Quick start · hspZ

Preview release. This example is deliberately one chromosome pair and one strand. Whole genomes need partition-aware orchestration.

1. Generate HSPs

The alignment profile used here is K=2400, L=3000, H=2000, Y=9400. K belongs to hspZ, so it must be passed explicitly—hspZ’s default is 3000.

docker run --rm --gpus all \
  -v "$PWD:/data" \
  ghcr.io/hillerlab/hspz:0.0.4 \
  hspZ run \
    --reference /data/reference.2bit \
    --query /data/query.2bit \
    --output /data/segments \
    --strand plus \
    --hspthresh 2400 \
    --time

For a non-empty first interval this writes segments/tmp1.block0.r0.plus.segments. No output file is created for an interval with no accepted HSP.

2. Run LASTZ from that segment set

Select the same reference and query records named inside the segment file.

lastz reference.2bit/chrRef query.2bit/chrQuery \
  --segments=segments/tmp1.block0.r0.plus.segments \
  --strand=plus \
  --gappedthresh=3000 \
  --inner=2000 \
  --ydrop=9400 \
  --allocate:traceback=1.99G \
  --format=axt+ \
  --output=chrRef.chrQuery.plus.axt

The parameter ownership is load-bearing:

parameter	value	consumer
K	2400	`hspZ run --hspthresh 2400`
L	3000	`lastz --gappedthresh=3000`
H	2000	`lastz --inner=2000`
Y	9400	`lastz --ydrop=9400`

hspZ short -H means --hspthresh—LASTZ K—not LASTZ --inner.

Whole genomes

A whole-genome .segments directory contains multiple partitions and both strands. Run LASTZ once per partition, split mixed files by reference/query chromosome pair, select those two 2bit records, and pass the strand encoded in the filename/record. Do not feed two whole-genome 2bit files and one mixed segment file to a single LASTZ process.

The hspZ backend under development in make_lastz_chains is the reference orchestration. It becomes the recommended public path only after that backend is released there.