Logical versus kinetic modeling of biological networks: applications in cancer research

Abstract

Mathematical modeling of biological networks is a promising approach to understand the complexity of cancer progression, which can be understood as accumulated abnormalities in the kinetics of cellular biochemistry. Two major modeling formalisms (languages) have been used for this purpose in the last couple of decades: one is based on the application of classical chemical kinetics of reaction networks and the other one is based on discrete kinetics representation (called logical formalism for simplicity here), governed by logical state update rules. In this short review, we remind the reader how these two methodologies complement each other but also present the fast and recent development of semi-quantitative approaches for modeling large biological networks, with a spectrum of complementary ideas each inheriting and combining features of both modeling formalisms. We also notice an increasing influence of the recent success of machine learning and artificial intelligence onto the methodology of mathematical network modeling in cancer research, leading to appearance of a number of pragmatic hybrid approaches. To illustrate the two approaches, logical versus kinetic modeling, we provide an example describing the same biological process with different description granularity in both discrete and continuous formalisms. The model focuses on a central question in cancer biology: understanding the mechanisms of metastasis appearance. We conclude that despite significant progress in development of modeling ideas, predicting response of large biological networks involved in cancer to various perturbations remains a major unsolved challenge in cancer systems biology.

Unsupervised logic-based mechanism inference for network-driven biological processes

“…Summary of the asynchronous updating results after 90 asynchronous updating simulations for each of the 256 possible initial states. The numbers on the right hand side are frequencies that the simulation reaches the corresponding steady state for the EMT model from [44]. This is an extension to Table 6 with the inclusion of the three selected models from the genetic optimizer EMT-A, EMT-B, and EMT-C, respectively.…”

Towards control of cellular decision-making networks in the epithelial-to-mesenchymal transition

“…As true in general, the continuous and discrete models of EMT are complementary of each other (Calzone, Barillot, and Zinovyev 2018). The continuous models usually describe each node at a more detailed resolution, for example they describe how a microRNA (e.g.…”

“…It is likely to be generally true that discrete models correspond to coarse-grained versions of continuous models that are mechanistic at the elementary reaction level (Calzone, Barillot, and Zinovyev 2018). The nodes of the discrete model are coarse-grained representations of modules or motifs of the continuous system, and the discrete states represent the attractors of these modules/motifs.…”

Computational Approaches in Theranostics: Mining and Predicting Cancer Data

“…These have been predicted using different approaches including mathematical modeling, stochastic search, global gene expression, and targeted phosphoproteomics profiling [108]. The mechanisms of action of drug combinations and synergies have been identified resorting to network-based models [15,16,17,108].…”

“…However, literature regarding cancer theranostics is lacking in comprehensive and systematic approaches to: (1) fully inspect the relevant interaction patterns and synergistic effects, (2) evaluate tumor heterogeneity and data-intensive theranostics technologies, (3) confirm the effectiveness of therapeutics, and (4) compare and validate specific mechanistic models. Fundamental aspects on the cellular and molecular basis of cancer have also been explored through the establishment of relevant biological networks [9,10,11,12,13,14,15,16,17]. This has been facilitated by combining information from cancer genomic, transcriptomic, proteomic, and metabolomic data and computational techniques, aiming at developing non-invasive methods for diagnostic purposes [9].…”