Feature Request: Inner-join for `modkit bedmethyl merge`

[SuhasSrinivasan]

Summary

modkit bedmethyl merge performs an outer join: a position is retained if it is present in any input.
When merging biological/technical replicates, it would be valuable to instead keep only positions that are reproducible across samples (present in at least N inputs), optionally with a minimum per-sample valid coverage.

Motivation

Pooling replicates with the current outer join keeps single-replicate positions, which can include non-reproducible/low-confidence calls. After the merge, replicate provenance is lost (the merged bedMethyl has no "number of contributing samples" field), so this filtering cannot be done downstream from the merged file alone. Doing it inside merge (which already reads each input per region via the tabix index) is natural and efficient.

Proposed feature

Two optional flags on bedmethyl merge, fully backward-compatible (defaults reproduce current outer join):

• --min-samples <N> (default 1): output a position only if it appears in at least N input files. Setting N = number of inputs performs an inner join (keep only positions present in all samples).
• --min-sample-coverage <C> (default 0): an input only counts toward a position (both for the --min-samples tally and for the summed counts) when that input's record has at least C valid coverage.

Example (3 replicates, require presence in all three with >= 5 valid coverage each):

modkit bedmethyl merge rep1.bed.gz rep2.bed.gz rep3.bed.gz \
  -o replicates_inner.bed -g genome_sizes.tsv \
  --min-samples 3 --min-sample-coverage 5