Extract_markers.py: input database path format

Microbial community profiling StrainPhlAn
1

From the extract_markers.py script docs, it is not clear how the --database path should be formatted. I originally tried using 4.0.3, which is the directory name holding all of the bowtie2 db files, but I received the error: Could not locate a Bowtie index corresponding to basename "4.0".

After looking at the generate_markers_fasta function, I see that one must supply the bowtie2 database basename + a file extension (e.g., mpa_vJan21_CHOCOPhlAnSGB_202103.md5). It would be helpful if the extract_markers.py script docs included that info.

Notably, for metaphlan, I could just use the directory containing the bowtie2 database (e.g., 4.0.3), so the UI seems to differ between metaphlan and extract_markers.py.

2

Also, it appears that extract_markers.py assumes that the *.pkl database file is bzip-compressed:

    def load_database(self, verbose=True):
        """Loads the MetaPhlAn PKL database"""
        if self.database_pkl is None:
            if verbose:
                info('Loading MetaPhlAn {} database...'.format(self.get_database_name()))
            self.database_pkl = pickle.load(bz2.BZ2File(self.database))
            if verbose:
                info('Done.')

Maybe a try - except would be helpful here, in order to allow for the input of an uncompressed pickle file? Does bzip2 compression really help reduce the size of the pickle file?

3

Hi @nick-youngblut
For all strainphlan-related scripts, the --database should point to the metaphlan PKL database (that is always bz2 compressed when exported by us even if not in the file extension). I will update the docs to make this fact clearer

4

Thank you for updating the docs. I’m surprised that bzip2 compression helps with pickled files, since they are already binary. Regardless, thank you for clarifying!