In the SCTransform vignette, inside Seurat (https://satijalab.org/seurat/v3.1/sctransform_vignette.html), they talk about choosing principal components, and how you have more leeway to select more using SCtransform, than you did with traditional Seurat methods. However, they don't talk about how to go about choosing so many PCs. Do you still use the same approaches such as JackStraw, Elbow Plots, or do people arbitrarily choose a high number.