Bayesian Data Integration Questions Classic Study on Protease Self-Digest Kinetics

Tötsch, Niklas; Hoffmann, Daniel

doi:10.1021/acsomega.0c01109

Cited by 2 publications

(10 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This result demonstrated how our approach can be used to generate mechanistic hypotheses explaining complex protein signaling networks. Similar studies concluded after proposing hypotheses (Hug et al, 2013; Tötsch and Hoffmann, 2020). In this work, we moved beyond hypothesis formation by testing the proposed SOCS dynamics.…”

Section: Resultssupporting

confidence: 83%

“…We introduced KL divergence-based clustering, an approach that facilitates hypothesis generation by prioritizing parameters for qualitative analysis. While other approaches for hypothesis formation using posteriors exist, they are largely qualitative and can be time consuming, as they typically require qualitative evaluation of every model parameter (Hug et al, 2013; Thijssen et al, 2018; Tötsch and Hoffmann, 2020). Our approach streamlines the qualitative evaluation of model parameters by using clustering to select only a subset of parameters for analysis.…”

Section: Resultsmentioning

confidence: 99%

“…Subsequently, rather than searching all model parameters at random, we use this ranking to prioritize searching the parameters that gained the most information from the data for evidence of alternative model hypotheses. Such evidence may manifest in the posterior, for example, as a bimodal posterior, or as a posterior that is inconsistent with different data subsets, as previously described (Hug et al, 2013;Tötsch and Hoffmann, 2020). Previous work has leveraged KL divergence as a metric of information gained from data to optimize experimental design (Chaloner and Verdinelli, 1995;Lomeli et al, 2021).…”

Section: Introductionmentioning

confidence: 93%

“…Such mathematical models must have accurate parameter values in order to generate reliable predictions. However, common methods of estimating model parameter values may not fully capture uncertainty, and thus overlook alternative hypotheses (Hug et al, 2013; Tötsch and Hoffmann, 2020). Bayesian statistics address this limitation by enabling the mathematical characterization of uncertainty in model parameters and subsequent predictions.…”

Section: Introductionmentioning

confidence: 99%

“…Indeed, previous studies have demonstrated the utility of posteriors for hypothesis formation based on mechanistic models of protein network dynamics. For example, Bayesian data integration has been used to reveal inconsistencies in a classic study on protease self-digest kinetics (Tötsch and Hoffmann, 2020). Bayesian data integration is the process of integrating subsets of data into the posterior inference and observing whether different data constrain the model in contradictory ways (Thijssen et al, 2018).…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Systematic Bayesian Posterior Analysis Guided by Kullback-Leibler Divergence Facilitates Hypothesis Formation

Huber

Georgia

Finley

2022

Preprint

View full text Add to dashboard Cite

Bayesian inference produces a posterior distribution for the parameters and predictions from a mathematical model that can be used to guide the formation of hypotheses, thereby increasing our understanding of a biological process. Previous approaches to such Bayesian hypothesis formation are largely qualitative and time consuming; further, demonstrations of these approaches typically stop at hypothesis formation, leaving the questions they raise unanswered. Here, we introduce a Kullback-Leibler (KL) divergence-based ranking to expedite Bayesian hypothesis formation and investigate the hypotheses it generates, ultimately generating novel, therapeutically-relevant biological insights. We test our approach on a computational model of prolactin-induced JAK2-STAT5 signaling, a pathway that mediates beta cell proliferation. Because diabetes is characterized by a decrease in functional beta cell mass, replenishing this mass by harnessing beta cell proliferation signaling pathways like this one is a viable potential therapeutic strategy. Our KL divergence-based approach selects only a subset of data-informed parameters to search for qualitative evidence of alternative hypotheses, thereby expediting hypothesis formation. Within this subset, we find two plausible ranges for the receptor degradation rate, each characterized by different dynamics of the SOCS protein, which mediates a negative feedback loop in the JAK2-STAT5 pathway. To investigate these dynamics, we refine the posterior of an existing model with additional data. This refined model out-performs the original model when predicting a test data set consisting of protein dynamics that are consistent across ligand and tissue types. Further, it predicts transient SOCS activation with high certainty. We demonstrate that our approach offers a novel quantitative framework for Bayesian hypothesis testing. Further, we establish an improved model of beta cell signaling and determine that the negative feedback controlling the prolactin-induced proliferative response in beta cells is likely transient. This understanding of SOCS dynamics is particularly relevant for harnessing the proliferation signaling response of beta cells.

show abstract

Section: Resultssupporting

confidence: 83%