Search for: All records

Biological systems exhibit complex dynamics that differential equations can often adeptly represent. Ordinary differential equation models are widespread; until recently their construction has required extensive prior knowledge of the system. Machine learning methods offer alternative means of model construction: differential equation models can be learnt from data via model discovery using sparse identification of nonlinear dynamics (SINDy). However, SINDy struggles with realistic levels of biological noise and is limited in its ability to incorporate prior knowledge of the system. We propose a data-driven framework for model discovery and model selection using hybrid dynamical systems: partial models containing missing terms. Neural networks are used to approximate the unknown dynamics of a system, enabling the denoising of the data while simultaneously learning the latent dynamics. Simulations from the fitted neural network are then used to infer models using sparse regression. We show, via model selection, that model discovery using hybrid dynamical systems outperforms alternative approaches. We find it possible to infer models correctly up to high levels of biological noise of different types. We demonstrate the potential to learn models from sparse, noisy data in application to a canonical cell state transition using data derived from single-cell transcriptomics. Overall, this approach provides a practical framework for model discovery in biology in cases where data are noisy and sparse, of particular utility when the underlying biological mechanisms are partially but incompletely known. 

In animal models,Nipbldeficiency phenocopies gene expression changes and birth defects seen in Cornelia de Lange syndrome, the most common cause of which isNipblhaploinsufficiency. Previous studies inNipbl^+/−mice suggested that heart development is abnormal as soon as cardiogenic tissue is formed. To investigate this, we performed single-cell RNA sequencing on wild-type andNipbl^+/−mouse embryos at gastrulation and early cardiac crescent stages.Nipbl^+/−embryos had fewer mesoderm cells than wild-type and altered proportions of mesodermal cell subpopulations. These findings were associated with underexpression of genes implicated in driving specific mesodermal lineages. In addition,Nanogwas found to be overexpressed in all germ layers, and many gene expression changes observed inNipbl^+/−embryos could be attributed toNanogoverexpression. These findings establish a link betweenNipbldeficiency,Nanogoverexpression, and gene expression dysregulation/lineage misallocation, which ultimately manifest as birth defects inNipbl^+/−animals and Cornelia de Lange syndrome. 

Single-cell genomic technologies offer vast new resources with which to study cells, but their potential to inform parameter inference of cell dynamics has yet to be fully realized. Here we develop methods for Bayesian parameter inference with data that jointly measure gene expression and Ca²⁺dynamics in single cells. We propose to share information between cells via transfer learning: for a sequence of cells, the posterior distribution of one cell is used to inform the prior distribution of the next. In application to intracellular Ca²⁺signalling dynamics, we fit the parameters of a dynamical model for thousands of cells with variable single-cell responses. We show that transfer learning accelerates inference with sequences of cells regardless of how the cells are ordered. However, only by ordering cells based on their transcriptional similarity can we distinguish Ca²⁺dynamic profiles and associated marker genes from the posterior distributions. Inference results reveal complex and competing sources of cell heterogeneity: parameter covariation can diverge between the intracellular and intercellular contexts. Overall, we discuss the extent to which single-cell parameter inference informed by transcriptional similarity can quantify relationships between gene expression states and signalling dynamics in single cells. 

Abstract Myeloid-derived suppressor cells (MDSCs) play a prominent role in the tumor microenvironment. A quantitative understanding of the tumor–MDSC interactions that influence disease progression is critical, and currently lacking. We developed a mathematical model of metastatic growth and progression in immune-rich tumor microenvironments. We modeled the tumor–immune dynamics with stochastic delay differential equations and studied the impact of delays in MDSC activation/recruitment on tumor growth outcomes. In the lung environment, when the circulating level of MDSCs was low, the MDSC delay had a pronounced impact on the probability of new metastatic establishment: blocking MDSC recruitment could reduce the probability of metastasis by as much as 50%. To predict patient-specific MDSC responses we fit to the model individual tumors treated with immune checkpoint inhibitors via Bayesian parameter inference. We reveal that control of the inhibition rate of natural killer (NK) cells by MDSCs had a larger influence on tumor outcomes than controlling the tumor growth rate directly. Posterior classification of tumor outcomes demonstrates that incorporating knowledge of the MDSC responses improved predictive accuracy from 63% to 82%. Investigation of the MDSC dynamics in an environment low in NK cells and abundant in cytotoxic T cells revealed, in contrast, that small MDSC delays no longer impacted metastatic growth dynamics. Our results illustrate the importance of MDSC dynamics in the tumor microenvironment overall and predict interventions promoting shifts toward less immune-suppressed states. We propose that there is a pressing need to consider MDSCs more often in analyses of tumor microenvironments. 

ABSTRACT Cells do not make fate decisions independently. Arguably, every cell-fate decision occurs in response to environmental signals. In many cases, cell-cell communication alters the dynamics of the internal gene regulatory network of a cell to initiate cell-fate transitions, yet models rarely take this into account. Here, we have developed a multiscale perspective to study the granulocyte-monocyte versus megakaryocyte-erythrocyte fate decisions. This transition is dictated by the GATA1-PU.1 network: a classical example of a bistable cell-fate system. We show that, for a wide range of cell communication topologies, even subtle changes in signaling can have pronounced effects on cell-fate decisions. We go on to show how cell-cell coupling through signaling can spontaneously break the symmetry of a homogenous cell population. Noise, both intrinsic and extrinsic, shapes the decision landscape profoundly, and affects the transcriptional dynamics underlying this important hematopoietic cell-fate decision-making system. This article has an associated ‘The people behind the papers’ interview. 

Abstract Motivation Methods to model dynamic changes in gene expression at a genome-wide level are not currently sufficient for large (temporally rich or single-cell) datasets. Variational autoencoders offer means to characterize large datasets and have been used effectively to characterize features of single-cell datasets. Here, we extend these methods for use with gene expression time series data. Results We present RVAgene: a recurrent variational autoencoder to model gene expression dynamics. RVAgene learns to accurately and efficiently reconstruct temporal gene profiles. It also learns a low dimensional representation of the data via a recurrent encoder network that can be used for biological feature discovery, and from which we can generate new gene expression data by sampling the latent space. We test RVAgene on simulated and real biological datasets, including embryonic stem cell differentiation and kidney injury response dynamics. In all cases, RVAgene accurately reconstructed complex gene expression temporal profiles. Via cross validation, we show that a low-error latent space representation can be learnt using only a fraction of the data. Through clustering and gene ontology term enrichment analysis on the latent space, we demonstrate the potential of RVAgene for unsupervised discovery. In particular, RVAgene identifies new programs of shared gene regulation of Lox family genes in response to kidney injury. Availability and implementation All datasets analyzed in this manuscript are publicly available and have been published previously. RVAgene is available in Python, at GitHub: https://github.com/maclean-lab/RVAgene; Zenodo archive: http://doi.org/10.5281/zenodo.4271097. Supplementary information Supplementary data are available at Bioinformatics online.