Closed XuewenWangUGA closed 2 years ago
Am having the same issue with pggb. Tried with 13 genomes each ~45Mbp on an HPC cluster but the tasks keep dying. I was hoping to see an answer on this issue here :(
As PGGB is getting more and more stable, we will extensively test this in the near future. We will suggest default parameters for different species like human, yeast, arabidopsis, mouse.
Key parameters which will eat most of your RAM are -w, --block-weight-max N
and -T, --poa-threads N
. Setting the first one to 170000
and the second one to 24
on a 48 threads node with 128GB RAM, we are able to produce chromosomal pangenome graphs for ~40 human assemblies plus the two reference grch38
and chm13
.
I hope this gives you a hint.
Dear Sir/Madam,
Is there any information on how to set up the memory and threads of pggb? For example, there are seven genomic sequences, each with 100 M bp. For a fast analysis, there is a balance between the memory for each thread and the number of threads. How to balance it? What is the mimimun memory for each thread, e.g. 4 G? More threads will play more important role than the the memory of each thread in the analysis? Thanks.