Transcriptomic forecasting with neural ordinary differential equations

Erbe, Rossin; Stein-O’Brien, Genevieve; Fertig, Elana J.

doi:10.1016/j.patter.2023.100793

Cited by 6 publications

(28 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Beyond filling the gaps between measurement times, computational techniques for temporal analysis of cellular states provide the potential to predict the future state of multicellular systems. While machine learning from single-cell datasets can make these predictions for individual cell types 9,10 , they cannot account for changes resulting from cell-cell interactions. Moreover, these predictive methods are data-driven and unable to predict cellular states from prior biological knowledge or mechanism alone.…”

Section: Introductionmentioning

confidence: 99%

Digitize your Biology! Modeling multicellular systems through interpretable cell behavior

Johnson,

Stein-O’Brien,

Booth

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

Cells are fundamental units of life. Recent technical advances have revolutionized our ability to quantify the state and identity of individual cells, and intercellular regulatory programs. However, these static measurements alone are limited in their ability to predict the complex collective behaviors that emerge from populations of many interacting cells over time. Mathematical models have a proven record of successfully predicting the behaviors of dynamic biological systems, e.g., therapeutic responses in cancer. Simulations from these models enable in silico visualization, examination, and refinement of biological models and can be used to generate new hypotheses about cells and their collective behavior. Agent-based modeling is particularly well-suited to studying communities of interacting cells, as it is intuitive to map a single cell to a single agent. Thus, we have developed a conceptual framing (with a reference implementation in the widely-used PhysiCell agent-based modeling framework) that can be initialized directly from single cell and spatial transcriptomic data, and that can be easily populated with interactive rules. Because the expert mathematical and computational knowledge required to build agent-based models has limited their widespread adoption in the biomedical research community, we engineered this framework to specify complex cellular responses to signals (or stimuli) using a single line of human readable text. This plain language text encodes cellular phenotypes and regulatory mechanisms from high throughput data and published literature, using a novel concept of hypothesis grammar. We motivate and fully describe this grammar and its philosophy, and then present a series of five example reference models of tumor growth and response to immunotherapy. Biologically, these examples demonstrate how mathematical models can predict from single cell and spatial transcriptomic data the cellular phenotypes responsible for tumor cell invasion and the simulation of immunotherapy treatment to overcome tumor cell growth. Computationally, these examples are designed to demonstrate how this conceptual framing and software implementation empower interdisciplinary teams to build an agent-based model of their experimental system, levering prior biological knowledge alone or in combination with information from spatial multi-omics technologies. Altogether, this approach provides an interface to bridge biological, clinical, and systems biology researchers for mathematical modeling of biological systems at scale, allowing the community to extrapolate from single-cell characterization to emergent multicellular behavior.

show abstract

Section: Introductionmentioning

confidence: 99%

Digitize your Biology! Modeling multicellular systems through interpretable cell behavior

Johnson,

Stein-O’Brien,

Booth

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Given that many dynamical systems can be described using ordinary differential equations (ODEs), a logical approach to modeling GRNs is to estimate ODEs for gene expression using an appropriate statistical learning technique [3][4][5][6]. Although estimating gene regulatory ODEs ideally requires time-course data, obtaining such data in biological systems might be difficult.…”

Section: Introductionmentioning

confidence: 99%

“…However, such methods impose several restrictions on the ODEs and cannot flexibly adjust to situations where the underlying assumptions do not hold; this increases the risk of model misspecification and hinders scalability to large networks, particularly given the enormous number of parameters necessary to specify a genome-scale model [10,11]. Other methods including Dynamo, PRESCIENT, and RNA-ODE are based on non-parametric approaches to learning regulatory ODEs, using tools such as sparse kernel regression [3], random forests [5], variational auto-encoders [7,12,13], diffusion processes [4], and neural ordinary differential equations [6,14], but these fail to include biologically relevant associations between regulatory elements and genes as constraints on the models. These latter models can be broadly placed into two classes based on the inputs required to estimate the gradient f of the gene regulatory dynamics, where f (x) = dx dt .…”

Section: Introductionmentioning

confidence: 99%

“…These latter models can be broadly placed into two classes based on the inputs required to estimate the gradient f of the gene regulatory dynamics, where f (x) = dx dt . The first class consists of methods like PRESCIENT [4] and RNAForecaster [6] that can learn f based only on time series gene expression input {x t0 ; x t1 ; . .…”

Section: Introductionmentioning

confidence: 99%

“…. ; x t T } without additional steps or consideration of other regulatory inputs [4,6,15]. In the process of learning transitions between consecutive time points, these "one-step" methods implicitly learn the local derivative (in the context of single cell sequencing often referred to as "RNA velocity" [16]) dx dt | x=xt m , as an intermediary to estimating f .…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Biologically informed NeuralODEs for genome-wide regulatory dynamics

Hossain

Fanfani

Quackenbush

et al. 2023

Preprint

View full text Add to dashboard Cite

Models that are formulated as ordinary differential equations (ODEs) can accurately explain temporal gene expression patterns and promise to yield new insights into important cellular processes, disease progression, and intervention design. Learning such ODEs is challenging, since we want to predict the evolution of gene expression in a way that accurately encodes the causal gene-regulatory network (GRN) governing the dynamics and the nonlinear functional relationships between genes. Most widely used ODE estimation methods either impose too many parametric restrictions or are not guided by meaningful biological insights, both of which impedes scalability and/or explainability. To overcome these limitations, we developed PHOENIX, a modeling framework based on neural ordinary differential equations (NeuralODEs) and Hill-Langmuir kinetics, that can flexibly incorporate prior domain knowledge and biological constraints to promote sparse, biologically interpretable representations of ODEs. We test accuracy of PHOENIX in a series of in silico experiments benchmarking it against several currently used tools for ODE estimation. We also demonstrate PHOENIX's flexibility by studying oscillating expression data from synchronized yeast cells and assess its scalability by modelling genome-scale breast cancer expression for samples ordered in pseudotime. Finally, we show how the combination of user-defined prior knowledge and functional forms from systems biology allows PHOENIX to encode key properties of the underlying GRN, and subsequently predict expression patterns in a biologically explainable way. PHOENIX will be available as open source at: https://github.com/QuackenbushLab/phoenix.

show abstract

Leveraging multi-omics data to empower quantitative systems pharmacology in immuno-oncology

Arulraj,

Wang,

Ippolito

et al. 2024

Briefings in Bioinformatics

View full text Add to dashboard Cite

Understanding the intricate interactions of cancer cells with the tumor microenvironment (TME) is a pre-requisite for the optimization of immunotherapy. Mechanistic models such as quantitative systems pharmacology (QSP) provide insights into the TME dynamics and predict the efficacy of immunotherapy in virtual patient populations/digital twins but require vast amounts of multimodal data for parameterization. Large-scale datasets characterizing the TME are available due to recent advances in bioinformatics for multi-omics data. Here, we discuss the perspectives of leveraging omics-derived bioinformatics estimates to inform QSP models and circumvent the challenges of model calibration and validation in immuno-oncology.

show abstract

Transcriptomic forecasting with neural ordinary differential equations

Cited by 6 publications

References 27 publications

Digitize your Biology! Modeling multicellular systems through interpretable cell behavior

Digitize your Biology! Modeling multicellular systems through interpretable cell behavior

Biologically informed NeuralODEs for genome-wide regulatory dynamics

Leveraging multi-omics data to empower quantitative systems pharmacology in immuno-oncology

Contact Info

Product

Resources

About