Running on multi-fasta about segalign

[rsharris]

With the original lastz, you can input your target as a multi-fasta, however it strongly recommended that you do not provide a multi-fasta as a query.

I'm the guy who wrote lastz, and what I'm about to write only applies to lastz.

I think that (i.e. the notion that you should not provide a multi-fasta as a query) is a misconception resulting from what in retrospect was a poor human-interface design choice I made. Leaving the [multiple] designation off the query file is recommended, regardless of whether the query file has one or more sequences. So if target and query both have more than one sequence, this is the recommended command:
lastz target.fa[multiple] query.fa

The [multiple] designation tells lastz to collect all the sequences from that file into one long sequence in memory. It is necessary that lastz have all the target sequences in memory, but for the query it only needs them in memory one at a time.

It would have made more sense for lastz to figure out whether the file contained more than one sequence and do what it needed to do without the user having to tell it. Unfortunately I decided there was a cost in doing that. In retrospect that cost is extremely minimal — it was a bad choice on my part, and it has led to confusion.

[gsneha26]

Thank you @niederhuth, we are glad that the tool is so helpful for your research.

I am slightly confused by what you mean by would you advice breaking the query fasta down by each sequence in the fasta file.

If I understand your question correctly, the following should help:
For SegAlign, we recommend that you do provide multiple chromosomes for both, the target and the query. There are internal optimizations that overlap the different alignment stages for different chromosome-pairs which helps in maximum resource utilization and provides higher speedup. ( If you are interested in knowing more about these optimizations, you can have a look at our paper - https://doi.ieeecomputersociety.org/10.1109/SC41405.2020.00043 ).

About the run_segalign.sh script -- it essentially handles the seed-filter and extend stages, where the seed-filter stages are run using the segalign executable (GPU accelerated), the output segment files are generated, and then, the extend stage is run using lastz. In order to get the alignments and not just the segment files, you always need to use run_segalign.sh and not just the segalign executable.