基于人工智能的药物-靶标相互作用预测

目的 建立一个高效的药物-靶标相互作用预测分类模型，为生物实验提供有力的补充工具。方法 研究开发一种基于深度学习的方法来预测药物-靶标相互作用：通过引入高维分子指纹和蛋白质描述符，并应用概率矩阵分解算法生成负样本集，构建一个高效的药物-靶标相互作用预测分类模型。结果 与其他已报道的方法相比，本方法具有可比性或优越性，预测准确性、特异性、敏感性以及AUC值均>90%，提示该方法在药物靶标预测方面具有良好的应用前景。结论 人工智能深度学习模型以及概率矩阵分解算法的结合有助于解决药物-靶标相互作用预测精度低、负样本选择不合理等问题。     

The analysis of the drug–targets based on the topological properties in the human protein–protein interaction network

Analyzing topological properties of drug-target proteins in the biology network is very helpful in understanding the mechanism of drug action. However, comprehensive studies to elaborately characterize the biological network features of drug-target proteins are still lacking. In this paper, we compared the topological properties of drug–targets with those of the non–drug-target sets, by mapping the drug–targets in DrugBank to the human protein interaction network. The results indicate that the topological properties of drug-targets are significantly distinguishable from those of non–drug-targets. Moreover, the potential possibility of drug-target prediction based on these properties is discussed. All proteins in the interaction network were ranked by their topological properties. Among the top 200 proteins, 94 overlapped with drug-targets in DrugBank and some novel predictions were found to be drug–targets in public literatures and other databases. In conclusion, our method explores the topological properties of drug-targets in the human protein interaction network by exploiting the large–scale drug-targets and protein interaction data.     

药物分子与靶蛋白相互作用的研究进展

药物发挥作用是以药物分子与其靶标的相互作用为基础的，对药物-靶点相互作用的定性分析与定量检测贯穿于从新药筛选发现到走向临床的整个过程。经过几十年的发展，研究药物分子与靶蛋白间相互作用的手段已经从传统的生化实验方法转变为以先进的分子生物学、生物物理学理论为支撑的高效、准确、多样化的技术体系。笔者从靶点发现与验证、亲和力测定、相互作用位点与结构分析几个方面对代表性的方法和技术进行介绍，以期为药物研发与机制探索提供参考。     

Target-mediated drug disposition and dynamics

Nonlinear pharmacokinetics and pharmacodynamics may result from several capacity-limited processes and often represent complicating factors in characterizing the pharmacological properties of drugs. Target-mediated drug disposition (TMDD) corresponds to a special case wherein a significant proportion of a drug (relative to dose) is bound with high affinity to a pharmacological target, such that this interaction is reflected in the pharmacokinetic properties of the drug. Dose-dependent effects on apparent pharmacokinetic parameters may manifest, including the steady-state volume of distribution and total systemic clearance. Although a few small molecular weight compounds have been identified to exhibit TMDD, the incidence of TMDD is likely to increase particularly among emerging biotechnology pharmaceuticals. The goal of this commentary is to describe the basic tenets of TMDD and discuss several mathematical modeling approaches for characterizing this phenomenon. Whereas traditional pharmacokinetic/pharmacodynamic models assume that the amount of the drug-target complex is negligible relative to the total amount of drug in the body, integrated mechanism-based models of TMDD incorporate the binding and stoichiometry of drug-target binding. These models may be utilized to infer the time-course of inaccessible system variables, such as the in vivo density of the drug-target complex, and provide a suitable platform for ascertaining the apparent pharmacodynamic implications of TMDD.     

网络药理学:认识药物及发现药物的新理念

网络药理学是指将药物作用网络与生物网络整合在一起,分析药物在此网络中与特定节点或模块的相互作用关系,从而理解药物和机体相互作用的科学。网络药理学突破传统的"一个药物一个靶标,一种疾病"理念,代表了现代生物医药研究的哲学理念与研究模式的转变。以系统生物学和网络生物学基本理论为基础的网络药理学具有整体性、系统性的特点,注重网络平衡(或鲁棒性)和网络扰动,强调理解某个单一生物分子(如基因、mRNA或蛋白等)在生物体系中的生物学地位和动力学过程要比理解其具体生物功能更为重要,揭示药物作用的生物学和动力学谱要比揭示其作用的单个靶标或几个"碎片化"靶标更重要,对认识药物和发现药物的理念产生了深远影响。     

Drug-target residence time and its implications for lead optimization

Much of drug discovery today is predicated on the concept of selective targeting of particular bioactive macromolecules by low-molecular-mass drugs. The binding of drugs to their macromolecular targets is therefore seen as paramount for pharmacological activity. In vitro assessment of drug-target interactions is classically quantified in terms of binding parameters such as IC(50) or K(d). This article presents an alternative perspective on drug optimization in terms of drug-target binary complex residence time, as quantified by the dissociative half-life of the drug-target binary complex. We describe the potential advantages of long residence time in terms of duration of pharmacological effect and target selectivity.     

Mechanistic models enable the rational use of in vitro drug-target binding kinetics for better drug effects in patients

Introduction: Drug-target binding kinetics are major determinants of the time course of drug action for several drugs, as clearly described for the irreversible binders omeprazole and aspirin. This supports the increasing interest to incorporate newly developed high-throughput assays for drug-target binding kinetics in drug discovery. A meaningful application of in vitro drug-target binding kinetics in drug discovery requires insight into the relation between in vivo drug effect and in vitro measured drug-target binding kinetics.

Areas covered: In this review, the authors discuss both the relation between in vitro and in vivo measured binding kinetics and the relation between in vivo binding kinetics, target occupancy and effect profiles.

Expert opinion: More scientific evidence is required for the rational selection and development of drug-candidates on the basis of in vitro estimates of drug-target binding kinetics. To elucidate the value of in vitro binding kinetics measurements, it is necessary to obtain information on system-specific properties which influence the kinetics of target occupancy and drug effect. Mathematical integration of this information enables the identification of drug-specific properties which lead to optimal target occupancy and drug effect in patients.     

The dynamics of drug-target interactions: drug-target residence time and its impact on efficacy and safety

The extent and duration of pharmacological action is determined by the lifetime of drug occupancy on a molecular target. This lifetime is defined by dynamic processes that control the rates of drug association and dissociation from the target. Recently, the term residence time has been coined to describe experimental measurements that can be related to the lifetime of the binary drug-target complex, and this in turn to durable, pharmacodynamic activity. The residence time concept and its impact on drug optimization are reviewed here. Examples are provided that demonstrate how a long residence time can improve drug efficacy in vivo. Additionally, optimization of drug-target residence time can help to mitigate off-target mediated toxicity, hence, improving drug safety and tolerability. Recent applications of the residence time concept to both drug discovery and development are also presented.     

药物重定位——网络药理学的重要应用领域

药物重定位是指发现已上市药物的新适应症,是网络药理学的重要应用领域。药物重定位策略是目前已知的药物研发策略中风险与效益比最好的策略之一,也是一种解决新药开发高投入低成功率困境的有效方法之一。目前已成功进行重定位的药物已超过百余种(国内有老药新用专著收载123种),药物重定位研究已超越了随机发现药物新适应症的阶段,进入了基于计算机技术的崭新研究阶段。现有研究方法主要有基于小分子(或配体)特征的方法、基于蛋白靶点(或受体)特征的方法、基于表型(或网络)特征的方法。随着对防治重大疾病有效药物需求的不断增加,以及系统生物学、计算生物学、网络药理学等相关学科的快速发展,面对新药研发难度越来越大的严峻形势,药物重定位已成为世界范围内关注的热点,在药物研发领域占据重要地位。     

The dynamics of drug-target interactions: drug-target residence time and its impact on efficacy and safety

The extent and duration of pharmacological action is determined by the lifetime of drug occupancy on a molecular target. This lifetime is defined by dynamic processes that control the rates of drug association and dissociation from the target. Recently, the term residence time has been coined to describe experimental measurements that can be related to the lifetime of the binary drug-target complex, and this in turn to durable, pharmacodynamic activity. The residence time concept and its impact on drug optimization are reviewed here. Examples are provided that demonstrate how a long residence time can improve drug efficacy in vivo. Additionally, optimization of drug-target residence time can help to mitigate off-target mediated toxicity, hence, improving drug safety and tolerability. Recent applications of the residence time concept to both drug discovery and development are also presented.     

Approximations of the target-mediated drug disposition model and identifiability of model parameters

Models for drugs exhibiting target-mediated drug disposition (TMDD) play an important role in the investigation of biological products (Mager and Jusko 2001). These models are often overparameterized and difficult to converge. A simpler quasi-equilibrium (QE) approximation of the general model has been suggested (Mager and Krzyzanski 2005), but even this simpler form can be overparameterized when, for example, drug target level is not available. This work (a) introduces quasi-steady-state (QSS) and Michaelis-Menten (MM) approximations of the TMDD model, (b) derives the relationships between the parameters of the TMDD, QE, QSS and MM models, (c) investigates the parameter ranges where the simplified approximations are equivalent to the TMDD model, (d) proposes an algorithm for establishing identifiability of these models, and (e) tests this algorithm on simulated datasets. The proposed QSS approximation is more general than the QE approximation: it degenerates into the QE approximation when the internalization rate of the drug-target complex is much smaller than its dissociation rate. The proposed identifiability analysis algorithm may be applied to provide justification for use of simplified approximations, avoiding use of incorrect parameter estimates of over-parameterized TMDD models while simultaneously saving time and resources required for the pharmacokinetics analysis of drugs with TMDD. The utility of the derived approximations and of the identifiability algorithm was demonstrated on the examples of the simulated data sets. The simulation examples indicated that the QSS model may be preferable to the QE model when the internalization rate of the drug-target complex significantly exceeds its dissociation rate. The MM approximation may be adequate when the drug concentration significantly exceeds the target concentrations or when the target occupancy is close to 100%.     

基于药-靶结合动力学的胰脂肪酶靶点占有率模型建立研究

目的：建立以胰脂肪酶和已上市药物为媒介的药-靶结合动力学模型,依据动力学参数与肠腔药动学构建靶点占有率模型,以评估药-靶结合动力学参数对体内药效的影响。方法：采用体外酶促反应体系测定胰脂肪酶抑制剂奥利司他和新利司他的半抑制浓度（IC₅₀）;通过反应进度曲线测定求解表观速率常数（k_obs）;采用快速稀释法测定药物-酶解离速率常数（k_off）;采用非线性拟合求解其他关键结合动力学参数;应用药物肠腔药代动力学模型,计算肠道中不同时间药物浓度;构建体内靶点占有率模型,并计算体内不同时间的靶点占有率。结果：测得奥利司他的IC₅₀为19.1 nmol·L^-1,新利司他的IC₅₀为76 nmol·L^-1;2个药物与胰脂肪酶呈时间、浓度依赖的缓慢结合反应,且结合反应属于两步结合模式,结合类型为机制B;可逆性研究结果显示奥利司他和新利司他为不可逆抑制;两药的结合速率常数（k_on）分别为76.5×10⁴和1.7×10⁴M^-1S^-1,k_off分别为0.33×10^-6和8.4×10^-6S^-1。奥利司他对脂肪酶的靶点占有率在24 h大于90%,新利司他对脂肪酶的靶点占有率在21 h时大于60%。结论：本研究所建立的基于药-靶结合动力学胰脂肪酶靶点占有率模型可用于评估动力学参数对体内药效的影响。     

Computational polypharmacology with text mining and ontologies

Huge volumes of data, produced by microarrays and next- generation sequencing, are now at the fingertips of scientists and allow to expand the scope beyond conventional drug de- sign. New promiscuous drugs directed at multiple targets promise increased therapeutic efficacy for treatment of multi- factorial diseases. At the same time, more systematic tests for unwanted side effects are now possible. In this paper, we focus on the application of text mining and ontologies to support experimental drug discovery. Text mining is a high- throughput technique to extract information from millions of scientific documents and web pages. By exploiting the vast number of extracted facts as well as the indirect links between them, text mining and ontologies help to generate new hypotheses on drug target interactions. We review latest applications of text mining and ontologies suitable for target and drug-target interaction discovery in addition to conventional approaches. We conclude that mining the literature on drugs and proteins offers unique opportunities to support the laborious and expensive process of drug development.     

DrugBank and its relevance to pharmacogenomics

DrugBank is a freely available web-enabled database that combines detailed drug data with comprehensive drug-target and drug-action information. It was specifically designed to facilitate in silico drug-target discovery, drug design, drug-metabolism prediction, drug-interaction prediction, and general pharmaceutical education. One of the most unique and useful components of the DrugBank database is the information it contains on drug metabolism, drug-metabolizing enzymes and drug-target polymorphisms. As pharmacogenomics is fundamentally concerned with the role of genes and genetic variation of how an individual responds to a drug, DrugBank is able to offer a convenient venue to explore pharmacogenomic questions in silico. This paper provides a brief overview on DrugBank and how it can facilitate pharmacogenomic research.     

The emergence of chemical genomics in drug discovery

The interaction of small organic molecules with proteins and other macromolecules is fundamental to drug action. Chemical genomics employs a combination of chemistry, genomics and informatics to study these drug-target interactions in a systematic and global manner in order to improve the efficiency of the drug discovery process.     

Camptothecins in clinical development

Following the realisation that DNA topoisomerase I is a useful therapeutic target to be exploited for the design of potential inhibitors, topoisomerase I inhibitors now represent an established class of effective agents. In spite of intense efforts in the field, only camptothecins have a clinical relevance. Several options in chemical manipulation of natural camptothecin have been explored to overcome the major drawbacks of the drug, which include water insolubility, lactone instability, reversibility of the drug-target interaction and drug resistance. Several analogues are currently in clinical development, including water soluble camptothecins, lipophilic camptothecins and polymer-bound camptothecins. The therapeutic advantages of novel camptothecins over the two analogues (topotecan and irinotecan) approved for clinical use remain to be defined. This article is an overview of the relevant features of the analogues that are undergoing clinical development.     

Pharmacogenetics of antiarrhythmic therapy

Individuals vary widely in their responses to therapy with most drugs. Indeed, responses to antiarrhythmic drugs are so highly variable that study of the underlying mechanisms has elucidated important lessons for understanding variable responses to drug therapy in general. Variability in drug response may reflect variability in the relationship between a drug dose and the concentrations of the drug and metabolite(s) at relevant target sites; this is termed pharmacokinetic variability. Another mechanism is that individuals vary in their response to identical exposures to a drug (pharmacodynamic variability). In this case, there may be variability in the target molecule(s) with which a drug interacts or, more generally, in the broad biological context in which the drug-target interaction occurs. Variants (polymorphisms and mutations) in the genes that encode proteins that are important for pharmacokinetics or for pharmacodynamics have now been described as important contributors to variable drug actions, including proarrhythmia, and these are described in this review. However, the translation of pharmacogenetics into clinical practice has been slow. To this end, the creation of large, well-characterised DNA databases and appropriate control groups, as well as large prospective trials to evaluate the impact of genetic variation on drug therapy, may hasten the impact of pharmacogenetics and pharmacogenomics in terms of delivering personalised drug therapy and to avoid therapeutic failure and serious side effects.     

Mapping of pharmacological space

The analysis of pharmacological space is becoming highly relevant in light of the emerging polypharmacology paradigm, that is, the increasing evidence that many drugs elicit therapeutic effects and adverse drug reactions through interactions with multiple targets. To better understand desired and undesired polypharmacology and identify new targets for existing drugs, computational methods are of critical importance. Herein we provide an overview of computational approaches for analyzing pharmacological space and put their opportunities and limitations in perspective. Insights into computational approaches for the study of target-ligand interactions and polypharmacology are provided and put into scientific context. The interplay between computational and experimental approaches is rationalized. Computational methods have become indispensable tools for the systematic analysis of drug-target interactions. Because currently most prominent predictive methods are knowledge-based, they are affected by data bias and sparseness. Predictions of drug-target interactions are already carried out on a large scale, but experimentally validated to a much lesser extent. In order to demonstrate true utility of pharmacological space analysis for drug discovery, it will be essential to closely interface computational and experimental target profiling efforts.     

整合化学和定量蛋白质组学方法用于鉴定药物靶点（英文）

Most drugs exert pharmacological effects through interaction with their target proteins.Therefore,drug target identification is a crucial step towards the understanding of the mechanism of drug action.It is also imperative to study the pharmacodynamics of a known drug,with an aim to discover the potentially unrevealed actions and thus refine its future clinical applications.Currently,drug-target identification is either through in vitro affinity chromatography-based approaches or in vivo activity-based protein profiling(ABPP)approaches.However,these approaches generally face difficulties discriminating specific drug targets from non-specific ones.To address this issue,we have come up with a strategy by coupling iTRAQTM(isobaric tags for relative and absolute quantitation)quantitative proteomics approach with clickable ABPP,to specifically and compre hensively identify drug targets in live cells.Using this approach,we identified the protein targets of andrographolide,a natural product with known anti-inflammation and anti-cancer effects,in live cancer cells.The identified target list not only confirmed the known functions of the drug but also revealed its potential novel application as a tumor metastasis inhibitor.We have also used this strategy,combining with a cleavable probe to identify the protein targets of aspirin and its binding sites.Our results revealed the roles of aspirin ininhibition of protein synthesis and induction of autophagy,which have been functionally validated.Our strategy is widely applicable to the identification of protein targets of covalent drugs.