BUSCO for bacterial lineages can't run on genomes with scaffolds >32000000 bases

[tkchafin]

Description of the bug

prodigal chokes on these, causing a BUSCO failure with an uninformative error message about prodigal failing to find any genes.

The solution was to break up long sequences, which I did using seqkit

cat in.fasta | seqkit sliding -W 31900000 -s 31900000 -g -j 8 > out.fasta

Command used and terminal output

Relevant files

No response

System information

No response

[muffato]

We then realised that the Busco output files are using the coordinates from the chopped up sequences. https://github.com/tkchafin/busco/blob/dev/bin/fix_windowed_busco.py is a script to change the coordinates back to the initial genome.

[muffato]

I found the reason: the pipeline runs Busco in batch mode, i.e. by passing all input Fasta files in a directory. In batch mode, Busco wants to run all genomes even if one fails so it captures and ignores the Prodigal did not recognize any genes matching the dataset archaea_odb10 in the input file error, whereas in single-Fasta mode, it re-raises the exception.

Comparing the Snakemake BTK pipeline, this behaviour is actually what we want since it catches and dismisses did not recognize any genes matching the dataset errors.