Stelling, J. | Search bioRxiv

bioRxiv2018

A simple and flexible computational framework for inferring sources of heterogeneity from single-cell dynamics

The availability of high-resolution single-cell data makes data analysis and interpretation an important open problem, for example, to disentangle sources of cell-to-cell and intra-cellular variability. Nonlinear mixed effects models (NLMEs), well established in pharmacometrics, account for such multiple sources of variations, but their estimation is often difficult. Single-cell analysis is an even more challenging application with larger data sets and models that are more complicated. Here, we show how to leverage the quality of time-lapse microscopy data with a simple two-stage method to estimate realistic dynamic NLMEs accurately. We demonstrate accuracy by benchmarking with a published model and dataset, and scalability with a new mechanistic model and corresponding dataset for amino acid transporter endocytosis in budding yeast. We also propose variation-based sensitivity analysis to identify time-dependent causes of cell-to-cell variability, highlighting important sub-processes in endocytosis. Generality and simplicity of the approach will facilitate customized extensions for analyzing single-cell dynamics.

systems biology↗

bioRxiv2017

Predicting Ligand-Dependent Tumors From Multi-Dimensional Signaling Features

Targeted therapies have shown significant patient benefit in about 5-10% of solid tumors that are addicted to a single oncogene. Here, we explore the idea of ligand addiction as a driver of tumor growth. High ligand levels in tumors have been shown to be associated with impaired patient survival, but targeted therapies have not yet shown great benefit in unselected patient populations. Using a novel approach of applying Bagged Decision Trees (BDT) to high-dimensional signaling features derived from a computational model, we can predict ligand dependent proliferation across a set of 58 cell lines. This mechanistic, multi-pathway model that features receptor heterodimerization, was trained on seven cancer cell lines and can predict signaling across two independent cell lines by adjusting only the receptor expression levels for each cell line. Interestingly, for patient samples the predicted tumor growth response correlates with high growth factor expression in the tumor microenvironment, which argues for a co-evolution of both factors in vivo.\n\nSummaryPrediction of ligand-induced growth of cancer cell lines, which correlates with ligand-blocking antibody efficacy, could be significantly improved by learning from features of a mechanistic signaling model, and was applied to reveal a correlation between growth factor expression and predicted response in patient samples.

systems biology↗