CDHIT para and cdhit slow speed

Question

[ahteshamabbasi1996]

I am writing to seek assistance regarding the usage of CD-HIT software for clustering a dataset of 135,000 nucleotide sequences. Currently, I am working on a cluster with 16 CPUs, and the maximum time limit available on this cluster is one week.

I have attempted to improve the performance of the CD-HIT process by employing various parameters, such as -T and -M, but unfortunately, none of them have proven to accelerate the execution time significantly.

Moreover, I have come across the option of using cd-hit-para, but I am uncertain about its usage since it requires specifying IP addresses, whereas I only have information regarding the number of CPUs available on the cluster.

I would greatly appreciate any assistance or guidance you can provide to help optimize the CD-HIT clustering process in my current setup.

Thank you in advance for your support.

[biofalconch]

Maybe not a solution but have you thought about using other clustering tools (e.g. UCLUST or MMseqs2)?

Answer 1

[Mensur Dlakic]

I think cd-hit-est is used for DNA sequences, but you should verify.

It is not just how many sequences you have - 135,000 is actually a small(ish) number - but how long they are. If they are prokaryotic genome-size long, or worse yet if they are eukaryotic chromosomes, it might take a very long time regardless of how many CPUs are available. Generally speaking, cd-hit is not meant for very long sequences.

[ahteshamabbasi1996]

Yes, you are right.

In my case, cd-hit-est is being used for nucleotide sequences.

These sequences are actually long eccDNA sequences belonging to HM. But even in 7 days, it is able to cluster only a few thousand sequences.

I am not sure what would be the role of -T, and -M parameters, if they can't provide any sort of acceleration to the clustering process.

Answer 2

[meyezili]

You can find some explanations for parameters in http://www.bioinformatics.org/cd-hit/cd-hit-user-guide.

By setting the CD-HIT parameter -T 0, all CPUs defined in the SLURM script will be used. Setting the parameter -M 0 allows unlimited usage of the available memory.

-T functions well in my work.