Our field has some very arbitrary threshold for noise on single features, sounds like there is some slightly more principled strategy in single cell genomics?