DIAMOND is a fast replacement for BLASTX and BLASTP.
Use DIAMOND in BLASTX mode to compare your microbiome short read datasets against the NR database.
This will allow MEGAN to bin your reads both taxonomically and also by function.
While most competing approaches either avoid alignment or use DNA alignment, we favor an protein-alignment-based approach for two reasons: (1) current databases only represent a small fraction of the microbial DNA that exists in the environment. So, unless you are studying a very well studied environment, there is a lot of stuff that you won't be able to bin. (2) Functional content and proteins are usually of great interest in a study.
When using DIAMOND, set it to output a DAA file and then use MEGAN's meganizer program to analyze the DAA file. Meganizer does not produce a new file but rather adds an additional block of classifications and indices to the end of the DAA file. That file can then be opened in MEGAN. This is the fastest and most space-efficient way process the output of DIAMOND.
MALT also does BLASTX and BLASTP mode alignments, but the program's memory usage doesn't scale favorably with the increasing size of protein databases. However, results published recently in Nature (Neanderthal diet) were obtained with the help of MALT in BLASTX mode.
MALT is more useful in the content of BLASTN mode alignment. In is particularly useful in the context of ancient DNA and a manuscript demonstrating this is available (Ancient Salmonella enteric genomes). A useful feature here is that MALT can do semi-global alignment. Also, MALT can produce an RMA file as output, which can immediately be opened in MEGAN. MEGAN's SAM output feeds into tools for processing ancient DNA such as ones developed in Johannes Krause's group.