MDPBiome

MDPBiome applies Markov Decision Processes (MDPs) to model longitudinal metagenomics data and prescribe optimal perturbation sequences to steer microbial communities toward defined goal states.

Key Features:

• Statistical Modeling: Models longitudinal metagenomics samples with Markov Decision Processes to simulate microbial state transitions in response to external perturbations.
• Policy Prescriptions: Suggests optimal sequences of perturbations required to transition a microbiome from a current state to a predefined goal state.
• Intermediate State Estimation: Estimates intermediate microbial states along transition paths to identify and avoid undesirable or unhealthy configurations.
• Robustness Analysis: Performs robustness analysis of prescribed policies to evaluate reliability across different scenarios and datasets.
• Visualization Tools: Provides visualization of state transitions to aid interpretation of microbiome dynamics.

Scientific Applications:

• Medicine: Suggests low-impact clinical interventions to achieve or maintain healthy microbial populations, exemplified by recommendations to avoid specific perturbations in bacterial vaginosis studies of vaginal microbiomes.
• Bioremediation and Industrial Scenarios: Predicts effects of external factors on microbial communities for applications in bioremediation and industrial microbiome management.
• Vaccine and Developmental Microbiome Studies: Provides insights into how interventions such as salmonella vaccines can accelerate gut microbiome maturation in chicks.

Methodology:

Constructs Markov Decision Process models from longitudinal metagenomics data to simulate impacts of perturbations, identifies optimal intervention policies from those models, and assesses policy robustness.