ncbi-genome-download VS UPIMAPI

If you have the taxonomic IDs of your organisms of interest, there are existing parallelized download tools that are more efficient like https://github.com/kblin/ncbi-genome-download or bit-dl-ncbi-assemblies from https://github.com/AstrobioMike/bit

Downloading from NCBI can be pretty wonky though, but third parties have made good programs to deal with. As an example - and a personal favorite - you have https://github.com/kblin/ncbi-genome-download which by far is the easiest way to download various assembled genomes.

I know you said Ensembl, but if you can live with NCBI, I would suggest https://github.com/kblin/ncbi-genome-download

Useful tool to download from NCBI: ncbi-genome-download

You may run MOSCA (https://github.com/iquasere/MOSCA), it performs all major steps of metagenomics analysis. It includes that functional classification you are looking for, since with UPIMAPI (https://github.com/iquasere/UPIMAPI) it annotates with UniProt DB as reference, and obtains information including taxonomy, EC numbers, and even those GOs, and reCOGnizer (https://github.com/iquasere/reCOGnizer), which annotates with CDD DB as reference, and obtains orthologous groups information (COG, Pfam, etc).

Prediction of coding sequences takes as input the contigs you obtained, and gives you the translated genes. Besides annotating with the KEGG database, you may also want to annotate with more general purpose databases (e.g. UniProt), as these provide more taxonomies and functional information. MOSCA includes UPIMAPI (https://github.com/iquasere/UPIMAPI) and reCOGnizer (https://github.com/iquasere/reCOGnizer), which annotate genes with reference to UniProt and CDD databases using two different methods, providing complementary information. This is the same methodology used by widely popular tools such as eggNOG-mapper and Prokka, but these use other databases.

Install with mamba instead of conda, no more problems. Or use UPIMAPI (https://github.com/iquasere/UPIMAPI) together with reCOGnizer (https://github.com/iquasere/reCOGnizer), since these tools obtain better results when annotating proteins