--junc-bonus option is ignored

[mparker2]

Thanks for writing minimap2. I have come to rely on it a lot for mapping nanopore direct RNA data. I was interested in trying some kind of two-pass alignment as mentioned in #348 so have been experimenting with the --junc-bed and --junc-bonus options. As far as I can tell, however, it seems like the --junc-bonus option is ignored. If the splice preset is used, --junc-bonus seems to get set as a positive number (according to the manual it is set to 9). Varying the --junc-bonus option does not seem to change the number (if spliced preset is used but --junc-bonus is set to zero the junctions in the bed file still seem to get used, whereas I would have expected the result to be the same as if no bed file was supplied). Furthermore if no presets are used the junction bonus seems to be set as zero regardless of whether --junc-bonus option is specified or not.

For example:

no splice preset, junc-bonus=0

# apart from the junc-bed related settings, these settings should be identical to -x splice preset

minimap2 -a -t12 -k15 -w5 --splice \
  -g2000 -G200k -A1 -B2 -O2,32 -E1,0 \
  -C9 -z200 -ub --splice-flank=yes \
  --cs=short --junc-bonus=0 --junc-bed juncs.bed \
  ref.fa test.fastq > aln_bonus_0.sam

paftools.js junceval annot.gtf aln_bonus_0.sam

# unmapped reads: 668070
# mapped reads: 979414
# primary alignments: 984638
# singletons: 246055
# predicted introns: 2896215
# non-overlapping introns: 4851
# correct introns: 2742705 (94.70%)

no splice preset, junc-bonus=9

minimap2 -a -t12 -k15 -w5 --splice \
  -g2000 -G200k -A1 -B2 -O2,32 -E1,0 \
  -C9 -z200 -ub --splice-flank=yes \
  --cs=short --junc-bonus=9 --junc-bed juncs.bed \
  ref.fa test.fastq > aln_bonus_9.sam

paftools.js junceval annot.gtf aln_bonus_9.sam

# unmapped reads: 668070
# mapped reads: 979414
# primary alignments: 984638
# singletons: 246055
# predicted introns: 2896215
# non-overlapping introns: 4851
# correct introns: 2742705 (94.70%)

splice preset, junc-bonus=0

minimap2 -x splice -a -t12 -k15 -w5 --splice \
  -g2000 -G200k -A1 -B2 -O2,32 -E1,0 \
  -C9 -z200 -ub --splice-flank=yes \
  --cs=short --junc-bonus=0 --junc-bed juncs.bed \
  ref.fa test.fastq > aln_bonus_0_splice.sam

paftools.js junceval annot.gtf aln_bonus_0_splice.sam

# unmapped reads: 666538
# mapped reads: 980946
# primary alignments: 986187
# singletons: 244051
# predicted introns: 2956054
# non-overlapping introns: 5016
# correct introns: 2932327 (99.20%)

splice preset, junc-bonus=9

minimap2 -x splice -a -t12 -k15 -w5 --splice \
  -g2000 -G200k -A1 -B2 -O2,32 -E1,0 \
  -C9 -z200 -ub --splice-flank=yes \
  --cs=short --junc-bonus=9 --junc-bed juncs.bed \
  ref.fa test.fastq > aln_bonus_9_splice.sam

paftools.js junceval annot.gtf aln_bonus_9_splice.sam

# unmapped reads: 666538
# mapped reads: 980946
# primary alignments: 986187
# singletons: 244051
# predicted introns: 2956054
# non-overlapping introns: 5016
# correct introns: 2932327 (99.20%)

I'm using minimap version 2.17-r941 from bioconda.

I've also attached a markdown file with some more examples that I tried:
junc_bed_test.txt

Thanks
Matt

[mparker2]

I'm not really a C programmer but I think that this can be solved by adding:

else if (c == 341) opt.junc_bonus = atoi(o.arg); // --junc-bonus

at or near line 212 in main.c?

[lh3]

Sorry for the late response. I think I have just fixed this issue. Let me know if it still doesn't work. Thanks.