search

getian107 / PRScsx

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

Performance improvements? #26

Open mglev1n opened 1 year ago

mglev1n commented 1 year ago

Although PRScsx allows for parallel computations of PRS across chromosomes, the time to generate score across all chromosomes is still substantial (eg. several hours of parallel computation). This is quite reasonable when dealing with a small number of traits, but for larger-scale analyses this becomes computationally prohibitive. Reducing the number of MCMC iterations seems like one way to improve performance, but I'm not sure about the "safety" of this approach. Are there other performance optimizations being considered for the future?

getian107 commented 1 year ago

Hi - Reducing the number of MCMC iterations will likely slightly reduce prediction accuracy. You might want to benchmark on a small number of traits. We are working on faster MCMC samplers and model fitting algorithms. This will take some time but hopefully will improve computational efficiency in the near future.