Closed privefl closed 3 years ago
Hi @privefl thanks for raising this point. In the latest version of the manuscript, we assume pgen files, to be consistent with the latest version of the package. To use bed files, one may need to use an earlier version of the package as in branch V1.0.
Thanks for your reply. In the latest version of the manuscript, you still talk about doing some optimized matrix computations because you have only 0s, 1s, 2s and NAs. Are you still using this when using pgen files? Do you handle dosage data?
Moreover, you still talk about
Moreover, the existing convenient data interface provided by the BEDMatrix package further facilitates our implementation.
whereas BEDMatrix does not handle pgen format yet, but only bed format for now, if I'm not mistaken.
Good catches. I take back the previous reply and sorry for the confusion. In fact the description and results in the paper should be all about the earlier version of the package https://github.com/junyangq/snpnet/tree/V1.0. We will correct the pgen assumption in the paper and add discussion about major changes and further optimizations made for the latest implementation.
Thank you @privefl for pointing out this issue!
Could I ask how far is it from being published?
It's under review after a major revision.
I think the preprint says that snpnet can be used on bed files directly. However, the documentation seems to talk only about pgen files.
Could you please clarify this point?