Some algorithms, such as RUV-III-NB, take integer counts as input and also output integer counts in gene space whereas others, such as scMerge, will produce logarithmic scale decimal numbers in gene space. May pseudoBulkDGE support this via a moderated t-test?