Dear Cactus developers,
I am planning to build a pangenome using cactus-pangenome for approximately 500 genomes, each around 2.7 Gb in size. However, I am concerned that this dataset may be too large and could require computational resources beyond what I currently have available.
My current server has 64 CPU cores and 1 TB of RAM. I would like to ask whether building such a pangenome directly with cactus-pangenome is feasible on this hardware. If it is feasible, would you have any recommendations for appropriate running parameters, such as --mgCores, --mapCores, --consCores, --indexCores, memory settings, filtering options, or output options?
My main downstream goals are structural variant analysis from the pangenome graph, for example using vg deconstruct or svwave, and potentially genotyping with tools such as PanGenie. Therefore, I would like to retain useful graph and VCF outputs while avoiding unnecessary resource usage.
If running the full dataset directly with cactus-pangenome is unlikely to be practical, do you have any suggestions for alternative strategies? For example, would you recommend reducing the number of assemblies, building chromosome-by-chromosome, using a subset first, changing output/indexing options, or using another pangenome construction approach?
Thank you very much for your time and advice.
Best regards,
Zhuoye Zheng
Dear Cactus developers,
I am planning to build a pangenome using cactus-pangenome for approximately 500 genomes, each around 2.7 Gb in size. However, I am concerned that this dataset may be too large and could require computational resources beyond what I currently have available.
My current server has 64 CPU cores and 1 TB of RAM. I would like to ask whether building such a pangenome directly with cactus-pangenome is feasible on this hardware. If it is feasible, would you have any recommendations for appropriate running parameters, such as --mgCores, --mapCores, --consCores, --indexCores, memory settings, filtering options, or output options?
My main downstream goals are structural variant analysis from the pangenome graph, for example using vg deconstruct or svwave, and potentially genotyping with tools such as PanGenie. Therefore, I would like to retain useful graph and VCF outputs while avoiding unnecessary resource usage.
If running the full dataset directly with cactus-pangenome is unlikely to be practical, do you have any suggestions for alternative strategies? For example, would you recommend reducing the number of assemblies, building chromosome-by-chromosome, using a subset first, changing output/indexing options, or using another pangenome construction approach?
Thank you very much for your time and advice.
Best regards,
Zhuoye Zheng