MUSiCC for Metagenomes and Lior Pachter's Code for Transcriptomes

MUSiCC for Metagenomes and Lior Pachter's Code for Transcriptomes

Lior Pachter is having a discussion on the mathematical aspects of digital normalization (Digital normalization revealed). This is related to his work with David Tse’s group and we posted the relevant tutorial slides here. Pachter and his student Sreeram are developing a code to analyze de novo assembled RNAseq data.

In the context of doing similar mathematical analysis for metagenomes, the readers may find the following paper useful.


Functional metagenomic analyses commonly involve a normalization step, where measured levels of genes or pathways are converted into relative abundances. Here, we demonstrate that this normalization scheme introduces marked biases both across and within human microbiome samples and systematically identify various sample- and gene-specific properties that contribute to these biases. We introduce an alternative normalization paradigm, MUSiCC, which combines universal single-copy genes with machine learning methods to correct these biases and to obtain a more accurate and biologically meaningful measure of gene abundances. Finally, we demonstrate that MUSiCC significantly improves downstream discovery of functional shifts in the microbiome. MUSiCC is available at

Written by M. //