A Scalable, Open-Source Implementation of a Large-Scale Mechanistic Model for Single Cell Proliferation and Death Signaling

Abstract

The current era of big biomedical data accumulation and availability brings data integration opportunities for leveraging its totality to make new discoveries and/or clinically predictive models. Black-box statistical and machine learning methods are powerful for such integration, but often cannot provide mechanistic reasoning, particularly on the single-cell level. While single-cell mechanistic models clearly enable such reasoning, they are predominantly “small-scale”, and struggle with the scalability and reusability required for meaningful data integration. Here, we present an open-source pipeline for scalable, single-cell mechanistic modeling from simple, annotated input files that can serve as a foundation for mechanistic data integration. As a test case, we convert one of the largest existing single-cell mechanistic models to this format, demonstrating robustness and reproducibility of the approach. We show that the model cell line context can be changed with simple replacement of input file parameter values. We next use this new model to test alternative mechanistic hypotheses for the experimental observations that interferon-gamma (IFNG) inhibits epidermal growth factor (EGF)-induced cell proliferation. Model- based analysis suggested, and experiments support that these observations are better explained by IFNG-induced SOCS1 expression sequestering activated EGF receptors, thereby downregulating AKT activity, as opposed to direct IFNG-induced upregulation of p21 expression. Overall, this new pipeline enables large-scale, single-cell, and mechanistically-transparent modeling as a data integration modality complementary to machine learning.

Predicting Anti-Cancer Drug Combination Responses with a Temporal Cell State Network Model

“…In other cases, such mapping may not be known a priori and/or more complex, i.e., a single drug may influence multiple transition rates. Biochemical network models that capture such complexities or mapping may prove useful in such situations [26][27][28][29][30]45,78 . Assumptions regarding the additivity (or not) of multi-drug action on transition rates would have to be asserted.…”

“…In principle, more comprehensive exploration of drug combination space could be achieved in silico. Various computational methods including mechanistic models and machine learning approaches have shown promise in predicting drug combination responses, especially taking into consideration context specific pathology and omics data as well as identifying specific biomarkers and drugtargets [44][45][46][47][48][49] . Regardless of the modeling methods being used, there is a widespread focus on using information about biochemical networks to facilitate drug combination response prediction 31,[50][51][52] .…”

Multi-Omics Binary Integration via Lasso Ensembles (MOBILE) for identification of context-specific networks and new regulatory mechanisms

“…The LINCS datasets analyzed during the current study are available in the Synapse repository, synapse.org/LINCS_MCF10A (67,128). Source Data are provided with this paper at doi:10.6084/m9.figshare.20294229.…”

Computational Speed-Up of Large-Scale, Single-Cell Model Simulations Via a Fully-Integrated SBML-Based Format

“…Algorithmic developments included rule-based modeling to specify reactions more compactly (Faeder et al, 2009), and model composition tools, (Lopez et al, 2013;Gyori et al, 2017;Hoops et al, 2006;Somogyi et al, 2015) but large-scale models often still presented challenges. More recent work has provided such tools like AMICI that enables SBML-specified models to be simulated quickly, PEtab (Stapor et al, 2018;Schmiester et al, 2021) and Datanator (Roth et al, 2021) that specifies data formats for parameter estimation, formalisms that can help with unambiguous species naming, (Lang et al, 2020) and composition approaches such as ours that simplify model aggregation and expansion in ways that are compatible with efficient largescale simulation algorithms and easy to reuse (Erdem et al, 2022). Not unexpectedly, however, there remains much work to be done to even technically enable large-scale and whole-cell modeling.…”