MetaPhlAn 4 appropriate for shallow shotgun sequenced data?

I was just wondering whether MetaPhlAn 4 supports shallow shotgun sequenced data.
What depth at the lowest does the MetaPhlAn 4 support for taxonomic profiling and other downstream analysis?

For example, if the mean read depth of at least 2 million sequences per sample were performed for sequencing, would it be okay to use MetaPhlAn 4 for taxonomic profiling at this read depth or does it require more deeper read depth?

Thank you.

It depends on the complexity of your microbiome community and the relative abundance of species in it. Are all species equally present or are some present at much less abundance than others? Can you sequence two or three of your biological samples much deeper (e.g. 100 million reads) and then randomly subsample FASTQ files to 2 million reads each to compare the number of species detected? That will inform you for the ecosystem of interest what read depth you need to be comprehensive.