Generative models for quantification of DNA modifications

Research output: Chapter in Book/Report/Conference proceedingChapterScientificpeer-review


Research units

  • Flatiron Institute
  • New York University


There are multiple chemical modifications of cytosine that are important to the regulation and ultimately the functional expression of the genome. To date no single experiment can capture these separate modifications, and integrative experimental designs are needed to fully characterize cytosine methylation and chemical modification. This chapter describes a generative probabilistic model, Lux, for integrative analysis of cytosine methylation and its oxidized variants. Lux simultaneously analyzes partially orthogonal bisulfite sequencing data sets to estimate proportions of different cytosine methylation modifications and estimate multiple cytosine modifications for a single sample by integrating across experimental designs composed of multiple parallel destructive genomic measurements. Lux also considers the variation in measurements introduced by different imperfect experimental steps; the experimental variation can be quantified by using appropriate spike-in controls, allowing Lux to deconvolve the measurements and recover accurately the underlying signal.


Original languageEnglish
Title of host publicationMethods in Molecular Biology
Publication statusPublished - 1 Jan 2018
MoE publication typeA3 Part of a book or another research book

Publication series

NameMethods in Molecular Biology
PublisherHumana Press
ISSN (Print)1064-3745

    Research areas

  • 5-methylcytosine oxidation, Bayesian analysis, Bisulfite sequencing, BS-seq/oxBS-seq/TAB-seq/fCAB-seq/CAB-seq/redBS-seq/MAB-seq, DNA methylation, Hierarchical generative modeling

ID: 27419996