When performing scanpy.pp.highly_variable_genes, documentation states that “Depending on flavor, this reproduces the R-implementations of Seurat…” does this mean it reproduces the latest version of Seurat? I am wondering what best practice would be for finding variable features in an object with multiple samples (layers). In Seurat V5, when finding variable features with an object with 2 layers for example (2 samples), it identifies each layer’s variable features, find the common variable features, and then supplement until n variable features are present in a list using the most variable features in the two lists that are also features in the other matrices. This is the same as the method it uses for identifying features for integration. If we specify v3, is the same true? Or would it be best practice to add batch_key in the case of multiple samples?