C&SP34: Development of a high-level, rule-based whole-cell modeling language

A. Collaborating Investigators: Jonathan Karr,¹ James R. Faeder²

B. Institutions: ¹Mt. Sinai and ²University of Pittsburgh

C. Funding Status of Project: NSF grant 1548123 "ERASynBio: MiniCell - A Model-driven Approach to Minimal Cell Engineering" (Karr) 12/17/2015–7/31/2018

D. Biomedical Research Problem: Despite decades of research and the growing wealth of data, we do not have a unified understanding of cell biology. Whole-cell (WC) models that represent every gene function and that predict the cell cycle dynamics of every molecular species have the potential to unify our understanding of cell biology.^111-113 Recently, Karr and others developed the first WC model.¹¹⁴ This was achieved by combining multiple mathematically distinct submodels of individual cellular pathways, and by developing custom software to simulate the model. However, this ad hoc approach is not easily reusable, expandable, or reproducible.¹¹⁵ New WC modeling tools are needed to build better models that can guide bioengineering and medicine.

Fig VIII.9 Transcription factor binding in WC-Lang. A. Rules define specific sequences of DNA and protein-DNA interactions in terms of DNA-sequence motifs and specific proteins. Evaluation of WC-Lang rules generates binding sites on specific sequences. B. Sequences and proteins are instantiated as BioNetGen species and rules. C. This system can be efficiently simulated using NFsim.

E. Methods and Procedures: This project will develop a high-level rule-based WC modeling language and a rule-based multi-algorithm simulator. The language will enable researchers to compactly describe biological processes, such as the translation of each amino acid, using high-level rule patterns and sequence, genomic, and biochemical data. WC-Lang will provide high-level data-based rules to succinctly describe large reactions networks in terms of reaction patterns and sequence, genomic, and other data, that will be encoded in BioNetGen.¹¹⁶ This high-level language will make it easier for researchers to develop and communicate WC models, and the BioNetGen implementation will make it possible to rapidly develop a reusable and efficient WC model simulator that will take advantage of BioNetGen’s network generation as well as the network-free simulation capabilities of NFsim.¹¹⁷ WC-Lang will be implemented as a Python library which encodes the high-level rules into lower-level BioNetGen rules. Use of rules to encode molecular interactions will open many possibilities for the visualization of WC models using contact maps and regulatory graphs (see Aim 2 of TR&D3). In the first phase of this project we will focus on implementing modules from a reference WC model the Karr lab is building that will be described completely independently from any simulation code. This is a major first step in the development of WC-lang. We will develop code to translate reference model modules into BioNetGen language code (Fig. VIII.9) and test simulation results against those obtained with reference simulators. In the second phase, we will expand BioNetGen and NFSim to support multi-algorithm modeling, which will be enabled by development of the libasal bodyNG API in Aim 3.1 of TR&D3. Submodels will be implemented as distinct processes, which is already possible for existing interfaces to BioNetGen and NFsim, and these will in turn be controlled by the multi-algorithm simulator that the Karr lab is developing in the above-referenced NSF-funded project. Together, the new language and expanded simulator will enable researchers to build much larger and more accurate WC models that can guide bioengineering and precision medicine.