基于支持向量机技术预测丙戊酸钠血药浓度

目的 基于支持向量机(SVM)技术,建立丙戊酸钠的血药浓度预测模型.方法 收集陆军军医大学第一附属医院2015年1月至2018年12月确诊为癫痫且服用丙戊酸钠缓释片的病人的血药浓度及16个血药浓度影响因素指标数据.利用随机数字表法将收集的206例病人共271个样本数据分为190个构成训练样本集以及81个构成测试样本集.基于SVM技术对190个训练样本进行训练,建立预测模型.再用外部验证法将81个测试样本的血药浓度模型预测值与实际观测值进行对比.结果 训练样本集和测试样本集中病人的各临床指标除胱抑素C外,其余指标差异无统计学意义(P≥0.05),训练样本集中病人胱抑素C为(1.17±1.23)mg/L,明显高于测试样本集中病人的(0.93±0.84)mg/L(P=0.012).基于SVM技术的血药浓度预测模型取得了较好的预测效果,模型预测值与实际观测值相对误差:小于5％的12个;5％～10％(含)的23个;10％～15％(含)的21个;15％～20％(含)的13个;20％～25％(含)的4个,超过25％的8个;平均相对误差为12.12％,相对误差小于20％(含)的样本占比达到85.18％.平均绝对误差为9.98 mg/L,绝对误差小于20 mg/L的样本占比达到95.06％.模型预测值与实际观测值的相关系数为0.788.结论 SVM技术在血药浓度预测方面具有良好的应用前景,基于该技术的丙戊酸钠血药浓度预测模型准确度较好,模型预测值与实际观测值的相关性较好,相对误差较小,可为临床制定个体化给药方案提供参考.     

Rapid and accurate assessment of seizure liability of drugs by using an optimal support vector machine method

Drug-induced seizures are a serious adverse effect and assessment of seizure risk usually takes place at the late stage of drug discovery process, which does not allow sufficient time to reduce the risk by chemical modification. Thus early identification of chemicals with seizure liability using rapid and cheaper approaches would be preferable. In this study, an optimal support vector machine (SVM) modeling method has been employed to develop a prediction model of seizure liability of chemicals. A set of 680 compounds were used to train the SVM model. The established SVM model was then validated by an independent test set comprising 175 compounds, which gave a prediction accuracy of 86.9%. Further, the SVM-based prediction model of seizure liability was compared with various preclinical seizure assays, including in vitro rat hippocampal brain slice, in vivo zebrafish larvae assay, mouse spontaneous seizure model, and mouse EEG model. In terms of predictability, the SVM model was ranked just behind the mouse EEG model, but better than the rat brain slice and zebrafish models. Nevertheless, the SVM model has considerable advantages compared with the preclinical seizure assays in speed and cost. In summary, the SVM-based prediction model of seizure liability established here offers potential as a cheaper, rapid and accurate assessment of seizure liability of drugs, which could be used in the seizure risk assessment at the early stage of drug discovery. The prediction model is freely available online at http://www.sklb.scu.edu.cn/lab/yangsy/download/ADMET/seizure_pred.tar.     

Qualitative identification of tea categories by near infrared spectroscopy and support vector machine

Near-infrared (NIR) spectroscopy has been successfully utilized for the rapid identification of green, black and Oolong tea. The spectral features of each tea category are reasonably differentiated in the NIR region, and the spectral differences provided enough qualitative spectral information for the identification of tea. Support vector machine (SVM) as the pattern recognition was applied to identify three tea categories in this study. The top five principal components (PCs) were extracted as the input of SVM classifiers by principal component analysis (PCA). The RBF SVM classifiers and the polynomial SVM classifiers were studied comparatively in this experiment. The best experimental results were obtained using the radial basis function (RBF) SVM classifier with sigma=0.5. The accuracies of identification were all more than 90% for three tea categories. Finally, compared with the back propagation artificial neural network (BP-ANN) approach, SVM algorithm showed its excellent generalization for identification results. The overall results show that NIR spectroscopy combined with SVM can be efficiently utilized for rapid and simple identification of the tea categories.     

In silico prediction of rhabdomyolysis of compounds by self-organizing map and support vector machine

Rhabdomyolysis is a potentially lethal syndrome resulting in leakage of myocyte intracellular contents into the plasma. Some drugs, such as lipid-lowering drugs and antihistamines, can cause rhabdomyolysis. In this work, a dataset containing 186 chemical compounds causing rhabdomyolysis and 117 drugs not causing rhabdomyolysis was collected. The dataset was split into a training set (containing 230 compounds) and a test set (containing 73 compounds). A Kohonen’s self-organizing map (SOM) and a support vector machine (SVM) were applied to develop classification models to differentiate compounds causing and not causing rhabdomyolysis. Using the SOM method, classification accuracies of 93.3% for the training set and 84.5% for the test set were achieved; using the SVM method, classification accuracies of 95.2% for the training set and 84.9% for the test set were achieved. In addition, the extended connectivity fingerprints (ECFP_4) for all the molecules were calculated and analyzed to find the important features of molecules relating to rhabdomyolysis.     

Predicting full thickness skin sensitization using a support vector machine

To assess the public’s propensity for allergic contact dermatitis (ACD), many alternatives to in vivo chemical screening have been developed which generally incorporate a small panel of cell surface and secreted dendritic cell biomarkers. However, given the underlying complexity of ACD, one cell type and limited cellular metrics may be insufficient to predict contact sensitizers accurately. To identify a molecular signature that can further characterize sensitization, we developed a novel system using RealSkin, a full thickness skin equivalent, in co-culture with MUTZ-3 derived Langerhan’s cells. This system was used to distinguish a model moderate pro-hapten isoeugenol (IE) and a model strong pre-hapten p-phenylenediamine (PPD) from irritant, salicylic acid (SA). Commonly evaluated metrics such as CD86, CD54, and IL-8 secretion were assessed, in concert with a 27-cytokine multi-plex screen and a functional chemotaxis assay. Data were analyzed with feature selection methods using ANOVA, hierarchical cluster analysis, and a support vector machine to identify the best molecular signature for sensitization. A panel consisting of IL-12, IL-9, VEGF, and IFN-γ predicted sensitization with over 90% accuracy using this co-culture system analysis. Thus, a multi-metric approach that has the potential to identify a molecular signature may be more predictive of contact sensitization.     

Predicting full thickness skin sensitization using a support vector machine

To assess the public’s propensity for allergic contact dermatitis (ACD), many alternatives to in vivo chemical screening have been developed which generally incorporate a small panel of cell surface and secreted dendritic cell biomarkers. However, given the underlying complexity of ACD, one cell type and limited cellular metrics may be insufficient to predict contact sensitizers accurately. To identify a molecular signature that can further characterize sensitization, we developed a novel system using RealSkin, a full thickness skin equivalent, in co-culture with MUTZ-3 derived Langerhan’s cells. This system was used to distinguish a model moderate pro-hapten isoeugenol (IE) and a model strong pre-hapten p-phenylenediamine (PPD) from irritant, salicylic acid (SA). Commonly evaluated metrics such as CD86, CD54, and IL-8 secretion were assessed, in concert with a 27-cytokine multi-plex screen and a functional chemotaxis assay. Data were analyzed with feature selection methods using ANOVA, hierarchical cluster analysis, and a support vector machine to identify the best molecular signature for sensitization. A panel consisting of IL-12, IL-9, VEGF, and IFN-γ predicted sensitization with over 90% accuracy using this co-culture system analysis. Thus, a multi-metric approach that has the potential to identify a molecular signature may be more predictive of contact sensitization.     

支持向量机与近红外光谱法鉴定大黄

目的：建立大黄真伪的鉴别方法。方法：本文对52个不同品种和不同产地的大黄样品进行了近红外谱图扫描，用支持向量机（SVM）中4种不同的核函数对近红外谱图进行了正品和非正品大黄的鉴别。结果：鉴别正确率均可达98．1％。结论：讨论了影响4个核函数预算的各个参数，表明多形式核函数更适合于本实验样品的鉴定。同时将此实验结果与用RBF神经网络鉴别的结果进行了比较，表明SVM使用简便，泛化能力强，核函数选择灵活性大，是大黄样本鉴别的简单可靠的方法。     

A novel approach using pharmacophore ensemble/support vector machine (PhE/SVM) for prediction of hERG liability

A novel approach by using a panel of plausible pharmacophore hypothesis candidates to constitute the pharmacophore ensemble (PhE) and subject them to regression by support vector machine (SVM) has been developed for predicting the liability of human ether-a-go-go-related gene (hERG). This PhE/SVM scheme takes into account the protein conformational flexibility while interacting with structurally diverse ligands, which is crucial yet often neglected by most of the analogue-based modeling methods. Thirty-nine molecules were carefully selected and cross-examined from the literature data for this study, of which 26 and 13 molecules were deliberately treated as the training set and the test set to generate the model and to validate the generated model, respectively. The final PhE/SVM model gave rise to an r(2) value of 0.97 for observed vs predicted pIC(50) values for the training set, a q(2) value of 0.89 by the 10-fold cross-validation and an r(2) value of 0.94 for the test set. Thus, this PhE/SVM model provides a fast and accurate tool for predicting liability of hERG and can be utilized to guide medicinal chemistry to avoid molecules with an inhibition potential of this potassium channel.     

Prediction of phospholipidosis-inducing potential of drugs by in vitro biochemical and physicochemical assays followed by multivariate analysis

An in vitro method to predict phospholipidosis-inducing potential of cationic amphiphilic drugs (CADs) was developed using biochemical and physicochemical assays. The following parameters were applied to principal component analysis, as well as physicochemical parameters: pK_a and clog P; dissociation constant of CADs from phospholipid, inhibition of enzymatic phospholipid degradation, and metabolic stability of CADs. In the score plot, phospholipidosis-inducing drugs (amiodarone, propranolol, imipramine, chloroquine) were plotted locally forming the subspace for positive CADs; while non-inducing drugs (chlorpromazine, chloramphenicol, disopyramide, lidocaine) were placed scattering out of the subspace, allowing a clear discrimination between both classes of CADs. CADs that often produce false results by conventional physicochemical or cell-based assay methods were accurately determined by our method. Basic and lipophilic disopyramide could be accurately predicted as a nonphospholipidogenic drug. Moreover, chlorpromazine, which is often falsely predicted as a phospholipidosis-inducing drug by in vitro methods, could be accurately determined. Because this method uses the pharmacokinetic parameters pK_a, clog P, and metabolic stability, which are usually obtained in the early stages of drug development, the method newly requires only the two parameters, binding to phospholipid, and inhibition of lipid degradation enzyme. Therefore, this method provides a cost-effective approach to predict phospholipidosis-inducing potential of a drug.     

Exploring QSARs for antiviral activity of 4-alkylamino-6-(2-hydroxyethyl)-2-methylthiopyrimidines by support vector machine

The support vector machine, which is a novel algorithm from the machine learning community, was used to develop quantitative structure activity relationship models to predict the antiviral activity of 4-alkylamino-6-(2-hydroxyethyl)-2-methylthiopyrimidines. The genetic algorithm was employed to select the variables that resulted in the best-fitted models. A comparison between the obtained results using support vector machine with those of multiple linear regression revealed that support vector machine model was much better than multiple linear regression. The root mean square errors of the training set and the test set for support vector machine model were calculated to be 0.102 and 0.205, and the correlation coefficients (r2) were 0.956 and 0.852, respectively. Furthermore, the obtained statistical parameter of leave-one-out (LOO) and leave-group-out (LGO) cross-validation test on support vector machine model were 0.893 and 0.881, respectively, which prove the reliability of this model. The results suggest that branching, volume and lipophilicity are the main independent factors contributing to the antiviral activities of the studied compounds.     

In vitro approach to assess the potential for risk of idiosyncratic adverse reactions caused by candidate drugs

Idiosyncratic adverse drug reactions (IADRs) in humans can result in a broad range of clinically significant toxicities leading to attrition during drug development as well as postlicensing withdrawal or labeling. IADRs arise from both drug and patient related mechanisms and risk factors. Drug related risk factors, resulting from parent compound or metabolites, may involve multiple contributory mechanisms including organelle toxicity, effects related to compound disposition, and/or immune activation. In the current study, we evaluate an in vitro approach, which explored both cellular effects and covalent binding (CVB) to assess IADR risks for drug candidates using 36 drugs which caused different patterns and severities of IADRs in humans. The cellular effects were tested in an in vitro Panel of five assays which quantified (1) toxicity to THLE cells (SV40 T-antigen-immortalized human liver epithelial cells), which do not express P450s, (2) toxicity to a THLE cell line which selectively expresses P450 3A4, (3) cytotoxicity in HepG2 cells in glucose and galactose media, which is indicative of mitochondrial injury, (4) inhibition of the human bile salt export pump, BSEP, and (5) inhibition of the rat multidrug resistance associated protein 2, Mrp2. In addition, the CVB Burden was estimated by determining the CVB of radiolabeled compound to human hepatocytes and factoring in both the maximum prescribed daily dose and the fraction of metabolism leading to CVB. Combining the aggregated results from the in vitro Panel assays with the CVB Burden data discriminated, with high specificity (78%) and sensitivity (100%), between 27 drugs, which had severe or marked IADR concern, and 9 drugs, which had low IADR concern, we propose that this integrated approach has the potential to enable selection of drug candidates with reduced propensity to cause IADRs in humans.     

A new approach to anti-inflammatory drugs.

All aspirin-like drugs so far tested inhibit prostaglandin biosynthesis but do not prevent the generation of the hydroxy acid 12-l-hydroxyeicosatetraenoic acid (HETE) by arachidonate lipoxygenase. HETE is chemotactic for polymorphonuclear leukocytes, and the failure to inhibit lipoxygenase may explain why the aspirin-like drugs have little or no effect on leukocyte migration at doses which are both anti-inflammatory and inhibit prostaglandin synthesis in vivo. 3-Amino-1-[m-(trifluoromethyl)-phenyl]-2-pyrazoline (BW755C) inhibits both pathways of arachidonic acid metabolism in vitro and causes a dose-dependent reduction in carrageenin-induced oedema in the rat paw. BW755C also reduces prostaglandin concentration in inflammatory exudates and has a significantly greater effect on leukocyte migration than indomethacin. The dual inhibition of arachidonate cyclo-oxygenase (prostaglandin synthetase) and lipoxygenase could lead, therefore, to increased anti-inflammatory activity.     

Stability of advanced life support drugs in the field.

PURPOSE: The effects of wide temperature variations on the stability of atropine, epinephrine, and lidocaine stored under field conditions in advanced life support (ALS) paramedic units were evaluated. METHODS: Vehicles from various ALS paramedic units were selected throughout Los Angeles County, California, including desert, marine, and helicopter-based divisions. A temperature-recording device was placed in the compartment where drugs are stored and used to record and store temperature data at 15-minute intervals. Three autoinjector-style syringes of atropine, epinephrine, and lidocaine were taken from stock for each ALS unit and placed in each vehicle, while three control syringes were stored in the laboratory under controlled conditions. Six samples of each drug were withdrawn at time 0 and on days 5, 10, 15, 30, and 45. Samples were analyzed using high-performance liquid chromatography. Stock solutions, created using analytical grade atropine, epinephrine, and lidocaine, were used to construct 5-point standard curves to determine the drug concentration of each sample. RESULTS: Seven sites exceeded 104 degrees F (40 degrees C) for as little as 30 minutes and as long as 795 minutes. Ten of the sites achieved a mean kinetic temperature (MKT) above 77 degrees F (25 degrees C), with the highest MKT calculated being 84.1 degrees F (28.9 degrees C) over a 45-day period. There was no evidence of drug degradation at any site, at any temperature, or at any time point. CONCLUSION: Atropine, epinephrine, and lidocaine can be stored at temperatures of up to 84.1 degrees F (28.9 degrees C) for up to 45 days and tolerate temperature spikes of up to 125 degrees F (51.7 degrees C) for a cumulative time of 795 minutes (13.25 hours) without undergoing degradation.