Minghong Liao

Enhanced Protein Fold Prediction Method Through a Novel Feature Extraction Technique

Information of protein 3-dimensional (3D) structures plays an essential role in molecular biology, cell biology, biomedicine, and drug design. Protein fold prediction is considered as an immediate step for deciphering the protein 3D structures. Therefore, protein fold prediction is one of fundamental problems in structural bioinformatics. Recently, numerous taxonomic methods have been developed for protein fold prediction. Unfortunately, the overall prediction accuracies achieved by existing taxonomic methods are not satisfactory although much progress has been made. To address this problem, we propose a novel taxonomic method, called PFPA, which is featured by combining a novel feature set through an ensemble classifier. Particularly, the sequential evolution information from the profiles of PSI-BLAST and the local and global secondary structure information from the profiles of PSI-PRED are combined to construct a comprehensive feature set. Experimental results demonstrate that PFPA outperforms the state-of-the-art predictors. To be specific, when tested on the independent testing set of a benchmark dataset, PFPA achieves an overall accuracy of 73.6%, which is the leading accuracy ever reported. Moreover, PFPA performs well without significant performance degradation on three updated large-scale datasets, indicating the robustness and generalization of PFPA. Currently, a webserver that implements PFPA is freely available on http://121.192.180.204:8080/PFPA/Index.html.

An Improved Protein Structural Classes Prediction Method by Incorporating Both Sequence and Structure Information

Protein structural classes information is beneficial for secondary and tertiary structure prediction, protein folds prediction, and protein function analysis. Thus, predicting protein structural classes is of vital importance. In recent years, several computational methods have been developed for low-sequence-similarity (25%-40%) protein structural classes prediction. However, the reported prediction accuracies are actually not satisfactory. Aiming to further improve the prediction accuracies, we propose three different feature extraction methods and construct a comprehensive feature set that captures both sequence and structure information. By applying a random forest (RF) classifier to the feature set, we further develop a novel method for structural classes prediction. We test the proposed method on three benchmark datasets (25PDB, 640, and 1189) with low sequence similarity, and obtain the overall prediction accuracies of 93.5%, 92.6%, and 93.4%, respectively. Compared with six competing methods, the accuracies we achieved are 3.4%, 6.2%, and 8.7% higher than those achieved by the best-performing methods, showing the superiority of our method. Moreover, due to the limitation of the size of the three benchmark datasets, we further test the proposed method on three updated large-scale datasets with different sequence similarities (40%, 30%, and 25%). The proposed method achieves above 90% accuracies for all the three datasets, consistent with the accuracies on the above three benchmark datasets. Experimental results suggest our method as an effective and promising tool for structural classes prediction. Currently, a webserver that implements the proposed method is available on http://121.192.180.204:8080/RF_PSCP/Index.html.Protein structural classes information is beneficial for secondary and tertiary structure prediction, protein folds prediction, and protein function analysis. Thus, predicting protein structural classes is of vital importance. In recent years, several computational methods have been developed for low-sequence-similarity (25%-40%) protein structural classes prediction. However, the reported prediction accuracies are actually not satisfactory. Aiming to further improve the prediction accuracies, we propose three different feature extraction methods and construct a comprehensive feature set that captures both sequence and structure information. By applying a random forest (RF) classifier to the feature set, we further develop a novel method for structural classes prediction. We test the proposed method on three benchmark datasets (25PDB, 640, and 1189) with low sequence similarity, and obtain the overall prediction accuracies of 93.5%, 92.6%, and 93.4%, respectively. Compared with six competing methods, the accuracies we achieved are 3.4%, 6.2%, and 8.7% higher than those achieved by the best-performing methods, showing the superiority of our method. Moreover, due to the limitation of the size of the three benchmark datasets, we further test the proposed method on three updated large-scale datasets with different sequence similarities (40%, 30%, and 25%). The proposed method achieves above 90% accuracies for all the three datasets, consistent with the accuracies on the above three benchmark datasets. Experimental results suggest our method as an effective and promising tool for structural classes prediction. Currently, a webserver that implements the proposed method is available on http://121.192.180.204:8080/RF_PSCP/Index.html.

A Sketching Interface for Sitting Pose Design in the Virtual Environment

Character pose design is one of the most fundamental processes in computer graphics authoring. Although there are many research efforts in this field, most existing design tools consider only character body structure, rather than its interaction with the environment. This paper presents an intuitive sketching interface that allows the user to interactively place a 3D human character in a sitting position on a chair. Within our framework, the user sketches the target pose as a 2D stick figure and attaches the selected joints to the environment (e.g., the feet on the ground) with a pin tool. As reconstructing the 3D pose from a 2D stick figure is an ill-posed problem due to many possible solutions, the key idea in our paper is to reduce solution space by considering the interaction between the character and environment and adding physics constraints, such as balance and collision. Further, we formulated this reconstruction into a nonlinear optimization problem and solved it via the genetic algorithm (GA) and the quasi-Newton solver. With the GPU implementation, our system is able to generate the physically correct and visually pleasing pose at an interactive speed. The promising experimental results and user study demonstrates the efficacy of our method.

A novel machine learning method for cytokine-receptor interaction prediction

Most essential functions are associated with various protein-protein interactions, particularly the cytokine-receptor interaction. Knowledge of the heterogeneous network of cytokine- receptor interactions provides insights into various human physiological functions. However, only a few studies are focused on the computational prediction of these interactions. In this study, we propose a novel machine-learning-based method for predicting cytokine-receptor interactions. A protein sequence is first transformed by incorporating the sequence evolutional information and then formulated with the following three aspects: (1) the k-skip-n-gram model, (2) physicochemical properties, and (3) local pseudo position-specific score matrix (local PsePSSM). The random forest classifier is subsequently employed to predict potential cytokine-receptor interactions. Experimental results on a dataset of Homo sapiens show that the proposed method exhibits improved performance, with 3.4% higher overall prediction accuracy, than existing methods.

mGOF-loc

To better understand the functions of proteins, it is a critical step to predict their subcellular locations. Recently, numerous computational methods have been developed for protein subcellular localization prediction. Most of existing methods rely on the Gene Ontology (GO) information for feature representation. Although the GO information is proved to be beneficial for the improved predictive performance of the methods in prior research, the following problem is that it generates a super-high dimensional feature space, and the dimension of the feature space will get higher and higher as the number of the terms in the GO database increase. To address this issue, we propose a novel feature representation method sufficiently exploring the sequence evolutional information rather than using the GO information. Using the proposed feature representation method, we generate a comprehensive feature set of 828 features from the following three aspects: physicochemical properties, position-specific score matrix (PSSM), and the k-skip-n-gram model. By featuring a multi-label ensemble classifier with the proposed features, we further develop a novel multi-label learning method, namely mGOF-loc. Results on an updated large-scale dataset distributed with 37 subcellular locations show that mGOF-loc outperforms existing methods. Currently, a webserver that implements mGOF-loc is freely available on http://server.malab.cn/mGOF-loc/Index.html.

Bloom filter based processing algorithms for the multi-dimensional event query in wireless sensor networks

To solve the multi-dimensional event based query in wireless sensor networks, this paper proposes four bloom filter based query processing algorithms UBP, BBP, SRBP and PBP. The four algorithms proposed can be classified into two classes: two bloom filter based precise algorithms, which are UBP and BBP, and two bloom filter based approximate algorithms, which are SRBP and PBP. By using the bloom filter and introducing the inaccuracy, the communication cost involved by the query processing can be reduced. For the two precise algorithms UBP and BBP, simulation results show that UBP consumes 51% less energy than BBP on average. UBP is better than BBP on energy consumption. For energy consumption comparison between the approximate algorithms and the precise algorithm UBP, simulation results show that SRBP consumes 18% less energy than UBP on average as while as PBP consume approximately the same energy as UBP on average. For query accuracy comparison between the approximate algorithms and the precise algorithm UBP, simulation results show that the average relative error between UBP and PBP is 14% and the average relative error between UBP and SRBP is 2%. SRBP is better than PBP on energy consumption and query accuracy respectively. UBP and SRBP are two preferred bloom filter based query processing algorithms.

A clustering patch hierarchical routing protocol for wireless sensor networks

Due to the limited power of the wireless sensor networks(WSNs), the network lifetime becomes a critical index in WSNs. However, network lifetime can not depict network effectively, there exist deficiency in currently evaluative standard. To solve the problem, two new indexes (network coverage rate and effective network lifetime) are introduced for evaluating WSNs. Newly evaluative standard can depict the performance of WSNs more effectively. Meanwhile, a Clustering Patch Hierarchical Routing Protocol(CPHRP) is proposed with purposes of improving network coverage rate and effective network lifetime in WSNs. It improves network coverage rate through clustering patch, and it has hierarchical multi-path tree routing property. Whats more, the nodes are partitioned into three classes: cluster node, sense node and non-sense node, which improves the energy conservation. Simulation results show that the CPHRP can guarantee more than 90% network coverage rate within most of network lifetime with comparison to HEED when the number of inner-cluster sense nodes is above 6. With the multiple growth of network nodes, the effective network lifetime of the CPHRP rises by more than 60%. When the number of inner-cluster nodes increases in multiple of 6, the growth of its network lifecycle is more than 50% in contrast of the less than 7% of HEED.

Exploring local discriminative information from evolutionary profiles for cytokinereceptor interaction prediction

Cytokinereceptor interaction is one of the most important types of proteinprotein interactions that are widely involved in cellular regulatory processes. Knowledge of cytokinereceptor interactions facilitates to deeply understand several physiological functions. In post-genomic era of sequence explosion, there is an increasing demand for developing machine learning based computational methods for the fast and accurate cytokinereceptor interaction prediction. However, the major problem lying on existing machine learning based methods is that the overall prediction accuracy is relatively low. To improve the accuracy, a crucial step is to establish a well-defined feature representation algorithm. Motivated on this perspective, we propose a novel feature representation method by integrating local information embedded in evolutionary profiles with the Pse-PSSM and AAC-PSSM-AC feature models. We further develop an improved prediction method, namely CRI-Pred, based on the proposed feature set using the Random Forest classifier. Experimental results evaluated with the jackknife test show that the CRI-Pred predictor outperforms the state-of-the-art methods, 5.1% higher in terms of the overall accuracy. This indicates the effectiveness and superiority of CRI-Pred. A webserver that implements CRI-Pred is now freely available at http://server.malab.cn/CRIPred/Index.html to the public to use in practical applications.

Interior structure transfer via harmonic 1-forms

As a natural extension of surface parameterizaiton, volumetric parameterization is becoming more and more popular and exhibiting great advantages in several applications such as medical image analysis, hexahedral meshing etc. This paper presents an efficient volume parameterization algorithm based on harmonic 1-form. Our new algorithm computes three harmonic 1-forms, which can be treated as three vector fields, such that both the divergence and circulation of them are zero. By integrating the three harmonic 1-forms over the entire volumes, we can bijectively map the volume to a cuboid domain. We demonstrate the power of the technique by introducing a new application, to transfer the interior structure during the morphing of two given shapes.

An iterative algorithm to process the top-k query for the wireless sensor networks

The top–k query is an important type of query for the wireless sensor network. In this paper, we present an iterative algorithm to process the top–k query, which is a distributed algorithm combining the in–network aggregation and the trace back techniques. By using the in–network aggregation technique, the iterative algorithm calculates the current maximum value in the network. By using the trace back technology, the maximum value just calculated is removed from the network. The two steps are repeated k times. As the current maximum value in the network is selected per round, the answer to a top–k query can be obtained after k repetitions. Experimental results show that the iterative algorithm can reduce the number of messages transmitted during the procedure of the top–k query processing.