Outputs

Files

The default sink creates a directory containing non-empty files named:

tmp<n>.block<query-bin>.r<reference-bin>.plus.segments
tmp<n>.block<query-bin>.r<reference-bin>.minus.segments

n is the one-based query-interval number. -D adds .split1, .split2, … before .segments; -Z writes the same logical entries directly into one reproducible .tar.gz archive with pinned mtimes, <output>.tar.gz unless a path follows -Z.

With --query-list, each job writes this layout into its own directory, OUT/000001/, OUT/000002/, … (or its own OUT/000001.tar.gz, … with -Z), and OUT/queries.tsv lists the jobs once all of them have finished (see GPUs and seeding).

Record format

Each tab-separated line has eight fields:

reference  ref_start  ref_end  query  query_start  query_end  strand  score

Coordinates are chromosome-relative, one-based, and inclusive at both ends. Minus-strand records use the reverse-complement query frame and are emitted in reverse anchor order, matching KegAlign/LASTZ expectations.

Ordering and deduplication

Within each MAX_HITS chunk, accepted GPU HSPs are stably sorted by diagonal, adjacent contained/duplicate records are removed with KegAlign’s previous-input semantics, and survivors are stably sorted into LASTZ query-major order before coordinate conversion.