Entity

Time filter

Source Type


Wang K.,Suzhou University of Science and Technology | Wu H.,Suzhou University of Science and Technology | Wu H.,Jiangsu Provincial Key Laboratory for Information Processing Technologies | Lu W.,Suzhou University of Science and Technology | And 4 more authors.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2015

Transmembrane proteins play an important role in cellular energy production, signal transmission, metabolism. Existing machine learning methods are difficult to model the global correlation of the membrane protein sequence, and they also can not improve the quality of the model from sophisticated sequence features. To address these problems, in this paper we proposed a novel method by a feedback conditional random fields (FCRF) to predict helix boundaries of α-helix transmembrane protein. A feedback mechanism was introduced into multi-level conditional random fields. The results of lower level model were used to calculate new feedback features to enhance the ability of basic conditional random fields. One wide-used dataset DB1 was used to validate the performance of the method. The method achieved 95% on helix location accuracy. Compared with the other predictors, FCRF ranks first on the accuracy of helix location. © Springer International Publishing Switzerland 2015. Source


Huang Y.,Soochow University of China | Lu Q.,Soochow University of China | Lu Q.,Jiangsu Provincial Key Laboratory for Information Processing Technologies | Zeng J.,Soochow University of China
Proceedings - 2013 IEEE International Conference on Bioinformatics and Biomedicine, IEEE BIBM 2013 | Year: 2013

We present an algorithm EcSymDock which optimized SymDock using Evolution couplings for better performance. SymDock is a protocol of Rosetta used to predict protein oligomer structure. Evolution Coupling (EC) is an optimized mutual information calculated by specific protein family MSA, it reveals certain contacts between residues. Today more and more researchers pay close attention to how to use ECs as constraints during protein structure prediction. We combined EC information with SymDock protocol and present the EcSymDock method to add EC as constraints into SymDock protocol. Testing on a list of dimers shows the method recovers the native dimeric configuration with better accuracy than SymDock and produces many decoys near the native structure in top 400 lowest energy. © 2013 IEEE. Source


Yang P.,Soochow University of China | Lu Q.,Soochow University of China | Lu Q.,Jiangsu Provincial Key Laboratory for Information Processing Technologies | Yang L.,Soochow University of China | Wu J.,Soochow University of China
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2010

Modeling protein loops is important for understanding characteristics and functions for protein, but remains an unsolved problem of computational biology. By employing a general Bayesian network, this paper constructs a fully probabilistic continuous model of protein loops, refered to as LoopBN. Direct affection between amino acids and backbone torsion angles can be learned under the framework of LoopBN. The continuous torsion angle pair of the loops can be captured by bivariate von Mises distribution. Empirical tests are conducted to evaluate the performance of LoopBN based on 8 free modeling targets of CASP8. Experimental results show that LoopBN not only performs better than the state-of-the-art modeling method on the quality of loop sample set, but also helps de novo prediction of protein structure by providing better sample set for loop refinement. © 2010 Springer-Verlag. Source


Li L.,Soochow University of China | Xu W.,Wujiang Rural Commercial Bank | Lu Q.,Soochow University of China | Lu Q.,Jiangsu Provincial Key Laboratory for Information Processing Technologies
Journal of Molecular Modeling | Year: 2015

Computational protein-ligand docking is of great importance in drug discovery and design. Conformational changes greatly affect the results of protein-ligand docking, especially when water molecules take part in mediating protein ligand interactions or when large conformational changes are observed in the receptor backbone interface. We have developed an improved protocol, SWRosettaLigand, based on the RosettaLigand protocol. This approach incorporates the flexibility of interfacial water molecules and modeling of the interface of the receptor into the original RosettaLigand. In a coarse sampling step, SWRosettaLigand pre-optimizes the initial position of the water molecules, docks the ligand to the receptor with explicit water molecules, and minimizes the predicted structure with water molecules. The receptor backbone interface is treated as a loop and perturbed and refined by kinematic closure, or cyclic coordinate descent algorithm, with the presence of the ligand. In two cross-docking test sets, it was identified that for 8 out of 14, and 16 out of 22, test instances, the top-ranked structures by SWRosettaLigand achieved better accuracy than other protocols. © 2015, Springer-Verlag Berlin Heidelberg. Source


Lu W.,Suzhou University of Science and Technology | Fu B.,Suzhou University of Science and Technology | Wu H.,Suzhou University of Science and Technology | Wu H.,Jiangsu Provincial Key Laboratory for Information Processing Technologies | And 4 more authors.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2015

Transmembrane proteins are important for cell transport biology and in the treatment of disease. Understanding the helix count and locations in transmembrane proteins is a key problem for structural and functional analyses. But there is a lack of high resolution three-dimensional structures. In this study, we propose a method based on conditional random fields for predicting the helix count and locations, CRF-TM, which reflects long-range correlations in the full-length sequence as joint probabilities. Two datasets are employed in the performance validation. Our results show that CRF-TM can rank the first group better compared with other widely used TM predictors. The results obtained by CRF-TM are also used to predict the three-dimensional structures of GPCRs, which is crucial drug targets and also a subclass of transmembrane with seven spanning α-helices. © Springer International Publishing Switzerland 2015. Source

Discover hidden collaborations