Home › Research › Microbiome

Microbiome

Computational methods for the microbiome: modeling its dynamics from sequencing data, and — with collaborators — correcting the biases and contamination that otherwise compromise microbiome analysis.

← Back to research

Microbial dynamics

Metagenomic sequencing captures only a snapshot of a living, changing ecosystem. The lab develops methods to recover its dynamics — how community composition, individual species' growth, and spatial structure change over time.

Highlighted work includes, for instance, modeling dynamics in the simplex of relative abundances (compositional Lotka–Volterra) and inferring mixed-population trajectories from longitudinal counts; reading personalized per-species growth rates from sequencing coverage (CoPTR); resolving gut-microbiome spatial structure; and MiSDEED, a synthetic-data engine for power analysis and study design. These are the lab's first-author lines of work in this area (Joseph, Chlenski, Pasarkar).

Compositional Lotka–Volterra describes microbial dynamics in the simplex

Joseph T, Shenhav L, Xavier J, Halperin E, Pe'er I · PLoS Comput. Biol. 16(5):e1007917 (2020)

Models community dynamics directly in the simplex of relative abundances, adapting Lotka–Volterra ecology to compositional data.

Efficient and accurate inference of mixed microbial population trajectories from longitudinal count data

Joseph TA, Pasarkar AP, Pe'er I · Cell Systems 10(6):463–469.e6 (2020)

Infers mixed microbial population trajectories from longitudinal count data, efficiently and accurately.

Directional Gaussian mixture models of the gut microbiome elucidate microbial spatial structure

Pasarkar AP, Joseph TA, Pe'er I · mSystems 6(6):e00817-21 (2021)

Directional Gaussian mixture models elucidate the spatial structure of the gut microbiome.

Accurate and robust inference of microbial growth dynamics from metagenomic sequencing reveals personalized growth rates (CoPTR)

Joseph TA, Chlenski P, Litman A, Korem T, Pe'er I · Genome Res. 32(3):558–568 (2022)

CoPTR reads per-species growth rates from sequencing-coverage peak-to-trough ratios, showing growth is personalized and disease-associated.

MiSDEED: a synthetic data engine for microbiome study power analysis and study design

Chlenski P, Hsu M, Pe'er I · Bioinformatics Advances 2(1):vbac043 (2022)

Generates synthetic longitudinal microbiome data under perturbation for power analysis, study design, and method benchmarking.

Methods developed with collaborators

In close collaboration with the labs of Tal Korem and Eran Halperin, the group has built widely used methods that make microbiome measurements trustworthy and comparable — for instance, removing contamination, correcting processing bias, guarding against inflated accuracy estimates, and tracing the sources that compose a community.

With the Korem lab

Contamination source modeling with SCRuB improves cancer phenotype prediction from microbiome data

Austin GI, Park H, Meydan Y, … Pe'er I, Uhlemann A-C, Shenhav L, Korem T · Nat. Biotechnol. 41(12):1820–1828 (2023)

SCRuB: a probabilistic decontamination method that shares information across samples and controls to remove contamination precisely.

Processing-bias correction with DEBIAS-M improves cross-study generalization of microbiome prediction models

Austin GI, Brown Kav A, ElNaggar S, … Pe'er I, Korem T · Nat. Microbiol. 10(4):897–911 (2025)

Corrects study-specific processing bias so microbiome-based prediction models transfer across cohorts.

Distributional bias compromises leave-one-out cross-validation

Austin GI, Pe'er I, Korem T · Sci. Adv. 11(48):eadx6976 (2025)

Shows how distributional bias inflates apparent accuracy under leave-one-out cross-validation in microbiome (and other) machine learning.

With the Halperin lab

FEAST: fast expectation-maximization for microbial source tracking

Shenhav L, Thompson M, Joseph TA, Briscoe L, Furman O, Bogumil D, Mizrahi I, Pe'er I, Halperin E · Nat. Methods 16(7):627–632 (2019)

FEAST estimates the contributions of thousands of candidate source environments to a microbial community, scalably and quickly.

Domain review

An introduction to whole-metagenome shotgun sequencing studies

Joseph TA, Pe'er I · Deep Sequencing Data Analysis (Methods Mol. Biol.), 107–122 (2021)

A primer on the design and analysis of whole-metagenome shotgun sequencing studies.

Pe'er Lab — Microbiome