Modeling semantics of inconsistent qualitative knowledge for quantitative Bayesian network inference

RPV Sealing Reliability Estimating Using a New Inconsistent Knowledge Fused Bayesian Network and Weighted Loss Function

The sealing system performance of a nuclear reactor pressure vessel (RPV) under different working circumstances is vital to the safe operation of the whole nuclear process; however, the mechanism, and especially the relationship among highly related variables for it, has remained unclear. Therefore, a new inconsistent knowledge fused Bayesian network and weighted loss function (iBWL) method was proposed to identify key variables and estimate the reliability of an RPV sealing system. In this method, a new inconsistent knowledge fusion method was proposed to make good use of available priceless field knowledge by considering its reliability and inconsistency seriously. The key variables identified by the Bayesian network structure were then used by a new weighted loss function to estimate the reliability of the RPV sealing system by comprehensively quantifying the deviations of them from their corresponding expected values. It is not only the quantified reliability of RPV sealing that can provide solid information for its operation status but also the weighted loss function can provide clues for how to tune the corresponding parameters to make sure RPV operating has an acceptable status. The application performed on the simulation samples based on the RPV of Liaoning Hongyanhe Nuclear Power Plant and another two RPV units in service strongly proved the outstanding performance of this advanced iBWL method.

Acquisition of causal models for local distributions in Bayesian networks

To specify a Bayesian network, a local distribution in the form of a conditional probability table, often of an effect conditioned on its n causes, needs to be acquired, one for each non-root node. Since the number of parameters to be assessed is generally exponential in n , improving the efficiency is an important concern in knowledge engineering. Non-impeding noisy-AND (NIN-AND) tree causal models reduce the number of parameters to being linear in n , while explicitly expressing both reinforcing and undermining interactions among causes. The key challenge in NIN-AND tree modeling is the acquisition of the NIN-AND tree structure. In this paper, we formulate a concise structure representation and an expressive causal interaction function of NIN-AND trees. Building on these representations, we propose two structural acquisition methods, which are applicable to both elicitation-based and machine learning-based acquisitions. Their accuracy is demonstrated through experimental evaluations.

Systematic search for recipes to generate induced pluripotent stem cells

Generation of induced pluripotent stem cells (iPSCs) opens a new avenue in regenerative medicine. One of the major hurdles for therapeutic applications is to improve the efficiency of generating iPSCs and also to avoid the tumorigenicity, which requires searching for new reprogramming recipes. We present a systems biology approach to efficiently evaluate a large number of possible recipes and find those that are most effective at generating iPSCs. We not only recovered several experimentally confirmed recipes but we also suggested new ones that may improve reprogramming efficiency and quality. In addition, our approach allows one to estimate the cell-state landscape, monitor the progress of reprogramming, identify important regulatory transition states, and ultimately understand the mechanisms of iPSC generation.

Converting somatic cells back to the stem cell state (called induced pluripotent stem cells or iPSCs) exemplifies the recent advancement of cellular reprogramming that holds great promise for developing regenerative medicine. Generation of iPSCs is often achieved by overexpressing three to four genes in somatic cells that are critical for regulating pluripotency. Developing optimal reprogramming recipe is a non-trivial task that requires significant effort. We present here a computational method that can facilitate discovery of effective recipes to generate iPSCs with high efficiency and better quality. In addition, our approach provides a new way to estimate the landscape in the cell-state space and monitor the trajectory of cellular reprogramming from a differentiated cell to an iPS cell. This work provides not only practical recipes for iPSC generation but also theoretical understanding of the reprogramming process.

A novel knowledge-driven systems biology approach for phenotype prediction upon genetic intervention

Deciphering the biological networks underlying complex phenotypic traits, e.g., human disease is undoubtedly crucial to understand the underlying molecular mechanisms and to develop effective therapeutics. Due to the network complexity and the relatively small number of available experiments, data-driven modeling is a great challenge for deducing the functions of genes/proteins in the network and in phenotype formation. We propose a novel knowledge-driven systems biology method that utilizes qualitative knowledge to construct a Dynamic Bayesian network (DBN) to represent the biological network underlying a specific phenotype. Edges in this network depict physical interactions between genes and/or proteins. A qualitative knowledge model first translates typical molecular interactions into constraints when resolving the DBN structure and parameters. Therefore, the uncertainty of the network is restricted to a subset of models which are consistent with the qualitative knowledge. All models satisfying the constraints are considered as candidates for the underlying network. These consistent models are used to perform quantitative inference. By in silico inference, we can predict phenotypic traits upon genetic interventions and perturbing in the network. We applied our method to analyze the puzzling mechanism of breast cancer cell proliferation network and we accurately predicted cancer cell growth rate upon manipulating (anti)cancerous marker genes/proteins.

Neural belief network

In this paper, a new neural belief network, which has considered backward inferences and the influence of the belief sources on belief propagations, is developed. In this new neural network, a link record set is built for every conclusion node for handling the multiple conditions of inference rules, and a route record set is built for every active node and every active link for handling the dependency of belief propagations on the belief sources. In addition, a temporary node is added for every evidence node. The assignment of the temporary nodes releases the evidence nodes from the role as belief sources and allows belief propagations in them. As a result, the new neural belief network can handle both definite evidences and indefinite evidences, and the evidences may come from observations or the prior knowledge of experts. The inference processes of the new neural belief network are based on available evidences and if...then rules. Therefore, it can solve the problems of Bayesian networks caused by the prior knowledge reliance and may be an alternative technique to the popular Bayesian networks.