Documentation and user control over PCA

[lwaldron]

The validate function and the function it calls .loadingCor use the first assay in a SummarizedExperiment to perform PCA. However, many users and standard single-cell workflows perform some kind of normalization (such as log-transformation) as a subsequent assay to the SummarizedExperiment (see for example https://bioconductor.org/books/release/OSCA.basic/normalization.html#normalization-transformation) and use normalized expression data for PCA. A few documentation + feature requests that came up:

• document in the validate function which assay of a SummarizedExperiment is used
• document what kind of normalization was used in training the RAVmodel, and what is recommended for user data
• document how PCA is done internally (it is different depending on input data type)
• provide an argument to validate that allows the user to specify which assay to use for PCA
• provide an argument to validate allowing the user specify the number of PCs from their dataset to compare against the RAVmodel (is 8 too small as a default?)

[lwaldron]

FYI @HelenaLC in case you have anything to add