| Title | : | From multi-modal data to systems-level modeling of cellular processes |
| Speaker | : | Nilesh Anantha Subramanian (IITM) |
| Details | : | Fri, 26 Apr, 2024 4:30 PM @ SSB-233 |
| Abstract: | : | An overarching goal of computational systems biology is to model interactions among bioprocesses like immune response, gene regulation (GRN), and metabolism (MM) in an organism. While established models of such bioprocesses under healthy and perturbed (disease) states exist, critical gaps around double-disease states and MMà GRN interactions remain. Towards this end, we propose computational frameworks based on statistical models and constraint-based (linear optimization) techniques in the context of these two studies: I) A systematic methodology using linear models to quantify disease-disease interaction (DDI) effects in immune response: Studies have shown that immune responses in a person suffering from two diseases (comorbidity) could be more complex than the union of responses to each disease occurring separately. A data-driven quantification of this complexity is lacking though. We propose a three-step computational framework based on linear models with interaction terms to analyze data from multiple groups of individuals (healthy group, single-disease patients, and double-disease group), to identify, quantify, and interpret the double-disease (DDI) markers. Application of this framework to helminth infection and diabetes revealed DDI markers like TNF-α, IFN-γ, and IL-2, etc., and IL-4/13 and IL-10 immune signaling pathways to be enriched in the target genes of these DDI markers. Our approach is generic and can be applied to dissect DDI mechanisms behind other comorbid conditions as well. II) Iterative surgery on a probabilistic causal network to model MMà GRN feedback: Integrated MM and GRN models can help predict the behavior of a complex biological system. Several works capture the information flow from GRN to MM via Boolean rules, however none model the feedback from MM to GRN. I will present an iterative algorithm that models this feedback via causal surgeries on a probabilistic GRN network and combines inference on this network with a known linear program representation of MM. I will also present preliminary results from the application of this integrated GRNßà MM model on simulated and real-world datasets. |
