共查询到20条相似文献,搜索用时 15 毫秒
1.
The conditions for diffusion-driven (Turing) instabilities in systems with two reactive species are well known. General methods for detecting potential Turing bifurcations in larger reaction schemes are, on the other hand, not well developed. We prove a theorem for a graph-theoretic condition originally given by Volpert and Ivanova [Mathematical Modeling (Nauka, Moscow, 1987) (in Russian), p. 57] for Turing instabilities in a mass-action reaction-diffusion system involving n substances. The method is based on the representation of a reaction mechanism as a bipartite graph with two types of nodes representing chemical species and reactions, respectively. The condition for diffusion-driven instability is related to the existence of a structure in the graph known as a critical fragment. The technique is illustrated using a substrate-inhibited bifunctional enzyme mechanism which involves seven chemical species. 相似文献
2.
3.
4.
5.
6.
Mingjian Wen Samuel M. Blau Xiaowei Xie Shyam Dwaraknath Kristin A. Persson 《Chemical science》2022,13(5):1446
Machine learning (ML) methods have great potential to transform chemical discovery by accelerating the exploration of chemical space and drawing scientific insights from data. However, modern chemical reaction ML models, such as those based on graph neural networks (GNNs), must be trained on a large amount of labelled data in order to avoid overfitting the data and thus possessing low accuracy and transferability. In this work, we propose a strategy to leverage unlabelled data to learn accurate ML models for small labelled chemical reaction data. We focus on an old and prominent problem—classifying reactions into distinct families—and build a GNN model for this task. We first pretrain the model on unlabelled reaction data using unsupervised contrastive learning and then fine-tune it on a small number of labelled reactions. The contrastive pretraining learns by making the representations of two augmented versions of a reaction similar to each other but distinct from other reactions. We propose chemically consistent reaction augmentation methods that protect the reaction center and find they are the key for the model to extract relevant information from unlabelled data to aid the reaction classification task. The transfer learned model outperforms a supervised model trained from scratch by a large margin. Further, it consistently performs better than models based on traditional rule-driven reaction fingerprints, which have long been the default choice for small datasets, as well as those based on reaction fingerprints derived from masked language modelling. In addition to reaction classification, the effectiveness of the strategy is tested on regression datasets; the learned GNN-based reaction fingerprints can also be used to navigate the chemical reaction space, which we demonstrate by querying for similar reactions. The strategy can be readily applied to other predictive reaction problems to uncover the power of unlabelled data for learning better models with a limited supply of labels.Contrastive pretraining of chemical reactions by matching augmented reaction representations to improve machine learning performance on small reaction datasets. 相似文献
7.
Maya Mincheva 《Journal of mathematical chemistry》2012,50(5):1111-1125
It is well known that oscillations in models of biochemical reaction networks can arise as a result of a single negative cycle. On the other hand, methods for finding general network conditions for potential oscillations in large biochemical reaction networks containing many cycles are not well developed. A biochemical reaction network with any number of species is represented by a simple digraph and is modeled by an ordinary differential equation (ODE) system with non-mass action kinetics. The obtained graph-theoretic condition generalizes the negative cycle condition for oscillations in ODE models to the existence of a pair of subnetworks, where each subnetwork contains an even number of positive cycles. The technique is illustrated with a model of genetic regulation. 相似文献
8.
9.
10.
A variety of computational models have been introduced recently that are based on the properties of DNA. In particular, branched junction molecules and graphlike DNA structures have been proposed as computational devices, although such models have yet to be confirmed experimentally. DNA branched junction molecules have been used previously to form graph-like three-dimensional DNA structures, such as a cube and a truncated octahedron, but these DNA constructs represent regular graphs, where the connectivities of all of the vertexes are the same. Here, we demonstrate the construction of an irregular DNA graph structure by a single step of self-assembly. A graph made of five vertexes and eight edges was chosen for this experiment. DNA branched junction molecules represent the vertexes, and duplex molecules represent the edges; in contrast to previous work, specific edge molecules are included as components. We demonstrate that the product is a closed cyclic single-stranded molecule that corresponds to a double cover of the graph and that the DNA double helix axes represent the designed graph. The correct assembly of the target molecule has been demonstrated unambiguously by restriction analysis. 相似文献
11.
12.
An intrinsic property of potential energy surfaces (PES) that holds within the adiabatic approximation is established: its structural stability.We derive the condition that ensures this property: There cannot be any integral curve of the gradient field of the PES that connects two classical transition state configurations without passing through another critical configuration in between.Under this situation, we can establish a one-to-one correspondence: a whole class of adiabatic PES defining one reaction mechanism is associated to a directed graph. Thus, the problem of finding a-priori pathways involving a given number m of chemical species narrows down to a classifying certain directed graphs with m sinks. The combinatorial method is derived in this paper.Detailed examples on a-priori pathways for degenerate thermal rearrangements and on 1-2 hydrocarbon shifts are worked out and found in agreement with experimental evidence. 相似文献
13.
A novel topological index W(F) is defined by the matrices X, W, and L as W(F) = XWL. Where L is a column vector expressing the characteristic of vertices in the molecule; X is a row vector expressing the bonding characteristics between adjacent atoms; W is a reciprocal distance matrix. The topological index W(F), based on the distance-related matrix of a molecular graph, is used to code the structural environment of each atom type in a molecular graph. The good QSPR/QSAR models have been obtained for the properties such as standard formation enthalpy of inorganic compounds and methyl halides, retention indices of gas chromatography of multiple bond-containing hydrocarbons, aqueous solubility, and octanol/water partition of benzene halides. These models indicate that the idea of using multiple matrices to define the modified Wiener index is valid and successful. 相似文献
14.
Solid-phase mix-and-split combinatorial synthesis is often used to produce large arrays of compounds to be tested during the various stages of the drug development process. This method can be represented by a synthesis graph in which nodes correspond to grow operations and arcs to beads transferred among the different reaction vessels. In this work, we address the problem of designing such a graph which maximizes the number of produced target compounds (namely, compounds out of an input library of desired molecules), given constraints on the number of beads used for library synthesis and on the number of reaction vessels available for concurrent grow steps. We present a heuristic based on a discrete search for solving this problem, test our solution on several data sets, explore its behavior, and show that it achieves good performance. 相似文献
15.
16.
Shiang-Tai Lin Min-Kang Hsieh Chieh-Ming Hsieh Chan-Chia Hsu 《The Journal of chemical thermodynamics》2009,41(10):1145-1153
We propose a general approach for developing liquid activity coefficient models based on the concept of local composition that satisfy at least two important physical conditions: (1) the total number of neighboring molecules around one molecule of species A must be a constant at any temperature for all possible mixture compositions, and (2) the number of pairs between any two species A–B determined from the local composition of B around A must be the same as that of A around B. Most commonly used liquid activity coefficient models (such as UNIQUAC) satisfy condition (1) but fail to meet condition (2), and thus are considered as fundamentally flawed. We propose a general formulation for the local composition equation containing a symmetric function of species A and B which ensures condition (2) be always satisfied. We show that the composition and temperature dependence of the symmetric functions can be completely obtained from condition (1), resulting in a new class of excess Gibbs free energy models. It is found that such a model can quantitatively reproduce the results of Monte Carlo simulation for various types of lattice fluids, while conventional models are merely qualitative. Furthermore, such a model is more accurate than the UNIQUAC model in correlating experimental data, especially in the dilute regime. Therefore, models developed based on this approach are theoretically sound and potentially applicable to a broader range of conditions. 相似文献
17.
J. Brocas 《Journal of mathematical chemistry》1994,15(1):389-395
The concept of mode of rearrangement is used to analyse the connectedness of the reaction graph for the Cope rearrangement in bullvalene. 相似文献
18.
A generalization of Wegschcider's condition concerning equilibrium constants in chemically reacting systems is formulated, which is then proved to be a necessary and sufficient condition for detailed balancing. In order to include a large multitude of rate laws, a generalized mass action kinetics is considered which comprises usual mass action kinetics and all reversible enzyme kinetics and which is consistent with basic postulates of irreversible thermodynamics for ideal mixtures. Reaction systems of arbitrary stoichiometry are considered. They may contain reactants with fixed concentrations, as is characteristic for models of biochemical reaction networks. Existence, uniqueness, and global asymptotic stability of equilibrium states for reaction systems endowed with generalized man action kinetics are proved. Using these results, he generalized Wegscheider condition is shown to be a sufficient criterion for the applicability of the quasi-steady-state approximation. 相似文献
19.
Topological and geometric images of invariant and monovariant reactions with the participation of melt in quaternary systems were presented. A network of invariant points and monovariant lines characterizes the structure of the liquidus hypersurface. Each invariant reaction produces an association of four solid phases, which is an element of the solidus hypersurface. Its topology is conveniently described by a graph of adjacency of tetrahedra that partition the composition tetrahedron of a system with a given set of binary, ternary, and quaternary compounds. It was demonstrated that it is possible to construct a spatial network of invariant points and monovariant lines on the liquidus hypersurface using the tetrahedron incidence graph with consideration for the phase reactions occurring in the system. 相似文献