Search

Breast cancer is a heterogeneous disease with many clinically distinguishable molecular subtypes each corresponding to a cluster of patients. Identification of prognostic and heterogeneous biomarkers for breast cancer is to detect cluster-specific gene biomarkers which can be used for accurate survival prediction of breast cancer outcomes. In this study, we proposed a FUsion Network-based method (FUNMarker) to identify prognostic and heterogeneous breast cancer biomarkers by considering the heterogeneity of patient samples and biological information from multiple sources. To reduce the affect of heterogeneity of patients, samples were first clustered using the K-means algorithm based on the principal components of gene expression. For each cluster, to comprehensively evaluate the influence of genes on breast cancer, genes were weighted from three aspects: biological function, prognostic ability and correlation with known disease genes. Then they were ranked via a label propagation model on a fusion network that combined physical protein interactions from seven types of networks and thus could reduce the impact of incompleteness of interactome. We compared FUNMarker with three state-of-the-art methods and the results showed that biomarkers identified by FUNMarker were biological interpretable and had stronger discriminative power than the existing methods in differentiating patients with different prognostic outcomes.

Raw sequencing reads of miRNAs contain machine-made substitution errors, or even insertions and deletions (indels). Although the error rate can be low at 0.1%, precise rectification of these errors is critically important because isoform variation analysis at single-base resolution such as novel isomiR discovery, editing events understanding, differential expression analysis, or tissue-specific isoform identification is very sensitive to base positions and copy counts of the reads. Existing error correction methods do not work for miRNA sequencing data attributed to miRNAs' length and per-read-coverage properties distinct from DNA or mRNA sequencing reads. We present a novel lattice structure combining kmers, (k - 1)mers and (k + 1)mers to address this problem. The method is particularly effective for the correction of indel errors. Extensive tests on datasets having known ground truth of errors demonstrate that the method is able to remove almost all of the errors, without introducing any new error, to improve the data quality from every-50-reads containing one error to every-1300-reads containing one error. Studies on experimental miRNA sequencing datasets show that the errors are often rectified at the 5' ends and the seed regions of the reads, and that there are remarkable changes after the correction in miRNA isoform abundance, volume of singleton reads, overall entropy, isomiR families, tissue-specific miRNAs, and rare-miRNA quantities., (© The Author(s) 2021. Published by Oxford University Press on behalf of Nucleic Acids Research.)

Biological targets are most commonly proteins such as enzymes, ion channels, and receptors. They are anything within a living organism to bind with some other entities (like an endogenous ligand or a drug), resulting in change in their behaviors or functions. Exploring potential drug-target interactions (DTIs) are crucial for drug discovery and effective drug development. Computational methods were widely applied in drug-target interactions, since experimental methods are extremely time-consuming and resource-intensive. In this paper, we proposed a novel deep learning-based prediction system, with a new negative instance generation, to identify DTIs. As a result, our method achieved an accuracy of 0.9800 on our created dataset. Another dataset derived from DrugBank was used to further assess the generalization of the model, which yielded a good performance with accuracy of 0.8814 and AUC value of 0.9527 on the dataset. The outcome of our experimental results indicated that the proposed method, involving the credible negative generation, can be employed to discriminate the interactions between drugs and targets. Website: http://www.dlearningapp.com/web/DrugCNN.htm.

Background: The early diagnosis of lung cancer has been a critical problem in clinical practice for a long time and identifying differentially expressed gene as disease marker is a promising solution. However, the most existing gene differential expression analysis (DEA) methods have two main drawbacks: First, these methods are based on fixed statistical hypotheses and not always effective; Second, these methods can not identify a certain expression level boundary when there is no obvious expression level gap between control and experiment groups., Methods: This paper proposed a novel approach to identify marker genes and gene expression level boundary for lung cancer. By calculating a kernel maximum mean discrepancy, our method can evaluate the expression differences between normal, normal adjacent to tumor (NAT) and tumor samples. For the potential marker genes, the expression level boundaries among different groups are defined with the information entropy method., Results: Compared with two conventional methods t-test and fold change, the top average ranked genes selected by our method can achieve better performance under all metrics in the 10-fold cross-validation. Then GO and KEGG enrichment analysis are conducted to explore the biological function of the top 100 ranked genes. At last, we choose the top 10 average ranked genes as lung cancer markers and their expression boundaries are calculated and reported., Conclusion: The proposed approach is effective to identify gene markers for lung cancer diagnosis. It is not only more accurate than conventional DEA methods but also provides a reliable method to identify the gene expression level boundaries.

Background: In silico prediction of potential drug side-effects is of crucial importance for drug development, since wet experimental identification of drug side-effects is expensive and time-consuming. Existing computational methods mainly focus on leveraging validated drug side-effect relations for the prediction. The performance is severely impeded by the lack of reliable negative training data. Thus, a method to select reliable negative samples becomes vital in the performance improvement., Methods: Most of the existing computational prediction methods are essentially based on the assumption that similar drugs are inclined to share the same side-effects, which has given rise to remarkable performance. It is also rational to assume an inverse proposition that dissimilar drugs are less likely to share the same side-effects. Based on this inverse similarity hypothesis, we proposed a novel method to select highly-reliable negative samples for side-effect prediction. The first step of our method is to build a drug similarity integration framework to measure the similarity between drugs from different perspectives. This step integrates drug chemical structures, drug target proteins, drug substituents, and drug therapeutic information as features into a unified framework. Then, a similarity score between each candidate negative drug and validated positive drugs is calculated using the similarity integration framework. Those candidate negative drugs with lower similarity scores are preferentially selected as negative samples. Finally, both the validated positive drugs and the selected highly-reliable negative samples are used for predictions., Results: The performance of the proposed method was evaluated on simulative side-effect prediction of 917 DrugBank drugs, comparing with four machine-learning algorithms. Extensive experiments show that the drug similarity integration framework has superior capability in capturing drug features, achieving much better performance than those based on a single type of drug property. Besides, the four machine-learning algorithms achieved significant improvement in macro-averaging F1-score (e.g., SVM from 0.655 to 0.898), macro-averaging precision (e.g., RBF from 0.592 to 0.828) and macro-averaging recall (e.g., KNN from 0.651 to 0.772) complimentarily attributed to the highly-reliable negative samples selected by the proposed method., Conclusions: The results suggest that the inverse similarity hypothesis and the integration of different drug properties are valuable for side-effect prediction. The selection of highly-reliable negative samples can also make significant contributions to the performance improvement.

Background: Hot spot residues are functional sites in protein interaction interfaces. The identification of hot spot residues is time-consuming and laborious using experimental methods. In order to address the issue, many computational methods have been developed to predict hot spot residues. Moreover, most prediction methods are based on structural features, sequence characteristics, and/or other protein features., Results: This paper proposed an ensemble learning method to predict hot spot residues that only uses sequence features and the relative accessible surface area of amino acid sequences. In this work, a novel feature selection technique was developed, an auto-correlation function combined with a sliding window technique was applied to obtain the characteristics of amino acid residues in protein sequence, and an ensemble classifier with SVM and KNN base classifiers was built to achieve the best classification performance., Conclusion: The experimental results showed that our model yields the highest F1 score of 0.92 and an MCC value of 0.87 on ASEdb dataset. Compared with other machine learning methods, our model achieves a big improvement in hot spot prediction., Availability: http://deeplearner.ahu.edu.cn/web/HotspotEL.htm .

Background: Distinction between pre-microRNAs (precursor microRNAs) and length-similar pseudo pre-microRNAs can reveal more about the regulatory mechanism of RNA biological processes. Machine learning techniques have been widely applied to deal with this challenging problem. However, most of them mainly focus on secondary structure information of pre-microRNAs, while ignoring sequence-order information and sequence evolution information., Results: We use new features for the machine learning algorithms to improve the classification performance by characterizing both sequence order evolution information and secondary structure graphs. We developed three steps to extract these features of pre-microRNAs. We first extract features from PSI-BLAST profiles and Hilbert-Huang transforms, which contain rich sequence evolution information and sequence-order information respectively. We then obtain properties of small molecular networks of pre-microRNAs, which contain refined secondary structure information. These structural features are carefully generated so that they can depict both global and local characteristics of pre-microRNAs. In total, our feature space covers 591 features. The maximum relevance and minimum redundancy (mRMR) feature selection method is adopted before support vector machine (SVM) is applied as our classifier. The constructed classification model is named MicroRNA -NHPred. The performance of MicroRNA -NHPred is high and stable, which is better than that of those state-of-the-art methods, achieving an accuracy of up to 94.83% on same benchmark datasets., Conclusions: The high prediction accuracy achieved by our proposed method is attributed to the design of a comprehensive feature set on the sequences and secondary structures, which are capable of characterizing the sequence evolution information and sequence-order information, and global and local information of pre-microRNAs secondary structures. MicroRNA -NHPred is a valuable method for pre-microRNAs identification. The source codes of our method can be downloaded from https://github.com/myl446/MicroRNA-NHPred .

Background: Protein secondary structure can be regarded as an information bridge that links the primary sequence and tertiary structure. Accurate 8-state secondary structure prediction can significantly give more precise and high resolution on structure-based properties analysis., Results: We present a novel deep learning architecture which exploits an integrative synergy of prediction by a convolutional neural network, residual network, and bidirectional recurrent neural network to improve the performance of protein secondary structure prediction. A local block comprised of convolutional filters and original input is designed for capturing local sequence features. The subsequent bidirectional recurrent neural network consisting of gated recurrent units can capture global context features. Furthermore, the residual network can improve the information flow between the hidden layers and the cascaded recurrent neural network. Our proposed deep network achieved 71.4% accuracy on the benchmark CB513 dataset for the 8-state prediction; and the ensemble learning by our model achieved 74% accuracy. Our model generalization capability is also evaluated on other three independent datasets CASP10, CASP11 and CASP12 for both 8- and 3-state prediction. These prediction performances are superior to the state-of-the-art methods., Conclusion: Our experiment demonstrates that it is a valuable method for predicting protein secondary structure, and capturing local and global features concurrently is very useful in deep learning.

Background: N 6 -methyladenosine (m 6 A) is an important epigenetic modification which plays various roles in mRNA metabolism and embryogenesis directly related to human diseases. To identify m 6 A in a large scale, machine learning methods have been developed to make predictions on m 6 A sites. However, there are two main drawbacks of these methods. The first is the inadequate learning of the imbalanced m 6 A samples which are much less than the non-m 6 A samples, by their balanced learning approaches. Second, the features used by these methods are not outstanding to represent m 6 A sequence characteristics., Results: We propose to use cost-sensitive learning ideas to resolve the imbalance data issues in the human mRNA m 6 A prediction problem. This cost-sensitive approach applies to the entire imbalanced dataset, without random equal-size selection of negative samples, for an adequate learning. Along with site location and entropy features, top-ranked positions with the highest single nucleotide polymorphism specificity in the window sequences are taken as new features in our imbalance learning. On an independent dataset, our overall prediction performance is much superior to the existing predictors. Our method shows stronger robustness against the imbalance changes in the tests on 9 datasets whose imbalance ratios range from 1:1 to 9:1. Our method also outperforms the existing predictors on 1226 individual transcripts. It is found that the new types of features are indeed of high significance in the m 6 A prediction. The case studies on gene c-Jun and CBFB demonstrate the detailed prediction capacity to improve the prediction performance., Conclusion: The proposed cost-sensitive model and the new features are useful in human mRNA m 6 A prediction. Our method achieves better correctness and robustness than the existing predictors in independent test and case studies. The results suggest that imbalance learning is promising to improve the performance of m 6 A prediction.

Motivation: Antigens that contain overlapping epitopes have been occasionally reported. As current algorithms mainly take a one-antigen-one-epitope approach to the prediction of epitopes, they are not capable of detecting these multiple and overlapping epitopes accurately, or even those multiple and separated epitopes existing in some other antigens., Results: We introduce a novel subgraph clustering algorithm for more accurate detection of epitopes. This algorithm takes graph partitions as seeds, and expands the seeds to merge overlapping subgraphs based on the term frequency-inverse document frequency (TF-IDF) featured similarity. Then, the merged subgraphs are each classified as an epitope or non-epitope. Tests of our algorithm were conducted on three newly collected datasets of antigens. In the first dataset, each antigen contains only a single epitope; in the second, each antigen contains only multiple and separated epitopes; and in the third, each antigen contains overlapping epitopes. The prediction performance of our algorithm is significantly better than the state-of-art methods. The lifts of the averaged f-scores on top of the best existing methods are 60, 75 and 22% for the single epitope detection, the multiple and separated epitopes detection, and the overlapping epitopes detection, respectively., Availability and Implementation: The source code is available at github.com/lzhlab/glep/., Supplementary Information: Supplementary data are available at Bioinformatics online.

Antimicrobial peptides (AMPs) are promising candidates in the fight against multidrug-resistant pathogens owing to AMPs' broad range of activities and low toxicity. Nonetheless, identification of AMPs through wet-lab experiments is still expensive and time consuming. Here, we propose an accurate computational method for AMP prediction by the random forest algorithm. The prediction model is based on the distribution patterns of amino acid properties along the sequence. Using our collection of large and diverse sets of AMP and non-AMP data (3268 and 166791 sequences, respectively), we evaluated 19 random forest classifiers with different positive:negative data ratios by 10-fold cross-validation. Our optimal model, AmPEP with the 1:3 data ratio, showed high accuracy (96%), Matthew's correlation coefficient (MCC) of 0.9, area under the receiver operating characteristic curve (AUC-ROC) of 0.99, and the Kappa statistic of 0.9. Descriptor analysis of AMP/non-AMP distributions by means of Pearson correlation coefficients revealed that reduced feature sets (from a full-featured set of 105 to a minimal-feature set of 23) can result in comparable performance in all respects except for some reductions in precision. Furthermore, AmPEP outperformed existing methods in terms of accuracy, MCC, and AUC-ROC when tested on benchmark datasets.

Frequently recurring RNA structural motifs play important roles in RNA folding process and interaction with other molecules. Traditional index-based and shape-based schemas are useful in modeling RNA secondary structures but ignore the structural discrepancy of individual RNA family member. Further, the in-depth analysis of underlying substructure pattern is insufficient due to varied and unnormalized substructure data. This prevents us from understanding RNAs functions and their inherent synergistic regulation networks. This article thus proposes a novel labeled graph-based algorithm RnaGraph to uncover frequently RNA substructure patterns. Attribute data and graph data are combined to characterize diverse substructures and their correlations, respectively. Further, a top-k graph pattern mining algorithm is developed to extract interesting substructure motifs by integrating frequency and similarity. The experimental results show that our methods assist in not only modelling complex RNA secondary structures but also identifying hidden but interesting RNA substructure patterns.

The carbonylation is found as an irreversible post-translational modification and considered a biomarker of oxidative stress. It plays major role not only in orchestrating various biological processes but also associated with some diseases such as Alzheimer's disease, diabetes, and Parkinson's disease. However, since the experimental technologies are costly and time-consuming to detect the carbonylation sites in proteins, an accurate computational method for predicting carbonylation sites is an urgent issue which can be useful for drug development. In this study, a novel computational tool termed predCar-Site has been developed to predict protein carbonylation sites by (1) incorporating the sequence-coupled information into the general pseudo amino acid composition, (2) balancing the effect of skewed training dataset by Different Error Costs method, and (3) constructing a predictor using support vector machine as classifier. This predCar-Site predictor achieves an average AUC (area under curve) score of 0.9959, 0.9999, 1, and 0.9997 in predicting the carbonylation sites of K, P, R, and T, respectively. All of the experimental results along with AUC are found from the average of 5 complete runs of the 10-fold cross-validation and those results indicate significantly better performance than existing predictors. A user-friendly web server of predCar-Site is available at http://research.ru.ac.bd/predCar-Site/., (Crown Copyright © 2017. Published by Elsevier Inc. All rights reserved.)

Background: MicroRNAs always function cooperatively in their regulation of gene expression. Dysfunctions of these co-functional microRNAs can play significant roles in disease development. We are interested in those multi-disease associated co-functional microRNAs that regulate their common dysfunctional target genes cooperatively in the development of multiple diseases. The research is potentially useful for human disease studies at the transcriptional level and for the study of multi-purpose microRNA therapeutics., Methods and Results: We designed a computational method to detect multi-disease associated co-functional microRNA pairs and conducted cross disease analysis on a reconstructed disease-gene-microRNA (DGR) tripartite network. The construction of the DGR tripartite network is by the integration of newly predicted disease-microRNA associations with those relationships of diseases, microRNAs and genes maintained by existing databases. The prediction method uses a set of reliable negative samples of disease-microRNA association and a pre-computed kernel matrix instead of kernel functions. From this reconstructed DGR tripartite network, multi-disease associated co-functional microRNA pairs are detected together with their common dysfunctional target genes and ranked by a novel scoring method. We also conducted proof-of-concept case studies on cancer-related co-functional microRNA pairs as well as on non-cancer disease-related microRNA pairs., Conclusions: With the prioritization of the co-functional microRNAs that relate to a series of diseases, we found that the co-function phenomenon is not unusual. We also confirmed that the regulation of the microRNAs for the development of cancers is more complex and have more unique properties than those of non-cancer diseases.

Background: Regulation mechanisms between miRNAs and genes are complicated. To accomplish a biological function, a miRNA may regulate multiple target genes, and similarly a target gene may be regulated by multiple miRNAs. Wet-lab knowledge of co-regulating miRNAs is limited. This work introduces a computational method to group miRNAs of similar functions to identify co-regulating miRNAsfrom a similarity matrix of miRNAs., Results: We define a novel information content of gene ontology (GO) to measure similarity between two sets of GO graphs corresponding to the two sets of target genes of two miRNAs. This between-graph similarity is then transferred as a functional similarity between the two miRNAs. Our definition of the information content is based on the size of a GO term's descendants, but adjusted by a weight derived from its depth level and the GO relationships at its path to the root node or to the most informative common ancestor (MICA). Further, a self-tuning technique and the eigenvalues of the normalized Laplacian matrix are applied to determine the optimal parameters for the spectral clustering of the similarity matrix of the miRNAs., Conclusions: Experimental results demonstrate that our method has better clustering performance than the existing edge-based, node-based or hybrid methods. Our method has also demonstrated a novel usefulness for the function annotation of new miRNAs, as reported in the detailed case studies.

Protein S-sulfenylation (SOH) is a type of post-translational modification through the oxidation of cysteine thiols to sulfenic acids. It acts as a redox switch to modulate versatile cellular processes and plays important roles in signal transduction, protein folding and enzymatic catalysis. Reversible SOH is also a key component for maintaining redox homeostasis and has been implicated in a variety of human diseases, such as cancer, diabetes, and atherosclerosis, due to redox imbalance. Despite its significance, the in situ trapping of the entire 'sulfenome' remains a major challenge. Yang et al. have recently experimentally identified about 1000 SOH sites, providing an enriched benchmark SOH dataset. In this work, we developed a new ensemble learning tool SOHPRED for identifying protein SOH sites based on the compositions of enriched amino acids and the physicochemical properties of residues surrounding SOH sites. SOHPRED was built based on four complementary predictors, i.e. a naive Bayesian predictor, a random forest predictor and two support vector machine predictors, whose training features are, respectively, amino acid occurrences, physicochemical properties, frequencies of k-spaced amino acid pairs and sequence profiles. Benchmarking experiments on the 5-fold cross validation and independent tests show that SOHPRED achieved AUC values of 0.784 and 0.799, respectively, which outperforms several previously developed tools. As a real application of SOHPRED, we predicted potential SOH sites for 193 S-sulfenylated substrates, which had been experimentally detected through a global sulfenome profiling in living cells, though the actual SOH sites were not determined. The web server of SOHPRED has been made publicly available at for the wider research community. The source codes and the benchmark datasets can be downloaded from the website.

A binding hot spot is a small area at a protein-protein interface that can make significant contribution to binding free energy. This work investigates the substantial contribution made by some special co-occurring atomic contacts at a binding hot spot. A co-occurring atomic contact is a pair of atomic contacts that are close to each other with no more than three covalent-bond steps. We found that two kinds of co-occurring atomic contacts can play an important part in the accurate prediction of binding hot spot residues. One is the co-occurrence of two nearby hydrogen bonds. For example, mutations of any residue in a hydrogen bond network consisting of multiple co-occurring hydrogen bonds could disrupt the interaction considerably. The other kind of co-occurring atomic contact is the co-occurrence of a hydrophobic carbon contact and a contact between a hydrophobic carbon atom and a π ring. In fact, this co-occurrence signifies the collective effect of hydrophobic contacts. We also found that the B-factor measurements of several specific groups of amino acids are useful for the prediction of hot spots. Taking the B-factor, individual atomic contacts and the co-occurring contacts as features, we developed a new prediction method and thoroughly assessed its performance via cross-validation and independent dataset test. The results show that our method achieves higher prediction performance than well-known methods such as Robetta, FoldX and Hotpoint. We conclude that these contact descriptors, in particular the novel co-occurring atomic contacts, can be used to facilitate accurate and interpretable characterization of protein binding hot spots.

Random walk on heterogeneous networks is a recently emerging approach to effective disease gene prioritization. Laplacian normalization is a technique capable of normalizing the weight of edges in a network. We use this technique to normalize the gene matrix and the phenotype matrix before the construction of the heterogeneous network, and also use this idea to define the transition matrices of the heterogeneous network. Our method has remarkably better performance than the existing methods for recovering known gene-phenotype relationships. The Shannon information entropy of the distribution of the transition probabilities in our networks is found to be smaller than the networks constructed by the existing methods, implying that a higher number of top-ranked genes can be verified as disease genes. In fact, the most probable gene-phenotype relationships ranked within top 3 or top 5 in our gene lists can be confirmed by the OMIM database for many cases. Our algorithms have shown remarkably superior performance over the state-of-the-art algorithms for recovering gene-phenotype relationships. All Matlab codes can be available upon email request., (Copyright © 2015 Elsevier Ltd. All rights reserved.)

Prokaryotic proteins are regulated by pupylation, a type of post-translational modification that contributes to cellular function in bacterial organisms. In pupylation process, the prokaryotic ubiquitin-like protein (Pup) tagging is functionally analogous to ubiquitination in order to tag target proteins for proteasomal degradation. To date, several experimental methods have been developed to identify pupylated proteins and their pupylation sites, but these experimental methods are generally laborious and costly. Therefore, computational methods that can accurately predict potential pupylation sites based on protein sequence information are highly desirable. In this paper, a novel predictor termed as pbPUP has been developed for accurate prediction of pupylation sites. In particular, a sophisticated sequence encoding scheme [i.e. the profile-based composition of k-spaced amino acid pairs (pbCKSAAP)] is used to represent the sequence patterns and evolutionary information of the sequence fragments surrounding pupylation sites. Then, a Support Vector Machine (SVM) classifier is trained using the pbCKSAAP encoding scheme. The final pbPUP predictor achieves an AUC value of 0.849 in 10-fold cross-validation tests and outperforms other existing predictors on a comprehensive independent test dataset. The proposed method is anticipated to be a helpful computational resource for the prediction of pupylation sites. The web server and curated datasets in this study are freely available at http://protein.cau.edu.cn/pbPUP/.

Protein-protein interfaces defined through atomic contact or solvent accessibility change are widely adopted in structural biology studies. But, these definitions cannot precisely capture energetically important regions at protein interfaces. The burial depth of an atom in a protein is related to the atom's energy. This work investigates how closely the change in burial level of an atom/residue upon complexation is related to the binding. Burial level change is different from burial level itself. An atom deeply buried in a monomer with a high burial level may not change its burial level after an interaction and it may have little burial level change. We hypothesize that an interface is a region of residues all undergoing burial level changes after interaction. By this definition, an interface can be decomposed into an onion-like structure according to the burial level change extent. We found that our defined interfaces cover energetically important residues more precisely, and that the binding free energy of an interface is distributed progressively from the outermost layer to the core. These observations are used to predict binding hot spots. Our approach's F-measure performance on a benchmark dataset of alanine mutagenesis residues is much superior or similar to those by complicated energy modeling or machine learning approaches.

Motivation: Deciphering gene interaction networks (GINs) from time-course gene expression (TCGx) data is highly valuable to understand gene behaviors (e.g., activation, inhibition, time-lagged causality) at the system level. Existing methods usually use a global or local proximity measure to infer GINs from a single dataset. As the noise contained in a single data set is hardly self-resolved, the results are sometimes not reliable. Also, these proximity measurements cannot handle the co-existence of the various in vivo positive, negative and time-lagged gene interactions., Methods and Results: We propose to infer reliable GINs from multiple TCGx datasets using a novel conserved subsequential pattern of gene expression. A subsequential pattern is a maximal subset of genes sharing positive, negative or time-lagged correlations of one expression template on their own subsets of time points. Based on these patterns, a GIN can be built from each of the datasets. It is assumed that reliable gene interactions would be detected repeatedly. We thus use conserved gene pairs from the individual GINs of the multiple TCGx datasets to construct a reliable GIN for a species. We apply our method on six TCGx datasets related to yeast cell cycle, and validate the reliable GINs using protein interaction networks, biopathways and transcription factor-gene regulations. We also compare the reliable GINs with those GINs reconstructed by a global proximity measure Pearson correlation coefficient method from single datasets. It has been demonstrated that our reliable GINs achieve much better prediction performance especially with much higher precision. The functional enrichment analysis also suggests that gene sets in a reliable GIN are more functionally significant. Our method is especially useful to decipher GINs from multiple TCGx datasets related to less studied organisms where little knowledge is available except gene expression data.

Background: Intensive research based on the inverse expression relationship has been undertaken to discover the miRNA-mRNA regulatory modules involved in the infection of Hepatitis C virus (HCV), the leading cause of chronic liver diseases. However, biological studies in other fields have found that inverse expression relationship is not the only regulatory relationship between miRNAs and their targets, and some miRNAs can positively regulate a mRNA by binding at the 5' UTR of the mRNA., Results: This work focuses on the detection of both inverse and positive regulatory relationships from a paired miRNA and mRNA expression data set of HCV patients through a 'change-to-change' method which can derive connected discriminatory rules. Our study uncovered many novel miRNA-mRNA regulatory modules. In particular, it was revealed that GFRA2 is positively regulated by miR-557, miR-765 and miR-17-3p that probably bind at different locations of the 5' UTR of this mRNA. The expression relationship between GFRA2 and any of these three miRNAs has not been studied before, although separate research for this gene and these miRNAs have all drawn conclusions linked to hepatocellular carcinoma. This suggests that the binding of mRNA GFRA2 with miR-557, miR-765, or miR-17-3p, or their combinations, is worthy of further investigation by experimentation. We also report another mRNA QKI which has a strong inverse expression relationship with miR-129 and miR-493-3p which may bind at the 3' UTR of QKI with a perfect sequence match. Furthermore, the interaction between hsa-miR-129-5p (previous ID: hsa-miR-129) and QKI is supported with CLIP-Seq data from starBase. Our method can be easily extended for the expression data analysis of other diseases., Conclusion: Our rule discovery method is useful for integrating binding information and expression profile for identifying HCV miRNA-mRNA regulatory modules and can be applied to the study of the expression profiles of other complex human diseases.

Background: Binding free energy and binding hot spots at protein-protein interfaces are two important research areas for understanding protein interactions. Computational methods have been developed previously for accurate prediction of binding free energy change upon mutation for interfacial residues. However, a large number of interrupted and unimportant atomic contacts are used in the training phase which caused accuracy loss., Results: This work proposes a new method, βACVASA, to predict the change of binding free energy after alanine mutations. βACVASA integrates accessible surface area (ASA) and our newly defined β contacts together into an atomic contact vector (ACV). A β contact between two atoms is a direct contact without being interrupted by any other atom between them. A β contact's potential contribution to protein binding is also supposed to be inversely proportional to its ASA to follow the water exclusion hypothesis of binding hot spots. Tested on a dataset of 396 alanine mutations, our method is found to be superior in classification performance to many other methods, including Robetta, FoldX, HotPOINT, an ACV method of β contacts without ASA integration, and ACVASA methods (similar to βACVASA but based on distance-cutoff contacts). Based on our data analysis and results, we can draw conclusions that: (i) our method is powerful in the prediction of binding free energy change after alanine mutation; (ii) β contacts are better than distance-cutoff contacts for modeling the well-organized protein-binding interfaces; (iii) β contacts usually are only a small fraction number of the distance-based contacts; and (iv) water exclusion is a necessary condition for a residue to become a binding hot spot., Conclusions: βACVASA is designed using the advantages of both β contacts and water exclusion. It is an excellent tool to predict binding free energy changes and binding hot spots after alanine mutation.

Many studies are aimed at identifying dense clusters/subgraphs from protein-protein interaction (PPI) networks for protein function prediction. However, the prediction performance based on the dense clusters is actually worse than a simple guilt-by-association method using neighbor counting ideas. This indicates that the local topological structures and properties of PPI networks are still open to new theoretical investigation and empirical exploration. We introduce a novel topological structure called k-partite cliques of protein interactions-a functionally coherent but not-necessarily dense subgraph topology in PPI networks-to study PPI networks. A k-partite protein clique is a maximal k-partite clique comprising two or more nonoverlapping protein subsets between any two of which full interactions are exhibited. In the detection of PPI's maximal k-partite cliques, we propose to transform PPI networks into induced K-partite graphs where edges exist only between the partites. Then, we present a maximal k-partite clique mining (MaCMik) algorithm to enumerate maximal k-partite cliques from K-partite graphs. Our MaCMik algorithm is then applied to a yeast PPI network. We observed interesting and unusually high functional coherence in k-partite protein cliques-the majority of the proteins in k-partite protein cliques, especially those in the same partites, share the same functions, although k-partite protein cliques are not restricted to be dense compared with dense subgraph patterns or (quasi-)cliques. The idea of k-partite protein cliques provides a novel approach of characterizing PPI networks, and so it will help function prediction for unknown proteins., (© 2013 Elsevier Ltd. All rights reserved.)

Mining Protein-Protein Interactions (PPIs) from the fast-growing biomedical literature resources has been proven as an effective approach for the identification of biological regulatory networks. This paper presents a novel method based on the idea of Interaction Relation Ontology (IRO), which specifies and organises words of various proteins interaction relationships. Our method is a two-stage PPI extraction method. At first, IRO is applied in a binary classifier to determine whether sentences contain a relation or not. Then, IRO is taken to guide PPI extraction by building sentence dependency parse tree. Comprehensive and quantitative evaluations and detailed analyses are used to demonstrate the significant performance of IRO on relation sentences classification and PPI extraction. Our PPI extraction method yielded a recall of around 80% and 90% and an F1 of around 54% and 66% on corpora of AIMed and BioInfer, respectively, which are superior to most existing extraction methods.

Coupling graphs are newly introduced in this paper to meet many application needs particularly in the field of bioinformatics. A coupling graph is a two-layer graph complex, in which each node from one layer of the graph complex has at least one connection with the nodes in the other layer, and vice versa. The coupling graph model is sufficiently powerful to capture strong and inherent associations between subgraph pairs in complicated applications. The focus of this paper is on mining algorithms of frequent coupling subgraphs and bioinformatics application. Although existing frequent subgraph mining algorithms are competent to identify frequent subgraphs from a graph database, they perform poorly on frequent coupling subgraph mining because they generate many irrelevant subgraphs. We propose a novel graph transformation technique to transform a coupling graph into a generic graph. Based on the transformed coupling graphs, existing graph mining methods are then utilized to discover frequent coupling subgraphs. We prove that the transformation is precise and complete and that the restoration is reversible. Experiments carried out on a database containing 10,511 coupling graphs show that our proposed algorithm reduces the mining time very much in comparison with the existing subgraph mining algorithms. Moreover, we demonstrate the usefulness of frequent coupling subgraphs by applying our algorithm to make accurate predictions of epitopes in antibody-antigen binding.

Characterization of binding hot spots of protein interfaces is a fundamental study in molecular biology. Many computational methods have been proposed to identify binding hot spots. However, there are few studies to assess the biological significance of binding hot spots. We introduce the notion of biological significance of a contact residue for capturing the probability of the residue occurring in or contributing to protein binding interfaces. We take a statistical Z-score approach to the assessment of the biological significance. The method has three main steps. First, the potential score of a residue is defined by using a knowledge-based potential function with relative accessible surface area calculations. A null distribution of this potential score is then generated from artifact crystal packing contacts. Finally, the Z-score significance of a contact residue with a specific potential score is determined according to this null distribution. We hypothesize that residues at binding hot spots have big absolute values of Z-score as they contribute greatly to binding free energy. Thus, we propose to use Z-score to predict whether a contact residue is a hot spot residue. Comparison with previously reported methods on two benchmark datasets shows that this Z-score method is mostly superior to earlier methods. This article is part of a Special Issue entitled: Computational Methods for Protein Interaction and Structural Prediction., (Crown Copyright © 2012. Published by Elsevier B.V. All rights reserved.)

Sequence-based understanding and identification of protein binding interfaces is a challenging research topic due to the complexity in protein systems and the imbalanced distribution between interface and noninterface residues. This paper presents an outlier detection idea to address the redundancy problem in protein interaction data. The cleaned training data are then used for improving the prediction performance. We use three novel measures to describe the extent a residue is considered as an outlier in comparison to the other residues: the distance of a residue instance from the center instance of all residue instances of the same class label (Dist), the probability of the class label of the residue instance (PCL), and the importance of within-class and between-class (IWB) residue instances. Outlier scores are computed by integrating the three factors; instances with a sufficiently large score are treated as outliers and removed. The data sets without outliers are taken as input for a support vector machine (SVM) ensemble. The proposed SVM ensemble trained on input data without outliers performs better than that with outliers. Our method is also more accurate than many literature methods on benchmark data sets. From our empirical studies, we found that some outlier interface residues are truly near to noninterface regions, and some outlier noninterface residues are close to interface regions.

Background: A protein binding hot spot is a cluster of residues in the interface that are energetically important for the binding of the protein with its interaction partner. Identifying protein binding hot spots can give useful information to protein engineering and drug design, and can also deepen our understanding of protein-protein interaction. These residues are usually buried inside the interface with very low solvent accessible surface area (SASA). Thus SASA is widely used as an outstanding feature in hot spot prediction by many computational methods. However, SASA is not capable of distinguishing slightly buried residues, of which most are non hot spots, and deeply buried ones that are usually inside a hot spot., Results: We propose a new descriptor called "burial level" for characterizing residues, atoms and atomic contacts. Specifically, burial level captures the depth the residues are buried. We identify different kinds of deeply buried atomic contacts (DBAC) at different burial levels that are directly broken in alanine substitution. We use their numbers as input for SVM to classify between hot spot or non hot spot residues. We achieve F measure of 0.6237 under the leave-one-out cross-validation on a data set containing 258 mutations. This performance is better than other computational methods., Conclusions: Our results show that hot spot residues tend to be deeply buried in the interface, not just having a low SASA value. This indicates that a high burial level is not only a necessary but also a more sufficient condition than a low SASA for a residue to be a hot spot residue. We find that those deeply buried atoms become increasingly more important when their burial levels rise up. This work also confirms the contribution of deeply buried interfacial atomic contacts to the energy of protein binding hot spot.

Background: Proteins interacting with each other as a complex play an important role in many molecular processes and functions. Directly detecting protein complexes is still costly, whereas many protein-protein interaction (PPI) maps for model organisms are available owing to the fast development of high-throughput PPI detecting techniques. These binary PPI data provides fundamental and abundant information for inferring new protein complexes. However, PPI data from different experiments do not overlap very much usually. The main reason is that the functions of proteins can activate only on certain environment or stimulus. In a short, PPI is condition-specific. Therefore specifying the conditions on when complexes are present is necessary for a deep understanding of their behaviours. Meanwhile, proteins have various interaction ways and control mechanisms to form different kinds of complexes. Thus the discovery of a certain type of complexes should depend on their own distinct biological or topological characteristics. We do not attempt to find all kinds of complexes by using certain features. Here, we integrate transcription regulation data (TR), gene expression data (GE) and protein-protein interaction data at the systems biology level to discover a special kind of protein complex called conditional co-regulated protein complexes. A conditional co-regulated protein complex has three remarkable features: the coding genes of the member proteins share the same transcription factor (TF), under a certain condition the coding genes express co-ordinately and the member proteins interact mutually as a complex to implement a common biological function., Results: A framework of discovering the conditional co-regulated protein complexes is proposed. Testing on the Yeast data sets under the Cell Cycle, DNA Damage and Dauxic Shift conditions, we identified a total of 29 conditional co-regulated complexes, among which the coding genes in 14 complexes show a strong association with their TFs activity. Based on the close relationship among co-regulation, co-expression and protein-protein interactions in the conditional co-regulated protein complexes, 39 novel TRs were predicted and explained., Conclusions: This paper was initiated to study conditional co-regulated protein complexes by integrating multiple data sources. Taking into consideration the influence of TFs activity on the protein interactions, we found that the expression coherence of the protein complexes' coding genes changed in accordance to their TFs' activity, which implied that the proteins' interactions also changed in response to the environments. Based on the three features of conditional co-regulated protein complexes, new transcriptional regulation interactions were predicted.

Unlabelled: Protein-protein interactions (PPIs) are one of the most important mechanisms in cellular processes. To model protein interaction sites, recent studies have suggested to find interacting protein group pairs from large PPI networks at the first step and then to search conserved motifs within the protein groups to form interacting motif pairs. To consider the noise effect and the incompleteness of biological data, we propose to use quasi-bicliques for finding interacting protein group pairs. We investigate two new problems that arise from finding interacting protein group pairs: the maximum vertex quasi-biclique problem and the maximum balanced quasi-biclique problem. We prove that both problems are NP-hard. This is a surprising result as the widely known maximum vertex biclique problem is polynomial time solvable [1]. We then propose a heuristic algorithm that uses the greedy method to find the quasi-bicliques from PPI networks. Our experiment results on real data show that this algorithm has a better performance than a benchmark algorithm for identifying highly matched BLOCKS and PRINTS motifs. We also report results of two case studies on interacting motif pairs that map well with two interacting domain pairs in iPfam., Availability: The software and supplementary information are available at http://www.cs.cityu.edu.hk/~lwang/software/ppi/index.html.

Classification which is a popular supervised machine learning method has many applications in computational biology, where data samples are automatically categorized into predefined labels with the aid of data mining. Often the training samples contain very few instances of interest (e.g., medical anomalies, rare disease in a population, and unusual syndromes, etc.), but many normal instances. Such imbalanced ratio of data distributions among the target labels hampers the efficacy of classification algorithms, because the induced model has not been trained with sufficient amount of instances of the interesting label(s), but overwhelmed with ordinary training records. Traditional remedies attempt to rebalance the data distributions of the target classes, by inflating the interesting instances artificially, reducing the majority of the common instances or a combination of both. Though the fundamental concept is effective, there is no clear guideline on how to strike a balance between fabricating the rare samples and reducing the norms, with the purpose of maximizing the classification accuracy. In this paper, an optimization model using different swarm strategies (Bat-inspired algorithm and PSO) is proposed for adaptively balancing the increase/decrease of the class distribution, depending on the properties of the biological datasets. The optimization is extended for achieving the highest possible accuracy and Kappa statistics at the same time as well. The optimization model is tested on five imbalanced medical datasets, which are sourced from lung surgery logs and virtual screening of bioassay data. Computer simulation results show that the proposed optimization model outperforms other class balancing methods in medical data classification. [ABSTRACT FROM AUTHOR]

Coupling graphs are newly introduced in this paper to meet many application needs particularly in the field of bioinformatics. A coupling graph is a two-layer graph complex, in which each node from one layer of the graph complex has at least one connection with the nodes in the other layer, and vice versa. The coupling graph model is sufficiently powerful to capture strong and inherent associations between subgraph pairs in complicated applications. The focus of this paper is on mining algorithms of frequent coupling subgraphs and bioinformatics application. Although existing frequent subgraph mining algorithms are competent to identify frequent subgraphs from a graph database, they perform poorly on frequent coupling subgraph mining because they generate many irrelevant subgraphs. We propose a novel graph transformation technique to transform a coupling graph into a generic graph. Based on the transformed coupling graphs, existing graph mining methods are then utilized to discover frequent coupling subgraphs. We prove that the transformation is precise and complete and that the restoration is reversible. Experiments carried out on a database containing 10,511 coupling graphs show that our proposed algorithm reduces the mining time very much in comparison with the existing subgraph mining algorithms. Moreover, we demonstrate the usefulness of frequent coupling subgraphs by applying our algorithm to make accurate predictions of epitopes in antibody-antigen binding. [ABSTRACT FROM PUBLISHER]

Context-awareness is a characteristic in the recognition between antigens and antibodies, highlighting the reconfiguration of epitope residues when an antigen interacts with a different antibody. A coarse binary classification of antigen regions into epitopes, or nonepitopes without specifying antibodies may not accurately reflect this biological reality. Therefore, we study an antibody-specified epitope prediction problem in line with this principle. This problem is new and challenging as we pinpoint a subset of the antigenic residues from an antigen when it binds to a specific antibody. We introduce two kinds of associations of the contextual awareness: 1) residues-residues pairing preference, and 2) the dependence between sets of contact residue pairs. Preference plays a bridging role to link interacting paratope and epitope residues while dependence is used to extend the association from one-dimension to two-dimension. The paratope/epitope residues' relative composition, cooperativity ratios, and Markov properties are also utilized to enhance our method. A nonredundant data set containing 80 antibody-antigen complexes is compiled and used in the evaluation. The results show that our method yields a good performance on antibody-specified epitope prediction. On the traditional antibody-ignored epitope prediction problem, a simplified version of our method can produce a competitive, sometimes much better, performance in comparison with three structure-based predictors. [ABSTRACT FROM PUBLISHER]

Co-regulations of miRNAs have been much less studied than the research on regulations between miRNAs and their target genes, although these two problems are equally important for understanding the entire mechanisms of complex post-transcriptional regulations. The difficulty to construct a miRNA–miRNA co-regulation network lies in how to determine reliable miRNA pairs from various resources of data related to the same disease such as expression levels, gene ontology (GO) databases, and protein–protein interactions. Here we take a novel integrative approach to the discovery of miRNA–miRNA co-regulation networks. This approach can progressively refine the various types of data and the computational analysis results. Applied to three lung cancer miRNA expression data sets of different subtypes, our method has identified a miRNA–miRNA co-regulation network and co-regulating functional modules common to lung cancer. An example of these functional modules consists of genes SMAD2, ACVR1B, ACVR2A and ACVR2B. This module is synergistically regulated by let-7a/b/c/f, is enriched in the same GO category, and has a close proximity in the protein interaction network. We also find that the co-regulation network is scale free and that lung cancer related miRNAs have more synergism in the network. According to our literature survey and database validation, many of these results are biologically meaningful for understanding the mechanism of the complex post-transcriptional regulations in lung cancer. [ABSTRACT FROM AUTHOR]