search

getian107 / PRScsx

Cross-population polygenic prediction
MIT License
65 stars 20 forks source link

Computation time and memory requirements #15

Closed opain closed 2 years ago

opain commented 2 years ago

Dear Tian Ge,

Thank you and your team for developing this exciting method!

Please can you indicate the required memory and computation time to run PRS-CSx, for example without parallel computing for one phi parameter on chromosome 1 (HapMap3 only), with default settings but --meta=True? Or please point me towards previous documentation regarding this aspect of PRS-CSx.

I imagine PRS-CSx with two sets of sumstats and --meta=True will be use slightly more memory and run for three times as long as PRScs with one set of sumstats?

Many thanks,

Ollie

getian107 commented 2 years ago

Hi Ollie- PRS-CSx with two sets of sumstats will roughly double the runtime and memory requirements of PRS-CS if the same number of MCMC iterations is used. We now recommend running a longer chain for PRS-CSx to improve convergence. For two sets of sumstats, if the length of the MCMC chain is doubled (default setting), the computation time is ~4x. --meta=True adds negligible computational cost.

opain commented 2 years ago

Great! Thank you Tian for your clear and prompt response.