Guohua Huang

Application of the Shortest Path Algorithm for the Discovery of Breast Cancer-Related Genes

Breast cancer, the most prevalent cancer in women, develops from breast tissue. Its incidence has increased in recent years due to environmental risk factors. Thus, it is urgent to uncover the mechanism underlying breast cancer to design effective treatments. Identification of all breast cancer-related genes is one way to help elucidate the underlying breast cancer mechanism. In this study, a computational method was built and applied to discover new candidate breast cancer-related genes. Based on the known breast cancer-related genes retrieved from public databases, the shortest path algorithm was applied to discover new candidate genes in the protein-protein interaction network. The analysis results of the selected genes suggest that some of them are deemed breast cancer-related genes according to the most recent published literature, while others have direct or indirect associations with the initiation and development of breast cancer.

A Hybrid Computational Method for the Discovery of Novel Reproduction-Related Genes

Uncovering the molecular mechanisms underlying reproduction is of great importance to infertility treatment and to the generation of healthy offspring. In this study, we discovered novel reproduction-related genes with a hybrid computational method, integrating three different types of method, which offered new clues for further reproduction research. This method was first executed on a weighted graph, constructed based on known protein-protein interactions, to search the shortest paths connecting any two known reproduction-related genes. Genes occurring in these paths were deemed to have a special relationship with reproduction. These newly discovered genes were filtered with a randomization test. Then, the remaining genes were further selected according to their associations with known reproduction-related genes measured by protein-protein interaction score and alignment score obtained by BLAST. The in-depth analysis of the high confidence novel reproduction genes revealed hidden mechanisms of reproduction and provided guidelines for further experimental validations.

Prediction of aptamer-target interacting pairs with pseudo-amino acid composition.

Aptamers are oligonucleic acid or peptide molecules that bind to specific target molecules. As a novel and powerful class of ligands, aptamers are thought to have excellent potential for applications in the fields of biosensing, diagnostics and therapeutics. In this study, a new method for predicting aptamer-target interacting pairs was proposed by integrating features derived from both aptamers and their targets. Features of nucleotide composition and traditional amino acid composition as well as pseudo amino acid were utilized to represent aptamers and targets, respectively. The predictor was constructed based on Random Forest and the optimal features were selected by using the maximum relevance minimum redundancy (mRMR) method and the incremental feature selection (IFS) method. As a result, 81.34% accuracy and 0.4612 MCC were obtained for the training dataset, and 77.41% accuracy and 0.3717 MCC were achieved for the testing dataset. An optimal feature set of 220 features were selected, which were considered as the ones that contributed significantly to the interacting aptamer-target pair predictions. Analysis of the optimal feature set indicated several important factors in determining aptamer-target interactions. It is anticipated that our prediction method may become a useful tool for identifying aptamer-target pairs and the features selected and analyzed in this study may provide useful insights into the mechanism of interactions between aptamers and targets.

Exploring Mouse Protein Function via Multiple Approaches.

Although the number of available protein sequences is growing exponentially, functional protein annotations lag far behind. Therefore, accurate identification of protein functions remains one of the major challenges in molecular biology. In this study, we presented a novel approach to predict mouse protein functions. The approach was a sequential combination of a similarity-based approach, an interaction-based approach and a pseudo amino acid composition-based approach. The method achieved an accuracy of about 0.8450 for the 1st-order predictions in the leave-one-out and ten-fold cross-validations. For the results yielded by the leave-one-out cross-validation, although the similarity-based approach alone achieved an accuracy of 0.8756, it was unable to predict the functions of proteins with no homologues. Comparatively, the pseudo amino acid composition-based approach alone reached an accuracy of 0.6786. Although the accuracy was lower than that of the previous approach, it could predict the functions of almost all proteins, even proteins with no homologues. Therefore, the combined method balanced the advantages and disadvantages of both approaches to achieve efficient performance. Furthermore, the results yielded by the ten-fold cross-validation indicate that the combined method is still effective and stable when there are no close homologs are available. However, the accuracy of the predicted functions can only be determined according to known protein functions based on current knowledge. Many protein functions remain unknown. By exploring the functions of proteins for which the 1st-order predicted functions are wrong but the 2nd-order predicted functions are correct, the 1st-order wrongly predicted functions were shown to be closely associated with the genes encoding the proteins. The so-called wrongly predicted functions could also potentially be correct upon future experimental verification. Therefore, the accuracy of the presented method may be much higher in reality.

Prediction of Multi-Type Membrane Proteins in Human by an Integrated Approach

Membrane proteins were found to be involved in various cellular processes performing various important functions, which are mainly associated to their types. However, it is very time-consuming and expensive for traditional biophysical methods to identify membrane protein types. Although some computational tools predicting membrane protein types have been developed, most of them can only recognize one kind of type. Therefore, they are not as effective as one membrane protein can have several types at the same time. To our knowledge, few methods handling multiple types of membrane proteins were reported. In this study, we proposed an integrated approach to predict multiple types of membrane proteins by employing sequence homology and protein-protein interaction network. As a result, the prediction accuracies reached 87.65%, 81.39% and 70.79%, respectively, by the leave-one-out test on three datasets. It outperformed the nearest neighbor algorithm adopting pseudo amino acid composition. The method is anticipated to be an alternative tool for identifying membrane protein types. New metrics for evaluating performances of methods dealing with multi-label problems were also presented. The program of the method is available upon request.

Prediction of cancer drugs by chemical-chemical interactions.

Cancer, which is a leading cause of death worldwide, places a big burden on health-care system. In this study, an order-prediction model was built to predict a series of cancer drug indications based on chemical-chemical interactions. According to the confidence scores of their interactions, the order from the most likely cancer to the least one was obtained for each query drug. The 1st order prediction accuracy of the training dataset was 55.93%, evaluated by Jackknife test, while it was 55.56% and 59.09% on a validation test dataset and an independent test dataset, respectively. The proposed method outperformed a popular method based on molecular descriptors. Moreover, it was verified that some drugs were effective to the ‘wrong’ predicted indications, indicating that some ‘wrong’ drug indications were actually correct indications. Encouraged by the promising results, the method may become a useful tool to the prediction of drugs indications.

Discriminating cirRNAs from other lncRNAs using a hierarchical extreme learning machine (H-ELM) algorithm with feature selection

As non-coding RNAs, circular RNAs (cirRNAs) and long non-coding RNAs (lncRNAs) have attracted an increasing amount of attention. They have been confirmed to participate in many biological processes, including playing roles in transcriptional regulation, regulating protein-coding genes, and binding to RNA-associated proteins. Until now, the differences between these two types of non-coding RNAs have not been fully uncovered. It is still quite difficult to detect cirRNAs from other lncRNAs using simple techniques. In this study, we investigated these two types of non-coding RNAs using several computational methods. The purpose was to extract important factors that could distinguish cirRNAs from other lncRNAs and build an effective classification model to distinguish them. First, we collected cirRNAs, lncRNAs and their representations from a previous study, in which each cirRNA or lncRNA was represented by 188 features derived from its graph representation, sequence and conservation properties. Second, these features were analyzed by the minimum redundancy maximum relevance (mRMR) method. The obtained mRMR feature list, incremental feature selection method and hierarchical extreme learning machine algorithm were employed to build an optimal classification model with sensitivity of 0.703, specificity of 0.850, accuracy of 0.789 and a Matthews correlation coefficient of 0.561. Finally, we analyzed the 16 most important features. Of them, the sequences and structures of the RNA molecule were top ranking, implying they can be potential indicators of differences between cirRNAs and other lncRNAs. Meanwhile, other features of evolutionary conversation, sequence consecution were also important.

Analysis and Prediction of Myristoylation Sites Using the mRMR Method, the IFS Method and an Extreme Learning Machine Algorithm

BACKGROUND Myristoylation is an important hydrophobic post-translational modification that is covalently bound to the amino group of Gly residues on the N-terminus of proteins. The many diverse functions of myristoylation on proteins, such as membrane targeting, signal pathway regulation and apoptosis, are largely due to the lipid modification, whereas abnormal or irregular myristoylation on proteins can lead to several pathological changes in the cell. OBJECTIVE To better understand the function of myristoylated sites and to correctly identify them in protein sequences, this study conducted a novel computational investigation on identifying myristoylation sites in protein sequences. MATERIALS AND METHODS A training dataset with 196 positive and 84 negative peptide segments were obtained. Four types of features derived from the peptide segments following the myristoylation sites were used to specify myristoylatedand non-myristoylated sites. Then, feature selection methods including maximum relevance and minimum redundancy (mRMR), incremental feature selection (IFS), and a machine learning algorithm (extreme learning machine method) were adopted to extract optimal features for the algorithm to identify myristoylation sites in protein sequences, thereby building an optimal prediction model. RESULTS As a result, 41 key features were extracted and used to build an optimal prediction model. The effectiveness of the optimal prediction model was further validated by its performance on a test dataset. Furthermore, detailed analyses were also performed on the extracted 41 features to gain insight into the mechanism of myristoylation modification. CONCLUSION This study provided a new computational method for identifying myristoylation sites in protein sequences. We believe that it can be a useful tool to predict myristoylation sites from protein sequences.

An Ensemble Prognostic Model for Colorectal Cancer

Colorectal cancer can be grouped into Dukes A, B, C, and D stages based on its developments. Generally speaking, more advanced patients have poorer prognosis. To integrate progression stage prediction systems with recurrence prediction systems, we proposed an ensemble prognostic model for colorectal cancer. In this model, each patient was assigned a most possible stage and a most possible recurrence status. If a patient was predicted to be recurrence patient in advanced stage, he would be classified into high risk group. The ensemble model considered both progression stages and recurrence status. High risk patients and low risk patients predicted by the ensemble model had a significant different disease free survival (log-rank test p-value, 0.0016) and disease specific survival (log-rank test p-value, 0.0041). The ensemble model can better distinguish the high risk and low risk patients than the stage prediction model and the recurrence prediction model alone. This method could be applied to the studies of other diseases and it could significantly improve the prediction performance by ensembling heterogeneous information.

Prediction of Substrate-Enzyme-Product Interaction Based on Molecular Descriptors and Physicochemical Properties

It is important to correctly and efficiently predict the interaction of substrate-enzyme and to predict their product in metabolic pathway. In this work, a novel approach was introduced to encode substrate/product and enzyme molecules with molecular descriptors and physicochemical properties, respectively. Based on this encoding method, KNN was adopted to build the substrate-enzyme-product interaction network. After selecting the optimal features that are able to represent the main factors of substrate-enzyme-product interaction in our prediction, totally 160 features out of 290 features were attained which can be clustered into ten categories: elemental analysis, geometry, chemistry, amino acid composition, predicted secondary structure, hydrophobicity, polarizability, solvent accessibility, normalized van der Waals volume, and polarity. As a result, our predicting model achieved an MCC of 0.423 and an overall prediction accuracy of 89.1% for 10-fold cross-validation test.