I don’t why my HUMAnN become much slower in recent days. For the same single sample, it took about 5h to finish the analysis before, but now it takes over 24h. My humann was installed by pip3, and I installed the Anaconda recently in order to install the fastqc by conda, which I am not sure whether caused the slow. I tried to deactivate the conda environment, but it doesn’t solve the problem.
Hi, Did anything change in your HUMAnN install? Or are there any new processes running on the same machine? If you updated your HUMAnN install from the demo to full databases or if there are other processes running on your machine, either of those could cause runs to take longer. Another possibility is were you using multiple cores prior but now running on a single core? Let me know if any of those sound like they might be possible!
Thanks so much for your prompt response.
I first ran my samples with the full uniref90 database, it took about 5 hours per sample. I deleted the diamond outputs in the temp_folder and rerun the samples with the full uniref50 database with --resume option. I expect the analysis should be less than 5 hours as it bypassed Metaphlan and Bowtie, and also the uniref50 database is smaller than uniref90 database.
My computer has nothing changed except installing the Anaconda installer and fastqc (conda). No other program is running when I run HUMAnN and I do use same threads as before.
Hi Chao, Thanks for the information! You are totally right that the UniRef50 database is smaller but when running with that database HUMAnN allows for less restrictive matching so you will likely have many more alignments for HUMAnN to process and with more alignments the runs should take longer. You can check the intermediate outputs to confirm they are larger than your prior runs.
Please post if you have any additional questions.
That makes sense. Thanks so much for clarifying.