1,721,237 research outputs found
Impacts of glycans attached to therapeutic glycoproteins
No full textHigh value-added therapeutic proteins have been
leading the biologics industry and occupied major portion of
the market. More than 60% of the currently available protein
therapeutics are glycoproteins attached with glycans which
play crucial roles for the protein folding, therapeutic efficacy,
in vivo half-life and immunogenecity. This review
introduces the process of glycosylation and the impacts of
glycans in the aspects of therapeutics. The important glycan
structures in therapeutic performances were also summarized
focusing on three representative categories of glycoproteins,
cytokines, therapeutic antibody and enzyme. Currently,
mammalian expression systems such as Chinese hamster
ovary cells are preferred for the production of therapeutic
glycoproteins due to their ability to synthesize glycans
having similar structures with human type glycans. However,
recent advances of plant glycoengineering to overcome
the limitation originating from different glycan structures
will soon allow to develop more efficient and economic
plant-based production systems for therapeutic glycoproteins.open
sj-docx-1-tag-10.1177_17562848241239551 – Supplemental material for Comparison of long-term outcomes of endoscopic ultrasound-guided hepaticogastrostomy and choledochoduodenostomy for distal malignant biliary obstruction: a multicenter retrospective study
No full textSupplemental material, sj-docx-1-tag-10.1177_17562848241239551 for Comparison of long-term outcomes of endoscopic ultrasound-guided hepaticogastrostomy and choledochoduodenostomy for distal malignant biliary obstruction: a multicenter retrospective study by Dongwook Oh, Sung Yong Han, Sang Hyub Lee, Seong-Hun Kim, Woo Hyun Paik, Hyung-Ku Chon, Tae Jun Song, Se Woo Park and Jae Hee Cho in Therapeutic Advances in Gastroenterology</p
Identification and functional characterization of the NanH extracellular sialidase from Corynebacterium diphtheriae
No full textCorynebacterium diphtheriae, a pathogenic Gram-positive bacterium, contains sialic acids on its cell surface, but no genes related to sialic acid decoration or metabolism have been reported in C. diphtheriae. In the present study, we have identified a putative sialidase gene, nanH, from C. diphtheriae KCTC3075 and characterized its product for enzyme activity. Interestingly, the recombinant NanH protein was secreted as a catalytically active sialidase into the periplasmic space in Escherichia coli, while the short region at its C-terminus was truncated by proteolysis. We reconstructed a truncated NanH protein His6-NanHΔN devoid of its signal sequence as a mature enzyme fused with the 6xHis tag at the N-terminal region. The purified His6-NanHΔN can cleave α-2,3- and α-2,6-linked sialic acid from sialic acid-containing substrates. In addition, even though the efficiency was low, the recombinant His6-NanHΔN was able to catalyse the transfer of sialic acid using several sialoconjugates as donor, suggesting that the reversible nature of C. diphtheriae NanH can be used for the synthesis of sialyl oligosaccharides via transglycosylation reaction.open
Construction of an in vitro trans-sialylation system: surface display of Corynebacterium diphtheriae sialidase on Saccharomyces cerevisiae
No full textSialidases can be used to transfer sialic acids from sialoglycans to asialoglycoconjugates via the trans-glycosylation reaction mechanism. Some pathogenic bacteria decorate their surfaces with sialic acids which were often scavenged from host sialoglycoconjugates using their surface-localized enzymes. In this study, we constructed an in vitro trans-sialylation system by reconstructing the exogenous sialoglycoconjugate synthesis system of pathogens on the surfaces of yeast cells. The nanH gene encoding an extracellular sialidase of Corynebacterium diphtheriae was cloned into the yeast surface display vector pYD1 based on the Aga1p-Aga2p platform to immobilize the enzyme on the surface of the yeast Saccharomyces cerevisiae. The surface-displayed recombinant NanH protein was expressed as a fully active sialidase and also transferred sialic acids from pNP-α-sialoside, a sialic acid donor substrate, to human-type asialo-N-glycans. Moreover, this system was capable of attaching sialic acids to the glycans of asialofetuin via α(2,3)- or α(2,6)-linkage. The cell surface-expressed C. diphtheriae sialidase showed its potential as a useful whole cell biocatalyst for the transfer of sialic acid as well as the hydrolysis of N-glycans containing α(2,3)- and α(2,6)-linked sialic acids for glycoprotein remodeling.open
Iron chelator-inducible expression system for Escherichia coli
No full textThe PentC promoter of the entCEBA operon encoding enzymes for enterobactin biosynthesis in Escherichia coli is tightly regulated by the availability of iron in the culture medium. In iron-rich conditions, the PentC promoter activity is strongly repressed by the global transcription regulator Fur (ferric uptake regulator), which complexes with ferrous ions and binds to the Fur box 19-bp inverted repeat. In this study, we have constructed the expression vector pOS2 containing the PentC promoter and characterized its repression, induction, and modulation by quantifying the expression of the lacZ reporter gene encoding β-galactosidase. β-Galactosidase activities of E. coli transformants harboring pOS2-lacZ were highly induced in the presence of divalent metal ion chelators such as 2,2′-dipyridyl and EDTA, and were strongly repressed in the presence of excess iron. It was also shown that the basal level β-galactosidase expression by the PentC promoter was drastically decreased by incorporating the fur gene into the expression vector. Since the newly developed iron chelator-inducible expression system is efficient and cost-effective, it has wide applications in recombinant protein production.open
Features and applications of bacterial sialidases
No full textSialidases, or neuraminidases (EC 3.2.1.18), belong to a class of glycosyl hydrolases that release terminal N-acylneuraminate residues from the glycans of glycoproteins, glycolipids, and polysaccharides. In bacteria, sialidases can be used to scavenge sialic acids as a nutrient from various sialylated substrates or to recognize sialic acids exposed on the surface of the host cell. Despite the fact that bacterial sialidases share many structural features, their biochemical properties, especially their linkage and substrate specificities, vary widely. Bacterial sialidases can catalyze the hydrolysis of terminal sialic acids linked by the α(2,3)-, α(2,6)-, or α(2,8)-linkage to a diverse range of substrates. In addition, some of these enzymes can catalyze the transfer of sialic acids from sialoglycans to asialoglycoconjugates via a transglycosylation reaction mechanism. Thus, some bacterial sialidases have been applied to synthesize complex sialyloligosaccharides through chemoenzymatic approaches and to analyze the glycan structure. In this review article, the biochemical features of bacterial sialidases and their potential applications in regioselective hydrolysis reactions as well as sialylation by transglycosylation for the synthesis of sialylated complex glycans are discussed.open
Biochemical charactrization of a glycosyltransferase homolog from an oral pathogen Fusobacterium nucleatum as a human glycan-modifying enzyme
No full textBacterial glycosyltransferases have drawn growing attention as economical enzymes for oligosaccharide synthesis, with their easy expression and relatively broad substrate specificity. Here, we characterized a glycosyltransferase homolog (Fnu_GT) from a human oral pathogen, Fusobacterium nucleatum. Bioinformatic analysis showed that Fnu_GT belongs to the glycosyltransferases family II. The recombinant Fnu_GT (rFnu_GT) expressed in Escherichia coli displayed the highest glycosylation activity when UDP-galactose (Gal) was used as a donor nucleotide-sugar with heptose or Nacetylglucosamine (GlcNAc) as an acceptor sugar. Interestingly, rFnu_GT transferred the galactose moiety of UDP-Gal to a nonreducing terminal GlcNAc attached to the trimannosyl core glycan, indicating its potential as an enzyme for humantype N-glycan synthesisopen
Going Beyond Counting First Authors in Author Co-citation Analysis
Full text linkThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
임의의 초기 정책에 대한 선형 시스템의 데이터 기반 최적 제어 및 쿠프만 연산자를 활용한 비선형 시스템에 대한 응용
Full text link학위논문(박사) -- 서울대학교대학원 : 공과대학 기계항공공학부, 2023. 8. 김유단.A model-free off-policy reinforcement learning algorithm is proposed for solving optimal control problems for dynamic systems. The algorithm is designed to converge to not only the optimal but also stabilizing policy, which is one of the most critical concerns in designing the controller for safety-critical systems such as unmanned aerial vehicles. Unlike typical approximate dynamic programming methods, an initial stabilizing policy is not required by the proposed algorithm, which is a key advantage.
In the first part of the dissertation, a data-driven surrogate Q-leaning algorithm is proposed for linear systems based on the extended Kleinman iteration that solves algebraic Riccati equation. To allow an initial unstable policy, the value function is redefined implicitly to evaluate the performance index of the unstable policy. Based on this implicit value function, an action-value function called the surrogate Q-function is proposed by augmenting virtual control dynamics in the state space to properly define values of state and control input pairs. An off-policy data-driven method called the surrogate Q-learning is then provided based on the surrogate Q-function, which enables the reuse of data obtained from an arbitrary control sources, e.g., trained human experts or fine-tuned PID controllers. The convergence of the extended Kleinman iteration to the unique positive definite solution, which yields the optimal stabilizing policy, is proven based on matrix inertia theory since the surrogate Q-learning is equivalent to the extended Kleinman iteration.
The second part of the dissertation is devoted to an application of the proposed reinforcement learning algorithm to nonlinear systems. The Koopman operator theory is employed to linearize nonlinear systems in an infinite-dimensional space, called the Koopman lifting linearization. The controllability and observability of linearized systems are investigated by assuming that there exists a finite-dimensional invariant subspace of the Koopman operator spanned by a mapping called the lifting. The equivalence between two optimal control problems for the original nonlinear system and the linearized system is shown under several conditions on the lifting. To find the lifting satisfying all of the conditions, a diffeomorphic lifting approximation by coupling flow-based invertible deep neural network is employed. A meta-learning framework is proposed to train the network to approximate a common lifting for a group of systems, and therefore the mode-free surrogate Q-learning can be applied to uncertain systems.
Numerical simulations using illustrative nonlinear systems with known optimal controllers are used to demonstrate the feasibility of the proposed framework, along with practical considerations and implementation details.본 논문에서는 최적제어 문제를 해결하기 위해 비모델(model-free) 강화학습 알고리듬을 제안하였다. 제어 시스템의 안정성은 제어기 설계 시 필수적으로 고려되어야 할 사항으로 본 논문에서 제안한 알고리듬은 학습되는 제어기가 최적일 뿐만 아니라 안정한 제어기로 수렴하도록 설계되었다. 기존의 근사 동적 프로그래밍 기법들과는 달리, 제안한 알고리듬은 안정한 초기 제어기를 필요로 하지 않는데, 이는 불안정한 평형점을 가지는 시스템의 비모델 학습 관점에서 주요한 장점이다.
논문의 전반부에서는 데이터만을 이용해 선형 시스템의 안정한 최적제어기를 학습할 수 있는 새로운 형태의 Q-학습 알고리듬을 제안한다. 초기 불안정한 제어 입력을 허용하기 위해 성능지수를 평가하기 위한 가치함수를 음함수 형태 재정의하고, 선형 시스템에 대해 존재성과 유일성을 보였다. 가상의 제어 동역학을 상태변수에 추가한 확장된 상태공간에서의 가치 음함수로 Q-함수를 정의하고, 이를 기반으로 하는 정책 반복법 기반의 Q-학습 알고리듬을 제안하였다. 이 알고리듬은 학습 중인 제어기로부터 데이터를 얻을 필요가 없는 off-policy 기법으로, 시스템의 숙련된 운영자나 실험적으로 설계된 PID 제어기를 통해 얻은 데이터를 사용할 수 있다는 장점이 있다. 제안한 Q-러닝 알고리듬을 이용하면 학습되는 제어기가 유한 단계 이내에 안정화 되며, 최종적으로 대수적 리카티(Riccati) 방정식의 안정한 선형 최적해로 수렴함을 행렬 관성 이론을 기반으로 증명하였다.
논문의 후반부는 제안된 강화학습 알고리즘을 비선형 시스템에 적용하는 문제를 다룬다. 이를 위해 비선형 시스템을 무한 차원 공간에서 선형화하는 쿠프만(Koopman) 연산자 이론을 활용한다. 리프팅(lifting)이라 불리는 매핑에 의해 생성되는 쿠프만 연산자의 유한 차원의 불변 부분공간이 존재한다고 가정할 때, 선형화된 시스템의 최적제어를 위해 가제어성과 가관측성을 가지기 위한 조건을 정립한다. 리프팅에 대한 여러 조건을 바탕으로 기존 비선형 시스템 최적제어 문제와 선형화된 시스템의 최적제어 문제 간의 동치성을 증명하고, 앞서 제안한 강화학습 알고리즘을 사용할 수 있는 이론적 근거를 마련한다. 모든 조건을 만족하는 리프팅을 찾기 위해 가역 심층신경망을 활용한 미분동형(diffeomorphic) 리프팅 근사법을 제안한다. 특정 시스템 그룹에 대해 공통된 리프팅이 존재한다면 그룹 내의 불확실한 시스템에 대해 제안한 비모델 강화학습을 활용할 수 있다는 점에 착안하여, 공통 리프팅을 학습하는 메타 러닝(meta learning) 프레임워크를 개발하였다.
마지막으로 이미 알려진 최적 제어기와 비선형 동역학을 갖는 비선형 시스템을 사용하여 수치 시뮬레이션을 수행하고, 제안된 프레임워크의 타당성과 구현 세부사항을 살펴보았다.1 Introduction 1
1.1 Problem Statement 1
1.2 Background, Motivation, and Necessities 3
1.3 Literature Review 6
1.3.1 Iterative Methods for Solving AREs 6
1.3.2 Model-Free Policy Iteration Methods 7
1.3.3 ADP Methods Without Initial Admissible Policies 8
1.3.4 The Koopman Operator for Control 8
1.3.5 Learning-Based Koopman Operator Applications 10
1.4 Objectives and Contributions 11
1.4.1 Objectives 11
1.4.2 Contributions 11
1.5 Dissertation Outline 15
2 Theoretical Backgrounds 17
2.1 Notation 17
2.2 Mathematical Preliminaries 18
2.2.1 The Matrix Inertia Theorem 18
2.2.2 Fréchet Derivatives 18
2.2.3 The Koopman Operator 19
2.3 Linear System Theory 22
2.3.1 Controllability and Observability 22
2.3.2 Algebraic Riccati Equations 23
2.3.3 Lyapunov Equations 25
2.4 The Kleinman Iteration 27
2.5 Meta-Learning 30
2.5.1 Optimization Problem Formulations 30
2.5.2 Closed-Form Base Learners 31
3 Data-Driven Optimal Control for Unknown Linear Systems 33
3.1 Implicit Value Functions 33
3.2 The Surrogate Q-Learning 38
3.2.1 Surrogate Q-Functions for Continuous-Time Systems 38
3.2.2 The Surrogate Q-Learning Algorithm 42
3.2.3 The Data-Driven Surrogate Q-Learning 46
3.3 The Extended Kleinman Iteration 49
3.3.1 Existence of Solutions 50
3.3.2 Selection of Design Parameters 52
3.4 Convergence Analysis 54
3.4.1 Monotonic Stabilization 54
3.4.2 Local Convergence 56
3.4.3 Global Convergence 59
3.5 Illustrative Numerical Examples 65
3.5.1 Validation of the Extended Kleinman Iteration 65
3.5.2 Validation of the Data-Driven Surrogate Q-Learning 66
4 Application to Nonlinear Optimal Control Problems 73
4.1 NonlinearOptimalControlProblems 74
4.2 Koopman Operators for Optimal Control Problems 76
4.2.1 Koopman Lifting Linearization 76
4.2.2 Equilibrium Points 78
4.2.3 Lifted Optimal Control Problems 79
4.3 The Meta-Learning Framework 85
4.3.1 Koopman Groups and Common Liftings 85
4.3.2 Diffeomorphic Lifting Approximation 86
4.3.3 Base Learner Formulation 89
4.3.4 Meta-Learner Formulation 91
4.3.5 Offline and Online Learning Synthesis 93
5 Numerical Simulation 95
5.1 Koopman Group of Nonlinear Systems 95
5.2 The Meta-Learning Stage 98
5.2.1 Meta-Learning Setups 98
5.2.2 Meta-Learning Results 99
5.3 The Surrogate Q-Learning Stage 105
5.3.1 Surrogate Q-Learning Setups 105
5.3.2 Surrogate Q-Learning Results 106
6 Conclusion 113
6.1 Concluding Remarks 113
6.2 Direction for Further Research 115
Bibliography 117
Appendix A The Glow Implementation 131
A.1 Flows 131
A.1.1 ActivationLayers 131
A.1.2 1×1 Convolution Layers 133
A.1.3 Affine Coupling Layers 134
국문초록 135박
- …
