Single cell network analysis with a mixture of Nested Effects Models

Abstract

MotivationNew technologies allow for the elaborate measurement of different traits of single cells under genetic perturbations. These interventional data promise to elucidate intra-cellular networks in unprecedented detail and further help to improve treatment of diseases like cancer. However, cell populations can be very heterogeneous.ResultsWe developed a mixture of Nested Effects Models (M&NEM) for single-cell data to simultaneously identify different cellular subpopulations and their corresponding causal networks to explain the heterogeneity in a cell population. For inference, we assign each cell to a network with a certain probability and iteratively update the optimal networks and cell probabilities in an Expectation Maximization scheme. We validate our method in the controlled setting of a simulation study and apply it to three data sets of pooled CRISPR screens generated previously by two novel experimental techniques, namely Crop-Seq and Perturb-Seq.Availability and implementationThe mixture Nested Effects Model (M&NEM) is available as the R-package at https://github.com/cbg-ethz/mnem/.Supplementary information Supplementary data are available at Bioinformatics online.

Analysis of mutational history of multidrug‐resistant genotypes with a mutagenetic tree model

“…We use a subsampling approach to balance the data by resistance and susceptible genotypes, account for uncertainty in the two classes, and increase the probability to find the optimal tree based on previous experience with such models. [19][20][21] The subsampling leaves us with 1000 trees, each of them containing a different set of up to 25 mutations. The trees are composed of directed edges with two mutations adjacent to each edge, a parent and a child mutation.…”

Hierarchy of TGFβ/SMAD, Hippo/YAP/TAZ, and Wnt/β-catenin signaling in melanoma phenotype switching

“…We applied NEM implemented in the Bioconductor R package “mnem” ( Pirkl & Beerenwinkel, 2018 ) to exhaustively compute the optimal network of the three S-genes TGFβ, siLATS1/2 and Wnt-3a, and the bootstrap support of all edges.…”

Inferring perturbation profiles of cancer samples

“…NEM and its extensions have been applied to various perturbation datasets. Most recent versions of the algorithm have been extended to combinatorial perturbations ( Pirkl et al, 2016 , 2017 ) probabilistic perturbations ( Srivatsa et al, 2018 ), time-series data ( Anchang et al , 2009 ; Froehlich et al, 2011 ; Wang et al, 2014 ), hidden player inference ( Sadeh et al, 2013 ), context specific signaling ( Sverchkov, 2018 ) and single cell perturbations ( Anchang et al , 2018 ; Pirkl and Beerenwinkel, 2018 ). NEM π is related to NEMiX ( Siebourg-Polster et al, 2015 ).…”