Identifying the Attractors of Gene Regulatory Networks from Expression Data under Uncertainty: An Interpretable Approach
This addresses the challenge of accurately determining cellular states in systems biology, with potential applications in understanding cell fates, but it is incremental as it builds on existing attractor analysis methods.
The paper tackles the problem of robustly identifying attractors in gene regulatory networks from temporal expression data under uncertainty, using a novel Zadeh Computing with Words approach, and demonstrates its efficiency and robustness in an in-silico study on a genetic toggle switch.
In systems biology, attractor landscape analysis of gene regulatory networks is recognized as a powerful computational tool for studying various cellular states from proliferation and differentiation to senescence and apoptosis. Therefore, accurate identification of attractors plays a critical role in determination of the cell fates. On the other hand, in a real biological circuit, genetic/epigenetic alterations as well as varying environmental factors drastically take effect on the location, characteristics, and even the number of attractors. The central question is: Given a temporal gene expression profile of a real gene regulatory network, how can the attractors be robustly identified in the presence of huge amount of uncertainty? This paper addresses this question using a novel approach based on Zadeh Computing with Words. The proposed scheme could effectively identify the attractors from temporal gene expression data in terms of both fuzzy logic-based and linguistic descriptions which are simply interpretable by human experts. Therefore, this method can be considered as an effective step towards interpretable artificial intelligence. Without loss of generality, genetic toggle switch is considered as the case study. The nonlinear dynamics of this benchmark gene regulatory network is computationally modeled by the notion of uncertain stochastic differential equations. The results of in-silico study demonstrate the efficiency and robustness of the proposed method.