Search

A state-of-the-art deep domain decomposition method (D3M) based on the variational principle is proposed for partial differential equations (PDEs). The solution of PDEs can be formulated as the solution of a constrained optimization problem, and we design a multi-fidelity neural network framework to solve this optimization problem. Our contribution is to develop a systematical computational procedure for the underlying problem in parallel with domain decomposition. Our analysis shows that the D3M approximation solution converges to the exact solution of underlying PDEs. Our proposed framework establishes a foundation to use variational deep learning in large-scale engineering problems and designs. We present a general mathematical framework of D3M, validate its accuracy and demonstrate its efficiency with numerical experiments.

Martynoside (MAR) is a bioactive glycoside of Rehmannia glutinosa, a traditional Chinese herb frequently prescribed for treating chemotherapy-induced pancytopenia. Despite its clinical usage in China for thousands of years, the mechanism of MAR's hematopoietic activity and its impact on chemotherapy-induced antitumor activity are still unclear. Here, we showed that MAR protected ex vivo bone marrow cells from 5-fluorouracil (5-FU)-induced cell death and inflammation response by down-regulating the TNF signaling pathway, in which II1b was the most regulatory gene. Besides, using mouse models with melanoma and colon cancer, we further demonstrated that MAR had protective effects against 5-FU-induced myelosuppression in mice without compromising its antitumor activity. Our results showed that MAR increased the number of bone marrow nucleated cells (BMNCs) and the percentage of leukocyte and granulocytic populations in 5-FU-induced myelosuppressive mice, accompanied by an increase in numbers of circulating white blood cells and platelets. The transcriptome profile of BMNCs further showed that the mode of action of MAR might be associated with the increased survival of BMNCs and the improvement of the bone marrow microenvironment. In summary, we revealed the potential molecular mechanism of MAR to counteract 5-FU-induced bone marrow cytotoxicity both ex vivo and in vivo, and highlighted its potential clinical usage in cancer patients experiencing chemotherapy-induced multi-lineage myelosuppression.

Failure probability evaluation for complex physical and engineering systems governed by partial differential equations (PDEs) are computationally intensive, especially when high-dimensional random parameters are involved. Since standard numerical schemes for solving these complex PDEs are expensive, traditional Monte Carlo methods which require repeatedly solving PDEs are infeasible. Alternative approaches which are typically the surrogate based methods suffer from the so-called “curse of dimensionality”, which limits their application to problems with high-dimensional parameters. For this purpose, we develop a novel hierarchical neural hybrid (HNH) method to efficiently compute failure probabilities of these challenging high-dimensional problems. Especially, multifidelity surrogates are constructed based on neural networks with different levels of layers, such that expensive highfidelity surrogates are adapted only when the parameters are in the suspicious domain. The efficiency of our new HNH method is theoretically analyzed and is demonstrated with numerical experiments. From numerical results, we show that to achieve an accuracy in estimating the rare failure probability (e.g., 10-5), the traditional Monte Carlo method needs to solve PDEs more than a million times, while our HNH only requires solving them a few thousand times.

In this paper we present an adaptive deep density approximation strategy based on KRnet (ADDA-KR) for solving the steady-state Fokker-Planck (F-P) equations. F-P equations are usually high-dimensional and defined on an unbounded domain, which limits the application of traditional grid based numerical methods. With the Knothe-Rosenblatt rearrangement, our newly proposed flow-based generative model, called KRnet, provides a family of probability density functions to serve as effective solution candidates for the Fokker-Planck equations, which has a weaker dependence on dimensionality than traditional computational approaches and can efficiently estimate general high-dimensional density functions. To obtain effective stochastic collocation points for the approximation of the F-P equation, we develop an adaptive sampling procedure, where samples are generated iteratively using the approximate density function at each iteration. We present a general framework of ADDA-KR, validate its accuracy and demonstrate its efficiency with numerical experiments.

In this work we propose a deep adaptive sampling (DAS) method for solving partial differential equations (PDEs), where deep neural networks are utilized to approximate the solutions of PDEs and deep generative models are employed to generate new collocation points that refine the training set. The overall procedure of DAS consists of two components: solving the PDEs by minimizing the residual loss on the collocation points in the training set and generating a new training set to further improve the accuracy of current approximate solution. In particular, we treat the residual as a probability density function and approximate it with a deep generative model, called KRnet. The new samples from KRnet are consistent with the distribution induced by the residual, i.e., more samples are located in the region of large residual and less samples are located in the region of small residual. Analogous to classical adaptive methods such as the adaptive finite element, KRnet acts as an error indicator that guides the refinement of the training set. Compared to the neural network approximation obtained with uniformly distributed collocation points, the developed algorithms can significantly improve the accuracy, especially for low regularity and high-dimensional problems. We demonstrate the effectiveness of the proposed DAS method with numerical experiments.

Backgroud: Controversy still remains in that whether sequential or concurrent regimen of anthracyclines and taxanes benefits more for breast cancer. Here we aimed to compare these two regimens in patients with operable breast cancer based on all published data of phase III randomized controlled trials in this topic.Methods: A literature search on PubMed, Web of science, Embase, ScienceDirect, Google scholar and clinicaltrials.gov databases was conducted up to May 2020. Meta-analysis was performed to evaluate the different efficacy on disease free survival (DFS) and overall survival (OS)for these two regimens. Subgroup analyses were further carried out in terms of node status and anthracycline selection.Results: Compared to concurrent regimen, sequential regimen did not improve the DFS or OS in the whole population included. Subgroup analysis showed that in node-positive patients, however, sequential regimen has better DFS, but not OS, than concurrent regimen.In sequential regimen, patients received doxorubicin and taxanes had improved DFS and OS than those were given epirubicin and taxanes. Futhermore, for patients receiving doxorubicin and taxanes,compared to sequential regimen,less cycles (4 cycles) of concurrent treatment showed worse DFS and OS, whereas more cycles(6 cycles) may rescue the loss.In addition,high grade toxicity response in both groups exhibited no difference except for neuropathy, which occured more frequently in the sequential regimen.Conclusions:Sequential regimen of anthracyclines and taxanes for patients with breast cancer did not promise a significant benefit in neither DFS nor OS over concurrent regimen. However, sequential regimen did provide a better DFS than that in concurrent regimen for node-positive patients. Interestingly, further subgroup analysis showed that for patients with node-positive and were given doxorubicin and taxanes, more cycles (6 cycles) of concurrent reigmen was not inferior to sequential regimen.

SUMMARY The type I interferon (IFN) pathway is a key component of innate immune response upon invasion of foreign pathogens. It is also under precise control to prevent excessive upregulation and undesired inflammation cascade. In the present study, we report that Riok3, an atypical kinase, negatively regulates retinoic acid-inducible gene-I (RIG-I)-like receptors (RLRs) sensing-induced type I IFN signaling. Riok3 deficiency selectively inhibits RNA viral replication in vitro, resulting from an upregulated type I IFN pathway. Mice with myeloid-specific Riok3 knockout also show a more robust induction of type I IFN upon RNA virus infection and are more resistant to RNA virus-induced pathogenesis. Mechanistically, Riok3 recruits and interacts with the E3 ubiquitin ligase TRIM40, leading to the degradation of RIG-I and melanoma differentiation-associated gene-5 (MDA5) via K48- and K27-linked ubiquitination. Collectively, our data reveal the mechanism that Riok3 employs to be a negative regulator of antiviral innate immunity., In brief Shen et al. show that Riok3 recruits TRIM40 to form a complex with both RIG-I and MDA5 during RNA virus infection, thus promoting TRIM40-mediated K27- and K48-linked ubiquitination and subsequent degradation of both RIG-I and MDA5 to inhibit the production of type I interferons., Graphical Abstract

Background Because of the increasing dysplasia rate in the lifelong course of inflammatory bowel disease (IBD) patients, it is imperative to characterize the crosstalk between IBD and colorectal cancer (CRC). However, there have been no reports revealing the occurrence of the ceRNA network in IBD-related CRC. Methods In this study, we conducted gene expression profile studies of databases and performed an integrated analysis to detect the potential of lncRNA-miRNA-mRNA ceRNA in regulating disease transformation. R packages were used to screen differentially expressed mRNA, lncRNA and miRNA among CRC, IBD and normal tissue. The lncRNA-miRNA-mRNA network was constructed based on predicted miRNA-targeted lncRNAs and miRNA-targeted mRNAs. Functional analyses were then conducted to identify genes involved in the ceRNA network, and key lncRNAs were evaluated based on several clinical outcomes. Results A total of three lncRNAs, 15 miRNAs, and 138 mRNAs were identified as potential mediators in the pathophysiological processes of IBD-related CRC. Gene Ontology annotation enrichment analysis confirmed that the dysplasia process was strongly associated with immune response, response to lipopolysaccharide, and inflammatory response. Survival analysis showed that LINC01106 (HR = 1.7; p

This work proposes a systematic model reduction approach based on rank adaptive tensor recovery for partial differential equation (PDE) models with high-dimensional random parameters. Since the standard outputs of interest of these models are discrete solutions on given physical grids which are high-dimensional, we use kernel principal component analysis to construct stochastic collocation approximations in reduced dimensional spaces of the outputs. To address the issue of high-dimensional random inputs, we develop a new efficient rank adaptive tensor recovery approach to compute the collocation coefficients. Novel efficient initialization strategies for non-convex optimization problems involved in tensor recovery are also developed in this work. We present a general mathematical framework of our overall model reduction approach, analyze its stability, and demonstrate its efficiency with numerical experiments.

Mutations in the human PAX6 gene produce various phenotypes, including aniridia, Peters' anomaly, autosomal dominant keratitis and familial foveal dysplasia. The various phenotypes may arise from different mutations in the same gene. To test this theory, we performed a functional analysis of two missense mutations in the paired domain: the R26G mutation, previously reported in a case of Peters' anomaly, and an unreported I87R mutation, which we identified in a patient with aniridia. While both the R26 and the I87 positions are conserved in the paired boxes of all known PAX genes, X-ray crystallography has shown that only R26 makes contact with DNA. We showed that the R26G mutant failed to bind a subset of paired domain binding sites but, surprisingly, bound other sites and successfully transactivated promoters containing those sites. In contrast, the I87R mutant had lost the ability to bind DNA at all tested sites and failed to transactivate promoters. Our data support the haploid-insufficiency hypothesis of aniridia, and the hypothesis that R26G is a hypomorphic allele.

PAX6 is required for proper development of the eye, central nervous system, and nose. PAX6 has two DNA binding domains, a glycine-rich region that links the two DNA binding domains, and a transactivation domain. There is evidence that the different DNA binding domains of PAX6 have different target genes. However, it is not clear if the two DNA binding domains function independently. We have studied the effect of structural changes in the paired domain on the function of PAX6 mediated through its homeodomain. The R26G and I87R mutations have been reported in different human patients with clinically different phenotypes and are in the N- and the C-terminal halves of the paired domain, respectively. Surprisingly, we found that the I87R mutant protein not only lost the transactivation function but also failed to bind DNA by either of its DNA binding domains. In contrast, the R26G mutant protein lost DNA binding through its paired domain but had greater DNA binding and transactivation than wild-type PAX6 on homeodomain binding sites. Like R26G, the 5a isoform showed higher DNA binding than wild-type PAX6. This study demonstrates that the two subdomains of the paired domain influence the function of the homeodomain differentially and also provides an explanation for the difference in phenotypes associated with these mutations.

PAX6 is a transcription factor with two DNA-binding domains (paired box and homeobox) and a proline-serine-threonine (PST)-rich transactivation domain. PAX6 regulates eye development in animals ranging from jellyfish to Drosophila to humans. Heterozygous mutations in the human PAX6 gene result in various phenotypes, including aniridia, Peter's anomaly, autosomal dominant keratitis, and familial foveal dysplasia. It is believed that the mutated allele of PAX6 produces an inactive protein and aniridia is caused due to genetic haploinsufficiency. However, several truncation mutations have been found to occur in the C-terminal half of PAX6 in patients with Aniridia resulting in mutant proteins that retain the DNA-binding domains but have lost most of the transactivation domain. It is not clear whether such mutants really behave as loss-of-function mutants as predicted by haploinsufficiency. Contrary to this theory, our data showed that these mutants are dominant-negative in transient transfection assays when they are coexpressed with wild-type PAX6. We found that the dominant-negative effects result from the enhanced DNA binding ability of these mutants. Kinetic studies of binding and dissociation revealed that various truncation mutants have 3-5-fold higher affinity to various DNA-binding sites when compared with the wild-type PAX6. These results provide a new insight into the role of mutant PAX6 in causing aniridia.