_junc(mi, rid, rs, re, junc);
-			if (is_sr) { // perform ungapped alignment
+			mm_get_junc(mi, rid, rs, re, !!(ksw_flag&KSW_EZ_SPLICE_REV), junc);
+			if (is_sr || (is_sr_rna && qe - qs == re - rs)) { // perform ungapped alignment
+				int32_t max_gapped_score = (qe - qs - 2) * opt->a - 2 * (opt->q + opt->e);
 				assert(qe - qs == re - rs);
 				ksw_reset_extz(ez);
 				for (j = 0, ez->score = 0; j < qe - qs; ++j) {
-					if (qseq[j] >= 4 || tseq[j] >= 4) ez->score += opt->e2;
+					if (qseq[j] >= 4 || tseq[j] >= 4) ez->score += opt->sc_ambi > 0? -opt->sc_ambi : opt->sc_ambi;
 					else ez->score += qseq[j] == tseq[j]? opt->a : -opt->b;
 				}
-				ez->cigar = ksw_push_cigar(km, &ez->n_cigar, &ez->m_cigar, ez->cigar, MM_CIGAR_MATCH, qe - qs);
+				if (ez->score > max_gapped_score)
+					ez->cigar = ksw_push_cigar(km, &ez->n_cigar, &ez->m_cigar, ez->cigar, MM_CIGAR_MATCH, qe - qs);
+				else
+					mm_align_pair(km, opt, qe - qs, qseq, re - rs, tseq, junc, mat, bw1, -1, opt->zdrop, ksw_flag|KSW_EZ_APPROX_MAX, ez);
 			} else { // perform normal gapped alignment
-				mm_align_pair(km, opt, qe - qs, qseq, re - rs, tseq, junc, mat, bw1, -1, opt->zdrop, extra_flag|KSW_EZ_APPROX_MAX, ez); // first pass: with approximate Z-drop
+				int32_t skip_full = 0;
+				if (is_sr_rna)
+					skip_full = mm_align_sr_rna(km, opt, qe - qs, qseq, re - rs, tseq, junc, tseq2, junc2, mat, bw1, -1, opt->zdrop, ksw_flag|KSW_EZ_APPROX_MAX, ez);
+				if (!skip_full)
+					mm_align_pair(km, opt, qe - qs, qseq, re - rs, tseq, junc, mat, bw1, -1, opt->zdrop, ksw_flag|KSW_EZ_APPROX_MAX, ez); // first pass: with approximate Z-drop
 			}
 			// test Z-drop and inversion Z-drop
 			if ((zdrop_code = mm_test_zdrop(km, opt, qseq, tseq, ez->n_cigar, ez->cigar, mat)) != 0)
-				mm_align_pair(km, opt, qe - qs, qseq, re - rs, tseq, junc, mat, bw1, -1, zdrop_code == 2? opt->zdrop_inv : opt->zdrop, extra_flag, ez); // second pass: lift approximate
+				mm_align_pair(km, opt, qe - qs, qseq, re - rs, tseq, junc, mat, bw1, -1, zdrop_code == 2? opt->zdrop_inv : opt->zdrop, ksw_flag, ez); // second pass: lift approximate
 			// update CIGAR
 			if (ez->n_cigar > 0)
 				mm_append_cigar(r, ez->n_cigar, ez->cigar);
@@ -808,8 +879,8 @@ static void mm_align1(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, int
 			qseq = &qseq0[rev][qe];
 			mm_idx_getseq(mi, rid, re, re0, tseq);
 		}
-		mm_idx_bed_junc(mi, rid, re, re0, junc);
-		mm_align_pair(km, opt, qe0 - qe, qseq, re0 - re, tseq, junc, mat, bw, opt->end_bonus, opt->zdrop, extra_flag|KSW_EZ_EXTZ_ONLY, ez);
+		mm_get_junc(mi, rid, re, re0, !!(ksw_flag&KSW_EZ_SPLICE_REV), junc);
+		mm_align_pair(km, opt, qe0 - qe, qseq, re0 - re, tseq, junc, mat, bw, opt->end_bonus, opt->zdrop, ksw_flag|KSW_EZ_EXTZ_ONLY, ez);
 		if (ez->n_cigar > 0) {
 			mm_append_cigar(r, ez->n_cigar, ez->cigar);
 			r->p->dp_score += ez->max;
@@ -832,11 +903,12 @@ static void mm_align1(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, int
 			mm_idx_getseq(mi, rid, rs1, re1, tseq);
 			qseq = &qseq0[r->rev][qs1];
 		}
-		mm_update_extra(r, qseq, tseq, mat, opt->q, opt->e, opt->flag & MM_F_EQX, !(opt->flag & MM_F_SR));
+		mm_update_extra(r, qseq, tseq, mat, opt->q, opt->e, opt->flag & MM_F_EQX, !(is_sr || is_sr_rna));
 		if (rev && r->p->trans_strand)
 			r->p->trans_strand ^= 3; // flip to the read strand
 	}
 
+	if (tseq2) kfree(km, tseq2);
 	kfree(km, tseq);
 	kfree(km, junc);
 }
@@ -891,7 +963,7 @@ static int mm_align1_inv(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, i
 	}
 	r_inv->rs = r1->re + t_off;
 	r_inv->re = r_inv->rs + ez->max_t + 1;
-	mm_update_extra(r_inv, &qseq[q_off], &tseq[t_off], mat, opt->q, opt->e, opt->flag & MM_F_EQX, !(opt->flag & MM_F_SR));
+	mm_update_extra(r_inv, &qseq[q_off], &tseq[t_off], mat, opt->q, opt->e, opt->flag & MM_F_EQX, !(opt->flag & (MM_F_SR|MM_F_SR_RNA)));
 	ret = 1;
 end_align1_inv:
 	kfree(km, tseq);
@@ -933,14 +1005,14 @@ double mm_event_identity(const mm_reg1_t *r)
 static int32_t mm_recal_max_dp(const mm_reg1_t *r, double b2, int32_t match_sc)
 {
 	uint32_t i;
-	int32_t n_gap = 0, n_gapo = 0, n_mis;
+	int32_t n_gap = 0, n_mis;
 	double gap_cost = 0.0;
 	if (r->p == 0) return -1;
 	for (i = 0; i < r->p->n_cigar; ++i) {
 		int32_t op = r->p->cigar[i] & 0xf, len = r->p->cigar[i] >> 4;
 		if (op == MM_CIGAR_INS || op == MM_CIGAR_DEL) {
 			gap_cost += b2 + (double)mg_log2(1.0 + len);
-			++n_gapo, n_gap += len;
+			n_gap += len;
 		}
 	}
 	n_mis = r->blen + r->p->n_ambi - r->mlen - n_gap;
@@ -992,24 +1064,36 @@ mm_reg1_t *mm_align_skeleton(void *km, const mm_mapopt_t *opt, const mm_idx_t *m
 	n_a = mm_squeeze_a(km, n_regs, regs, a);
 	memset(&ez, 0, sizeof(ksw_extz_t));
 	for (i = 0; i < n_regs; ++i) {
-		mm_reg1_t r2;
+		mm_reg1_t r2; // only used for inversion
 		if ((opt->flag&MM_F_SPLICE) && (opt->flag&MM_F_SPLICE_FOR) && (opt->flag&MM_F_SPLICE_REV)) { // then do two rounds of alignments for both strands
-			mm_reg1_t s[2], s2[2];
-			int which, trans_strand;
+			mm_reg1_t s[2], s2[2], *r;
 			s[0] = s[1] = regs[i];
-			mm_align1(km, opt, mi, qlen, qseq0, &s[0], &s2[0], n_a, a, &ez, MM_F_SPLICE_FOR);
-			mm_align1(km, opt, mi, qlen, qseq0, &s[1], &s2[1], n_a, a, &ez, MM_F_SPLICE_REV);
-			if (s[0].p->dp_score > s[1].p->dp_score) which = 0, trans_strand = 1;
-			else if (s[0].p->dp_score < s[1].p->dp_score) which = 1, trans_strand = 2;
-			else trans_strand = 3, which = (qlen + s[0].p->dp_score) & 1; // randomly choose a strand, effectively
-			if (which == 0) {
+			mm_align1(km, opt, mi, qlen, qseq0, &s[0], &s2[0], n_a, a, &ez, MM_F_SPLICE_FOR); // assume the transcript is on the + strand of the genome
+			if ((opt->flag&MM_F_SR_RNA) && regs[i].qe - regs[i].qs == regs[i].re - regs[i].rs && s[0].qe - s[0].qs == s[0].re - s[0].rs && s[0].qs == 0 && s[0].qe == qlen) {
 				regs[i] = s[0], r2 = s2[0];
-				free(s[1].p);
+				regs[i].p->trans_strand = 0;
 			} else {
-				regs[i] = s[1], r2 = s2[1];
-				free(s[0].p);
+				int which, trans_strand;
+				mm_align1(km, opt, mi, qlen, qseq0, &s[1], &s2[1], n_a, a, &ez, MM_F_SPLICE_REV); // assume the transcript on the - strand
+				if (s[0].p->dp_score > s[1].p->dp_score) which = 0, trans_strand = 1;
+				else if (s[0].p->dp_score < s[1].p->dp_score) which = 1, trans_strand = 2;
+				else trans_strand = 3, which = (qlen + s[0].p->dp_score) & 1; // randomly choose a strand, effectively
+				if (which == 0) {
+					regs[i] = s[0], r2 = s2[0];
+					free(s[1].p);
+				} else {
+					regs[i] = s[1], r2 = s2[1];
+					free(s[0].p);
+				}
+				r = &regs[i];
+				r->p->trans_strand = trans_strand;
+				if (r->is_spliced) {
+					if (trans_strand == 1 || trans_strand == 2) // this is an *approximate* way to tell if there are splice signals.
+						r->p->dp_max += (opt->a + opt->b) + ((opt->a + opt->b) >> 1);
+					else if (trans_strand == 3)
+						r->p->dp_max -= opt->a + opt->b;
+				}
 			}
-			regs[i].p->trans_strand = trans_strand;
 		} else { // one round of alignment
 			mm_align1(km, opt, mi, qlen, qseq0, &regs[i], &r2, n_a, a, &ez, opt->flag);
 			if (opt->flag&MM_F_SPLICE)
@@ -1027,7 +1111,7 @@ mm_reg1_t *mm_align_skeleton(void *km, const mm_mapopt_t *opt, const mm_idx_t *m
 	kfree(km, qseq0[0]);
 	kfree(km, ez.cigar);
 	mm_filter_regs(opt, qlen, n_regs_, regs);
-	if (!(opt->flag&MM_F_SR) && !opt->split_prefix && qlen >= opt->rank_min_len) {
+	if (!(opt->flag&(MM_F_SR|MM_F_SR_RNA|MM_F_ALL_CHAINS)) && !opt->split_prefix && qlen >= opt->rank_min_len) {
 		mm_update_dp_max(qlen, *n_regs_, regs, opt->rank_frac, opt->a, opt->b);
 		mm_filter_regs(opt, qlen, n_regs_, regs);
 	}

data/ext/minimap2/cookbook.md

@@ -31,8 +31,8 @@ To acquire the data used in this cookbook and to install minimap2 and paftools,
 please follow the command lines below:
 ```sh
 # install minimap2 executables
-curl -L https://github.com/lh3/minimap2/releases/download/v2.27/minimap2-2.27_x64-linux.tar.bz2 | tar jxf -
-cp minimap2-2.27_x64-linux/{minimap2,k8,paftools.js} .  # copy executables
+curl -L https://github.com/lh3/minimap2/releases/download/v2.29/minimap2-2.29_x64-linux.tar.bz2 | tar jxf -
+cp minimap2-2.29_x64-linux/{minimap2,k8,paftools.js} .  # copy executables
 export PATH="$PATH:"`pwd`                               # put the current directory on PATH
 # download example datasets
 curl -L https://github.com/lh3/minimap2/releases/download/v2.10/cookbook-data.tgz | tar zxf -