new module: custom/bed12codonpositions

[pinin4fjords]

Why

A pipeline that needs codon-level genomic positions along a spliced transcript (ribo-seq P-site counts per codon, frame / periodicity QC, novel-ORF tiling) has nothing off-the-shelf to reach for. bedtools makewindows takes BED3 and ignores BED12 blocks; chaining bedtools bed12tobed6 | bedtools makewindows re-anchors each exon at offset 0, so codons crossing an intron land in the wrong frame. No existing nf-core module emits per-codon BED from a BED12.

What it does

For each BED12 record, walks the blocks in mRNA order (5'→3'), emits every --step-th mRNA position starting at --frame, and projects each back to genomic coordinates. With --width N > 1 it emits an N-nt span per position, splitting at block boundaries so a codon that crosses an intron becomes two rows whose union still maps to a contiguous mRNA region.

Worked example

Two-exon - strand record with one intron (spliced mRNA length 20):

chr1    100    210    geneX    500    -    100    210    0    2    10,10    0,100

Default args (--step 3 --width 1, one row per codon at its 5'-most nt) — i.e. the BED you'd intersect with an offset-corrected ribo-seq BAM to count P-sites per codon:

chr1    209    210    geneX    500    -
chr1    206    207    geneX    500    -
chr1    203    204    geneX    500    -
chr1    200    201    geneX    500    -
chr1    107    108    geneX    500    -
chr1    104    105    geneX    500    -
chr1    101    102    geneX    500    -

Rows are in 5'→3' mRNA order, so on the - strand the genomic coordinates count down. The jump from 200 back to 107 is the spliced intron.

With --width 3 (full 3-nt codon spans), the codon at mRNA position 9 straddles the intron — its 5' nucleotide is in the upstream exon and the other two are in the downstream exon, so it's split into two BED rows (4th and 5th below):

chr1    207    210    geneX    500    -
chr1    204    207    geneX    500    -
chr1    201    204    geneX    500    -
chr1    200    201    geneX    500    -
chr1    108    110    geneX    500    -
chr1    105    108    geneX    500    -
chr1    102    105    geneX    500    -

Score (column 5) is preserved. Frame is taken from the record's own start, so the module works regardless of any GTF phase annotation — running it three times with --frame 0/1/2 gives you the three per-frame BEDs needed for ribo-seq periodicity QC.

I/O

• Input: tuple val(meta), path(bed12)
• Output: tuple val(meta), path("${prefix}.bed") (BED6) + versions topic
• ext.args: --frame INT (0), --step INT (3), --width INT (1), --keep-duplicates

Container

Wave-built community.wave.seqera.io/library/python_pandas_pyyaml:75514f9f977be607 (with matching community-cr-prod.seqera.io singularity blob).

Test plan

• nf-test (default, width 3, frame 1, intron fixture at width 1 and 3, keep-duplicates, stub)
• nf-core modules lint clean (2 Wave-container false positives)
• pre-commit clean
• CI green