为了提高蛋白质相互作用位点预测的精度,该文基于蛋白质位置特异性得分矩阵(Position specific scoring matrix,PSSM),提出了一种新的加权得分矩阵直方图特征表示方法;针对训练数据的极端不平衡,结合下采样和分类器集成方法,训练随机森...为了提高蛋白质相互作用位点预测的精度,该文基于蛋白质位置特异性得分矩阵(Position specific scoring matrix,PSSM),提出了一种新的加权得分矩阵直方图特征表示方法;针对训练数据的极端不平衡,结合下采样和分类器集成方法,训练随机森林集成分类器。相对于传统的特征,该文所提新特征具有更低的维数,同时拥有更好的鉴别性。分类器集成则缓解了下采样带来的信息丢失,并提高了分类精度。实验结果验证了所述方法是有效的,在标准数据集上的结果优于其他最新的蛋白质相互作用位点预测方法。展开更多
The knowledge of subnuclear localization in eukaryotic cells is indispensable for under-standing the biological function of nucleus, genome regulation and drug discovery. In this study, a new feature representation wa...The knowledge of subnuclear localization in eukaryotic cells is indispensable for under-standing the biological function of nucleus, genome regulation and drug discovery. In this study, a new feature representation was pro-posed by combining position specific scoring matrix (PSSM) and auto covariance (AC). The AC variables describe the neighboring effect between two amino acids, so that they incorpo-rate the sequence-order information;PSSM de-scribes the information of biological evolution of proteins. Based on this new descriptor, a support vector machine (SVM) classifier was built to predict subnuclear localization. To evaluate the power of our predictor, the benchmark dataset that contains 714 proteins localized in nine subnuclear compartments was utilized. The total jackknife cross validation ac-curacy of our method is 76.5%, that is higher than those of the Nuc-PLoc (67.4%), the OET- KNN (55.6%), AAC based SVM (48.9%) and ProtLoc (36.6%). The prediction software used in this article and the details of the SVM parameters are freely available at http://chemlab.scu.edu.cn/ predict_SubNL/index.htm and the dataset used in our study is from Shen and Chou’s work by downloading at http://chou.med.harvard.edu/ bioinf/Nuc-PLoc/Data.htm.展开更多
Identification of the drug-binding residues on the surface of proteins is a vital step in drug discovery and it is important for understanding protein function. Most previous researches are based on the structural inf...Identification of the drug-binding residues on the surface of proteins is a vital step in drug discovery and it is important for understanding protein function. Most previous researches are based on the structural information of proteins, but the structures of most proteins are not available. So in this article, a sequence-based method was proposed by combining the support vector machine (SVM)-based ensemble learning and the improved position specific scoring matrix (PSSM). In order to take the local environment information of a drug-binding site into account, an improved PSSM profile scaled by the sliding window and smoothing window was used to improve the prediction result. In addition, a new SVM-based ensemble learning method was developed to deal with the imbalanced data classification problem that commonly exists in the binding site predictions. When performed on the dataset of 985 drug-binding residues, the method achieved a very promising prediction result with the area under the curve (AUC) of 0.9264. Furthermore, an independent dataset of 349 drug- binding residues was used to evaluate the pre- diction model and the prediction accuracy is 84.68%. These results suggest that our method is effective for predicting the drug-binding sites in proteins. The code and all datasets used in this article are freely available at http://cic.scu.edu.cn/bioinformatics/Ensem_DBS.zip.展开更多
文摘为了提高蛋白质相互作用位点预测的精度,该文基于蛋白质位置特异性得分矩阵(Position specific scoring matrix,PSSM),提出了一种新的加权得分矩阵直方图特征表示方法;针对训练数据的极端不平衡,结合下采样和分类器集成方法,训练随机森林集成分类器。相对于传统的特征,该文所提新特征具有更低的维数,同时拥有更好的鉴别性。分类器集成则缓解了下采样带来的信息丢失,并提高了分类精度。实验结果验证了所述方法是有效的,在标准数据集上的结果优于其他最新的蛋白质相互作用位点预测方法。
文摘The knowledge of subnuclear localization in eukaryotic cells is indispensable for under-standing the biological function of nucleus, genome regulation and drug discovery. In this study, a new feature representation was pro-posed by combining position specific scoring matrix (PSSM) and auto covariance (AC). The AC variables describe the neighboring effect between two amino acids, so that they incorpo-rate the sequence-order information;PSSM de-scribes the information of biological evolution of proteins. Based on this new descriptor, a support vector machine (SVM) classifier was built to predict subnuclear localization. To evaluate the power of our predictor, the benchmark dataset that contains 714 proteins localized in nine subnuclear compartments was utilized. The total jackknife cross validation ac-curacy of our method is 76.5%, that is higher than those of the Nuc-PLoc (67.4%), the OET- KNN (55.6%), AAC based SVM (48.9%) and ProtLoc (36.6%). The prediction software used in this article and the details of the SVM parameters are freely available at http://chemlab.scu.edu.cn/ predict_SubNL/index.htm and the dataset used in our study is from Shen and Chou’s work by downloading at http://chou.med.harvard.edu/ bioinf/Nuc-PLoc/Data.htm.
文摘Identification of the drug-binding residues on the surface of proteins is a vital step in drug discovery and it is important for understanding protein function. Most previous researches are based on the structural information of proteins, but the structures of most proteins are not available. So in this article, a sequence-based method was proposed by combining the support vector machine (SVM)-based ensemble learning and the improved position specific scoring matrix (PSSM). In order to take the local environment information of a drug-binding site into account, an improved PSSM profile scaled by the sliding window and smoothing window was used to improve the prediction result. In addition, a new SVM-based ensemble learning method was developed to deal with the imbalanced data classification problem that commonly exists in the binding site predictions. When performed on the dataset of 985 drug-binding residues, the method achieved a very promising prediction result with the area under the curve (AUC) of 0.9264. Furthermore, an independent dataset of 349 drug- binding residues was used to evaluate the pre- diction model and the prediction accuracy is 84.68%. These results suggest that our method is effective for predicting the drug-binding sites in proteins. The code and all datasets used in this article are freely available at http://cic.scu.edu.cn/bioinformatics/Ensem_DBS.zip.
