Milot Mirdita

This is a translated search against a protein (profile) db, so this corresponds to blastx/2. (kind of I don’t think there is a actually translated profile search mode in blast)....

1) I am looking through the code and seeing some bugs in how coverage works within the alignment for taxonomy. Ignoring if this makes sense or not, its definitely broken...

First: both the sensitivity parameter and the iteration parameter do not do anything for nucleotide MMseqs2 searches. sensitivity is the parameter for adjusting the length of the similar k-mer lists,...

Are all of your sequences so short? I see average length 13, that might be an issue with MMseqs2. Are these protein or nucleotide sequences?

You can try to use the parameters we used in our [SpacePHARER manuscript](https://academic.oup.com/bioinformatics/article/37/19/3364/6207963) for finding CRISPR spacers with MMseqs2 (see the supplement). These should correspond to the following parameters: ```...

Could you try a clustering with `--min-seq-id 0.99`? Does that have the same issue?

If you give amino-acid sequences produced by prodigal to MMseqs2, it will not rerun any ORF calling.

I don't think you need much changing. The defaults are set to optimize sensitivity while keeping false positives low. You can further decrease the E-value threshold, but you can do...

The non-easy modules are intended to be used with MMseqs2 databases only. To use `search` please call `createdb` on the FASTA files first: ``` mmseqs createdb /mount-nfs/mydataset/single_train_sequences.fasta qdb mmseqs createdb...

Also I recommend to search with as many queries as possible. Single query searches are very slow (except if you do a very specialized setup, similar to our server setups,...