mnem

mnem extends Nested Effects Models to analyze single cell perturbation data by simultaneously clustering cell populations into k clusters and inferring k causal networks linking perturbed genes for each cluster.

Key Features:

• Mixture Model Clustering: Simultaneously clusters cell populations into k clusters while inferring separate causal networks for each cluster.
• Expectation Maximization Inference: Infers mixture components via an expectation maximization algorithm that iteratively updates optimal networks and cell probabilities.
• Multi-trait Read-out Support: Handles multi-trait observed read-outs including gene expression profiles from single cell perturbation experiments.

Scientific Applications:

• Perturb-Seq Data Analysis: Analyzes single cell perturbation data generated by Perturb-Seq experiments where individual cells are perturbed by knock-down of specific genes.
• Crop-Seq Data Analysis: Analyzes single cell perturbation data generated by Crop-Seq experiments with pooled CRISPR screens.
• Heterogeneous Cell Population Analysis: Identifies different cellular subpopulations and their corresponding causal networks to explain heterogeneity in cell populations.
• Cancer Research: Helps elucidate intra-cellular networks to improve understanding and treatment of diseases like cancer.

Methodology:

mnem uses a mixture of Nested Effects Models with an Expectation Maximization scheme that assigns each cell to a network with a certain probability and iteratively updates the optimal networks and cell probabilities.