Closed brendanwee closed 2 years ago
Also, do you have any insight on the amount of memory the estimation step would take with 2000+ coverage maps?
I would recommend running on all time points together. More samples should allow CoPTR-Ref to better estimate the position of the replication origin, and CoPTR-Contig to better reorder bins along the genome.
Also, do you have any insight on the amount of memory the estimation step would take with 2000+ coverage maps?
For the data in the paper (~1300 samples), I was able to run CoPTR on a 16GB machine. I tried to keep memory requirements low. In the estimate step, CoPTR groups coverage maps by species, writes them to disk, then loads them one at a time to estimate PTRs across samples. This way it doesn't need to store all coverage maps at once.
Awesome! Thank you Tyler!
I have a dataset with two timepoints. I am thinking that it is probably best to run the estimation on the full dataset (t0+t1) to give CoPTR the best chance at predicting PTRs (reordering bins). Are there any other assumptions in
estimate
that would suggest I separate the two timepoints during this step?