When should I use HoneyBADGER about honeybadger

[JEFworks]

Both are very good questions and ones that we are addressing extensively in our paper revision. Briefly, yes, the 3' bias will limit the number of SNPs detectable and thus limit the power to identify smaller CNVs using the allele-based approach hb$calcAlleleCnvBoundaries. My benchmarks so far suggests that the method will still be useful for whole chromosome or chromosome arm level alterations. The expression-based approach hb$calcGexpCnvBoundaries would provide comparable resolution under the assumption that your 10X data contains primarily one cell type and thus an appropriate reference expression dataset for expression normalization can be identified. Otherwise, you may need to cluster your data by cell type first, and then normalize to the appropriate reference expression dataset. I am also currently benchmarking speed, which is tricky for 10X data because you have more cells but fewer SNPs and fewer genes. Speed is also highly dependent on the complexity of the subclonal structure.

I would always recommend assessing your data visually using hb$plotGexpProfile and hb$plotAlleleProfile as sanity checks.

[saeedsaberi]

Thanks for the useful and prompt comment. Last question, do you have a favourite normalization for 10X data?