runEigDecomp function results in NaN
marvinquiet opened this issue · comments
Thank you for your great work in developing such a wonderful pipeline for analyzing scATAC-seq.
When using it from my own experience, I noticed that in the runEigDecomp function under https://rdrr.io/github/r3fang/SnapATAC/src/R/utilities.R. The transitions <- as.matrix(d_rot1 %*% norm_p1 %*% d_rot1);
may generate NaNs which hinders the calculation on obj@smat@sdev
. This can result in an incorrect diffusion map.
I wonder if I can simply replace those NaNs as 0s? Thank you in advance!