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。     

Large-Scale Prediction of Drug-Target Interaction: a Data-Centric Review

The prediction of drug-target interactions (DTIs) is of extraordinary significance to modern drug discovery in terms of suggesting new drug candidates and repositioning old drugs. Despite technological advances, large-scale experimental determination of DTIs is still expensive and laborious. Effective and low-cost computational alternatives remain in strong need. Meanwhile, open-access resources have been rapidly growing with massive amount of bioactivity data becoming available, creating unprecedented opportunities for the development of novel in silico models for large-scale DTI prediction. In this work, we review the state-of-the-art computational approaches for identifying DTIs from a data-centric perspective: what the underlying data are and how they are utilized in each study. We also summarize popular public data resources and online tools for DTI prediction. It is found that various types of data were employed including properties of chemical structures, drug therapeutic effects and side effects, drug-target binding, drug-drug interactions, bioactivity data of drug molecules across multiple biological targets, and drug-induced gene expressions. More often, the heterogeneous data were integrated to offer better performance. However, challenges remain such as handling data imbalance, incorporating negative samples and quantitative bioactivity data, as well as maintaining cross-links among different data sources, which are essential for large-scale and automated information integration.     

Recent developments in antithrombotic therapy: will sodium warfarin be a drug of the past?

Warfarin and heparin have formed the mainstay in the prophylaxis of deep vein thrombosis (DVT), stroke prevention in atrial fibrillation, and treatment of thromboembolic disease (TED). However, these choices are hampered by difficult administration, interactions with other medications, side effect profile, and limited indications for treatment. Anti-factor Xa (anti-Xa) inhibitors have already entered the drug market with the drug Fondaparinux being the first anti-Xa inhibitor to be approved for use in the U.S. by the Food and Drug Administration (FDA), and other drugs such as idraparinux being currently in development. A new class of medications, known as direct thrombin inhibitors (DTI), includes the parental agents lepirudin, argatroban and bivalirudin which have been approved by the FDA and the oral agents ximelagatran, melagatran and dabigatran. The latter three drugs which are oral DTIs may soon replace warfarin and heparin as the preferred medications for DVT prophylaxis and for reducing the relative risk of stroke. These drugs do not rely on blocking serine proteases nor do they require a co-factor (antithrombin III) like unfractionated heparin (UFH) or low molecular weight heparin (LMWH). DTIs are rapid in onset, easy to administer, do not interact with other medications or foods, have limited side effects, and can be administered in a fixed dose. The DTI ximelagatran has already been approved in several European and Asian countries, and over a dozen randomized clinical trials have been conducted demonstrating its performance to be on par with warfarin. However, approval by the FDA in the U.S. remains pending in view of reported incidences of elevations in hepatic enzymes that are currently under evaluation. This review examines the role of DTIs in the prevention and treatment of TED and the recent patents reported in the literature.     

In silico prediction of drug solubility in water-dioxane mixtures using the Jouyban-Acree model

A numerical method based on the Jouyban-Acree model was presented for prediction of drug solubility in water-dioxane mixtures at various temperatures. The method requires drug solubility in monosolvent systems, i.e. two data points for each temperature of interest. The mean percentage deviation (MPD) of predicted solubilities was calculated to show the accuracy of the predicted data and 27% was found as the average MPD for 36 data sets studied. The proposed numerical method reduced the number of required experimental data from five to two points and could also be extended to predict solubility at various temperatures.     

Novel oral anticoagulants in the treatment of acute coronary syndromes: is there any room for new anticoagulants?

Thrombosis plays a key role in the pathophysiology of acute coronary syndromes (ACS). The management of patients with ACS includes interventional procedures and use of antithrombotic agents acutely, and dual antiplatelet therapy (aspirin and a P2Y12 receptor antagonist) for secondary prevention. However, patients with recent ACS remain at a substantial residual risk for recurrent ischemic events or death. The idea of follow-up treatment with an oral anticoagulant on top of standard therapy seems promising. Warfarin was the first oral anticoagulant thoroughly investigated in this direction, but the widespread long-term use of warfarin in ACS has been limited by challenges associated with pharmacodynamic/pharmacokinetic deficiencies of the drug and the risk of bleeding. Novel oral anticoagulants, such as direct thrombin inhibitors (DTIs) and FXa inhibitors overcome the downsides of VKAs. Ximelagatran was the first DTI, investigated and proven to be effective in prevention of recurrent ischemic events in ACS patients, but the drug association with hepatotoxicity prompted its withdrawal. Dabigatran etexilate, apixaban, darexaban (YM150) and TAK-442 were studied in phase II dose-escalation trials in order to determine the balance between clinical effectiveness and bleeding risk in daily use with dual antiplatelet therapy, with both positive and negative results. Rivaroxaban is the only agent that completed a phase III trial, showing reduction in recurrent ischemic events rate and death from cardiovascular causes as well as all-cause death. This review summarizes the data from completed and ongoing clinical trials of the new oral anticoagulants in patients with ACS.     

Development of a decision tree to classify the most accurate tissue-specific tissue to plasma partition coefficient algorithm for a given compound

Physiologically based pharmacokinetic (PBPK) modeling is a tool used in drug discovery and human health risk assessment. PBPK models are mathematical representations of the anatomy, physiology and biochemistry of an organism and are used to predict a drug’s pharmacokinetics in various situations. Tissue to plasma partition coefficients (Kp), key PBPK model parameters, define the steady-state concentration differential between tissue and plasma and are used to predict the volume of distribution. The experimental determination of these parameters once limited the development of PBPK models; however, in silico prediction methods were introduced to overcome this issue. The developed algorithms vary in input parameters and prediction accuracy, and none are considered standard, warranting further research. In this study, a novel decision-tree-based Kp prediction method was developed using six previously published algorithms. The aim of the developed classifier was to identify the most accurate tissue-specific Kp prediction algorithm for a new drug. A dataset consisting of 122 drugs was used to train the classifier and identify the most accurate Kp prediction algorithm for a certain physicochemical space. Three versions of tissue-specific classifiers were developed and were dependent on the necessary inputs. The use of the classifier resulted in a better prediction accuracy than that of any single Kp prediction algorithm for all tissues, the current mode of use in PBPK model building. Because built-in estimation equations for those input parameters are not necessarily available, this Kp prediction tool will provide Kp prediction when only limited input parameters are available. The presented innovative method will improve tissue distribution prediction accuracy, thus enhancing the confidence in PBPK modeling outputs.     

Machine learning techniques for in silico modeling of drug metabolism

The computational assessment of drug metabolism has gained considerable interest in pharmaceutical research. Amongst others, machine learning techniques have been employed to model relationships between the chemical structure of a compound and its metabolic fate. Examples for these techniques, which were originally developed in fields far from drug discovery, are artificial neural networks or support vector machines. This paper summarizes the application of various machine learning techniques to predict the interaction between organic molecules and metabolic enzymes. More complex endpoints such as metabolic stability or in vivo clearance will also be addressed. It is shown that the prediction of metabolic endpoints with machine learning techniques has made considerable progress over the past few years. Depending on the procedure used, either classification or quantitative prediction is possible for even large and diverse compound sets. Together with the expanding experimental data basis, these in silico methods have become valuable tools in the drug discovery and development process.     

Prediction of drug clearance in children: impact of allometric exponents, body weight, and age

In recent years, with the advent of pediatric exclusivity and requirements for conducting clinical studies involving children, emphasis has been placed on finding safe and efficacious doses of drugs for children. It has been suggested that one can predict the clearance (CL) of a drug in children according to this equation: CL in the child = Adult CL * (Weight of the child/70)0.75. In light of the controversy surrounding the exponent of 0.75 for the prediction of clearance, the objectives of the study were as follows: (1) to develop allometric equations based on body weight or age to predict clearance of a drug in children; (2) to determine if the fixed exponent of 0.75 is a suitable exponent for the prediction of clearance in children from adult data, as compared with the allometric exponent generated for individual drugs; (3) to determine if the allometric equation generated on the basis of age predicts clearance in children better or worse than the allometric equation generated on the basis of body weight; and (4) to propose a new approach based on the findings of the current evaluation. Five methods were used to predict drug clearance in children. Six drugs were used in the evaluation, and drug clearance in each child was predicted for a given drug. Besides evaluating the exponent of 0.75, allometric equations were developed using double log plots of clearance versus body weight or age. The exponents of the allometric equations were then used to predict drug clearance by replacement of 0.75 in the aforementioned equation. The results of the study indicate that 0.75 is not the best exponent for prediction of drug clearance in children, and a more suitable approach is to develop an allometric relationship for a given drug in children. For all 6 drugs, there were 77 children in whom the clearance was predicted. There were 48 observations for which error in the predicted clearance was 50% or more with use of the exponent 0.75, whereas there were only 13 observations with prediction error > or = 50% when 0.75 was replaced by an allometric exponent developed for a given drug. In order to predict drug clearance in children with reasonable accuracy, an allometric equation should be developed for every drug and the exponent 0.75 should be replaced by the exponent of the allometric equation developed for that drug.     

Predicting drug clearance from recombinantly expressed CYPs: intersystem extrapolation factors

1. Recombinantly expressed human cytochromes P450 (rhCYPs) have been underused for the prediction of human drug clearance (CL). 2. Differences in intrinsic activity (per unit CYP) between rhCYP and human liver enzymes complicate the issue and these discrepancies have not been investigated systematically. We define intersystem extrapolation factors (ISEFs) that allow the use of rhCYP data for the in vitro-in vivo extrapolation of human drug CL and the variance that is associated with interindividual variation of CYP abundance due to genetic and environmental effects. 3. A large database (n = 451) of metabolic stability data has been compiled and used to derive ISEFs for the most commonly used expression systems and CYP enzymes. 4. Statistical models were constructed for the ISEFs to determine major covariates in order to optimize experimental design to increase prediction accuracy. 5. Suggestions have been made for the conduct of future studies using rhCYP to predict human drug clearance.     

基于GM(1,1)预测模型的中药材价格指数预测

目的:选择合适的预测模型对中药材价格进行预测。方法:应用离散形式的差分方程求解法建立中药材价格指数的灰色GM(1,1)差分预测模型,用程序简洁、预测精度高的美国The MathWorks公司出品的商业数学软件(MATLAB)实现模型算法。结果:模型预测结果显示,小误差几率、均方差比检验等级为好,相对误差检验等级为合格,可见预测效果十分理想。结论:该模型适用中药材价格指数的中期预测。     

Intracerebral microdialysis technique and its application on brain pharmacokinetic-pharmacodynamic study

Intracerebral microdialysis (IC-MD) has been developed as a well-validated and powerful technique for decades. As a practical sampling tool, it can gain the continuous dialysates of endogenous and exogenous substances in extracellular fluid (ECF) of awake freely moving animals. Also, variform IC-MD probes (IC-MDPs) have grown more exquisite. The implantation of the IC-MDP in certain tissue of brain allows monitor drug distribution and measure drug and corresponding neurotransmitters levels in brain ECF after administration for brain pharmacokinetic-pharmacodynamic (B-PK-PD) study. So it is suitable for IC-MD to B-PK-PD study (IC-MD/B-PK-PD). The performance of IC-MD/B-PK-PD can not only elevate the degree of precision and accuracy of experimental data, minimize the individual difference by reduced number of animals, but also give important information for the prediction and optimization of drug effective dose in preclinical study. In this review, we have discussed various IC-MD/B-PK-PD studies of analgesic, antiepileptic and antidepressant drug. The role of IC-MD/B-PK-PD in confirming and assessing the drug effect before clinic trials is highlighted.     

Analytical pharmacology: the impact of numbers on pharmacology

Analytical pharmacology strives to compare pharmacological data to detailed quantitative models. The most famous tool in this regard is the Black/Leff operational model, which can be used to quantify agonism in a test system and predict it in any other system. Here we give examples of how and where analytical pharmacology has been used to classify drugs and predict mechanism of action in pharmacology. We argue for the importance of analytical pharmacology in drug classification and in prediction of drug mechanisms of action. Although some of the specifics of Black's models have been updated to account for new developments, the principles of analytical pharmacology should shape drug discovery for many years to come.     

自回归整合移动平均模型在医院药库采购预测中的应用

目的:探讨利用自回归整合移动平均模型(ARIMA)预测的采购新模式提高医院药库工作质量和效率。方法:采集2008年01～47周药品消耗数据,根据ABC分类法确定A类品种,并从中随机抽样10个品种,基于2008年01～44周的数据,应用SPSS13软件作ARIMA模型建模拟合,用所得到的模型作45～47周预测,并比较实际消耗数据。结果:利用ARIMA模型预测的采购件数与实际消耗数基本相符,数量预测准确率为89.19%,整件预测准确率为97.56%。结论:ARIMA模型能够很好地拟合并可获得较高的中短期预测精度,能够为采购提供科学合理的决策支持,做到既不断货也不积压,合理控制药品库存量。     

Entering the era of computationally driven drug development

Abstract

Historically, failure rates in drug development are high; increased sophistication and investment throughout the process has shifted the reasons for attrition, but the overall success rates have remained stubbornly and consistently low. Only 8% of new entities entering clinical testing gain regulatory approval, indicating that significant obstacles still exist for efficient therapeutic development. The continued high failure rate can be partially attributed to the inability to link drug exposure with the magnitude of observed safety and efficacy-related pharmacodynamic (PD) responses; frequently, this is a result of nonclinical models exhibiting poor prediction of human outcomes across a wide range of disease conditions, resulting in faulty evaluation of drug toxicology and efficacy. However, the increasing quality and standardization of experimental methods in preclinical stages of testing has created valuable data sets within companies that can be leveraged to further improve the efficiency and accuracy of preclinical prediction for both pharmacokinetics (PK) and PD. Models of Quantitative structure–activity relationships (QSAR), physiologically based pharmacokinetics (PBPK), and PK/PD relationships have also improved efficiency. Founded on a core understanding of biochemistry and physiological interactions of xenobiotics, these in silico methods have the potential to increase the probability of compound success in clinical trials. Integration of traditional computational methods with machine-learning approaches and existing internal pharma databases stands to make a fundamental impact on the speed and accuracy of predictions during the process of drug development and approval.     

Data fusion of Fourier transform infrared spectra and powder X-ray diffraction patterns for pharmaceutical mixtures

Fusing complex data from two disparate sources has been demonstrated to improve the accuracy in quantifying active ingredients in mixtures of pharmaceutical powders. A four-component simplex-centroid design was used to prepare blended powder mixtures of acetaminophen, caffeine, aspirin and ibuprofen. The blends were analyzed by Fourier transform infra-red spectroscopy (FTIR) and powder X-ray diffraction (PXRD). The FTIR and PXRD data were preprocessed and combined using two different data fusion methods: fusion of preprocessed data (FPD) and fusion of principal component scores (FPCS). A partial least square (PLS) model built on the FPD did not improve the root mean square error of prediction. However, a PLS model built on the FPCS yielded better accuracy prediction than PLS models built on individual FTIR and PXRD data sets. The improvement in prediction accuracy of the FPCS may be attributed to the removal of noise and data reduction associated with using PCA as a preprocessing tool. The present approach demonstrates the usefulness of data fusion for the information management of large data sets from disparate sources.     

Prediction of skin sensitization potency using machine learning approaches

The replacement of animal use in testing for regulatory classification of skin sensitizers is a priority for US federal agencies that use data from such testing. Machine learning models that classify substances as sensitizers or non‐sensitizers without using animal data have been developed and evaluated. Because some regulatory agencies require that sensitizers be further classified into potency categories, we developed statistical models to predict skin sensitization potency for murine local lymph node assay (LLNA) and human outcomes. Input variables for our models included six physicochemical properties and data from three non‐animal test methods: direct peptide reactivity assay; human cell line activation test; and KeratinoSens™ assay. Models were built to predict three potency categories using four machine learning approaches and were validated using external test sets and leave‐one‐out cross‐validation. A one‐tiered strategy modeled all three categories of response together while a two‐tiered strategy modeled sensitizer/non‐sensitizer responses and then classified the sensitizers as strong or weak sensitizers. The two‐tiered model using the support vector machine with all assay and physicochemical data inputs provided the best performance, yielding accuracy of 88% for prediction of LLNA outcomes (120 substances) and 81% for prediction of human test outcomes (87 substances). The best one‐tiered model predicted LLNA outcomes with 78% accuracy and human outcomes with 75% accuracy. By comparison, the LLNA predicts human potency categories with 69% accuracy (60 of 87 substances correctly categorized). These results suggest that computational models using non‐animal methods may provide valuable information for assessing skin sensitization potency. Copyright © 2017 John Wiley & Sons, Ltd.