Incorporating scale uncertainty in microbiome and gene expression analysis as an extension of normalization

Statistical normalizations are used in differential analyses to address sample-to-sample variation in sequencing depth. Yet normalizations make strong, implicit assumptions about the scale of biological systems, such as microbial load, leading to false positives and negatives. We introduce scale models as a generalization of normalizations, which allows researchers to model potential errors in these modeling assumptions, thereby enhancing the transparency and robustness of data analyses. In practice, scale models can drastically reduce false positives and false negatives rates. We introduce updates to the popular ALDEx2 software package, available on Bioconductor, facilitating scale model analysis.