Can I run HUMAnN with 12 GB RAM?

Microbial community profiling HUMAnN
1

Hi, I have the following configuration in my machine. Can run HUMAnN in this configuration with uniref90_ec_filtered database?
RAM: 12GB
Processors: Intel i7 ; 8 Cores (2.4 GHz x 8)

Thanks,
DC7

2

Hello DC7, Yes with the filtered database you might be able to run HUMAnN with 12GB of memory depending on the size of your input files. If you run out of memory it will likely be in the translated search portion of the workflow. You could run bypassing translated search which will not identify any “unclassified” UniRef sequences but would still generate gene families and pathways output files stratified by species.

Thank you,
Lauren

1 Like
3

Thanks a lot @lauren.j.mciver for your reply. But, if I bypass that step how it may affect the analysis? Will it give drastically different result? I mean, normally, around what percentage of the unaligned sequences are identified in the translated search?

May the reviewer consider the bypassing as a backdraw of the study?

4

Bypassing that step would prevent any unclassified proteins from being identified in the results. The total number of reads being passed to the translated search portion would depend on the origin of your samples (check out Fig. S3 of the HUMAnN 2.0 paper for details on runtime/alignment rates/memory based on total number of input reads and sample type). My suggestion would be to try running with translated search and see how the memory looks. You might be okay based on the smaller filtered ec database!

Thank you,
Lauren

1 Like
5

okay @lauren.j.mciver. Many thanks for your quick reply.

6

Hi @lauren.j.mciver !!! One additional question. In metaphlan, by default, reads shorter than 70bp are discarded. Does the same thing happen in HUMAnN 3.0?
Actually, we can remove reads shorter than 70bp in trimmomatic. But, I am loosing a very large number of sequences while running trimmomatic . Thus I am bypassing the trimmomatic and trf step. So, I am looking for some other way to remove sequences shorter than 70bp (or 45bp in some case).

Thanks

7

Hi, HUMAnN will not discard short reads. In kneaddata you could use the option --trimmomatic-options <options> to customize how trimmomatic is run. You should be able to modify this to fit the use case you need.

Thanks!
Lauren