Abstract Graph neural networks (GNN) has been considered as an attractive modelling method for molecular property prediction, and numerous studies have shown that GNN could yield more promising results than traditional descriptor-based methods. In this study, based on 11 public datasets covering various endpoints, the predictive capacity computational efficiency of prediction models developed by eight machine learning (ML) algorithms, including four (SVM, XGBoost, RF DNN) graph-based (GCN,...

Accurate quantification of protein–ligand interactions remains a key challenge to structure-based drug design. However, traditional machine learning (ML)-based methods based on handcrafted descriptors, one-dimensional protein sequences, and/or two-dimensional graph representations limit their capability learn the generalized molecular in 3D space. Here, we proposed novel deep representation framework named InteractionGraphNet (IGN) from structures complexes. In IGN, two independent...

In many real-world tasks, multiple agents must learn to coordinate with each other given their private observations and limited communication ability. Deep multiagent reinforcement learning (Deep-MARL) algorithms have shown superior performance in such challenging settings. One representative class of work is value decomposition, which decomposes the global shared Q-value $Q_{tot}$ into individual Q-values $Q^{i}$ guide individuals' behaviors, i.e. VDN imposing an additive formation QMIX...

Abstract Graph neural networks (GNN) has been considered as an attractive modelling method for molecular property prediction, and numerous studies have shown that GNN could yield more promising results than traditional descriptor-based methods. In this study, based on 11 public datasets covering various endpoints, the predictive capacity computational efficiency of prediction models developed by eight machine learning (ML) algorithms, including four (SVM, XGBoost, RF DNN) graph-based (GCN,...

Abstract Accurate predictions of druggability and bioactivities compounds are desirable to reduce the high cost time drug discovery. After more than five decades continuing developments, quantitative structure–activity relationship (QSAR) methods have been established as indispensable tools that facilitate fast, reliable affordable assessments physicochemical biological properties in drug-discovery programs. Currently, there mainly two types QSAR methods, descriptor-based graph-based...

Development of accurate machine-learning-based scoring functions (MLSFs) for structure-based virtual screening against a given target requires large unbiased dataset with structurally diverse actives and decoys. However, most datasets the development MLSFs were designed traditional SFs may suffer from hidden biases data insufficiency. Hereby, we developed new approach named Topology-based Conformation-based decoys generation (TocoDecoy), which integrates two strategies to generate by...

Many deep learning (DL)-based molecular generative models have been proposed to design novel molecules. These may perform well on benchmarks, but they usually do not take real-world constraints into account, such as available training data set, synthetic accessibility, and scaffold diversity in drug discovery. In this study, a new algorithm, ChemistGA, was by combining the traditional heuristic algorithm with DL, which crossover of genetic (GA) redefined DL conjunction GA, an innovative...

We study noncommutative maximal inequalities and ergodic theorems for group actions on von Neumann algebras. Consider a locally compact G of polynomial growth with symmetric subset V. Let α be continuous action algebra M by trace-preserving automorphisms. then show that the operators defined Anx=1m(Vn)∫Vnαgxdm(g),x∈Lp(M),n∈N,1≤p≤∞, are weak type (1,1) strong (p,p) 1<p<∞. Consequently, sequence (Anx)n≥1 converges almost uniformly x∈Lp(M) 1≤p<∞. Also, we establish individual associated more...

Chatbot models have achieved remarkable progress in recent years but tend to yield contradictory responses. In this paper, we exploit the advantage of contrastive learning technique mitigate issue. To endow model with ability discriminating patterns, minimize similarity between target response and contradiction related negative example. The example is generated learnable latent noise, which receives feedback from pretrained critic. Experimental results show that our method helps avoid...

Fairness in recommendation has attracted increasing attention due to bias and discrimination possibly caused by traditional recommenders. In Interactive Recommender Systems (IRS), user preferences the system's fairness status are constantly changing over time. Existing fairness-aware recommenders mainly consider static settings. Directly applying existing methods IRS will result poor recommendation. To resolve this problem, we propose a reinforcement learning based framework, FairRec,...

Retrosynthesis prediction is a crucial task for organic synthesis. In this work, we propose template-free and Transformer-based method dubbed RetroPrime, integrating chemists’ retrosynthetic strategy of (1) decomposing molecule into synthons then (2) generating reactants by attaching leaving groups. These two steps are accomplished with versatile Transformer models, respectively. While RetroPrime performs competitively against all state-of-the art models on the standard USPTO-50K dataset, it...

Retrosynthesis prediction is a crucial task for organic synthesis. In this work, we propose template-free and Transformer-based method dubbed RetroPrime, integrating chemists’ retrosynthetic strategy of (1) decomposing molecule into synthons then (2) generating reactants by attaching leaving groups. These two steps are accomplished with versatile Transformer models, respectively. While RetroPrime performs competitively against all state-of-the art models on the standard USPTO-50K dataset, it...

Molecular de novo design is a critical yet challenging task in scientific fields, aiming to novel molecular structures with desired property profiles. Significant progress has been made by resorting generative models for graphs. However, limited attention paid hierarchical models, which can exploit the inherent structure (with rich semantic information) of graphs and generate complex molecules larger size that we shall demonstrate be difficult most existing models. The primary challenge...

We prove a Chernoff-type bound for sums of matrix-valued random variables sampled via regular (aperiodic and irreducible) finite Markov chain. Specially, consider walk on chain Hermitian function its state space. Our result gives exponentially decreasing bounds the tail distributions extreme eigenvalues sample mean matrix. proof is based matrix expander (regular undirected graph) Chernoff [Garg et al. STOC '18] scalar Chernoff-Hoeffding chains [Chung STACS '12]. can be applied to analyze...

Abstract Graph neural networks (GNN) has been considered as an attractive modelling method for molecular property prediction, and numerous studies have shown that GNN could yield more promising results than traditional descriptor-based methods. In this study, based on 11 public datasets covering various endpoints, the predictive capacity computational efficiency of prediction models developed by eight machine learning (ML) algorithms, including four (SVM, XGBoost, RF DNN) graph-based (GCN,...

The notion of word embedding plays a fundamental role in natural language processing (NLP). However, pre-training for very large-scale vocabulary is computationally challenging most existing methods. In this work, we show that with merely small fraction contexts (Q-contexts)which are typical the whole corpus (and their mutual information words), one can construct high-quality negligible errors. Mutual between and words be encoded canonically as sampling state, thus, Q-contexts fast...

Molecular de novo design is a critical yet challenging task in scientific fields, aiming to novel molecular structures with desired property profiles. Significant progress has been made by resorting generative models for graphs. However, limited attention paid hierarchical models, which can exploit the inherent structure (with rich semantic information) of graphs and generate complex molecules larger size that we shall demonstrate be difficult most existing models. The primary challenge...

Retrosynthesis prediction is a crucial task for organic synthesis. In this work, we propose template-free and Transformer-based method dubbed RetroPrime, integrating chemists’ retrosynthetic strategy of (1) decomposing molecule into synthons then (2) generating reactants by attaching leaving groups. These two steps are accomplished with versatile Transformer models, respectively. While RetroPrime performs competitively against all state-of-the art models on the standard USPTO-50K dataset, it...

In this paper, we present LiQuID, a tool for seeing lighting quality in design.Photographs are useful vehicles both describing and making assessments of architectural systems.A significant barrier to using photographs during the design process relates sheer volume renderings that needs be analyzed.Although there have been efforts produce novel visualization systems manage large sets photographs, research aims reduce complexity by classifying data into representative prototypes.A hypothetical...