Predicting protein–protein interactions from protein sequences using meta predictor

A novel method is proposed for predicting protein–protein interactions (PPIs) based on the meta approach, which predicts PPIs using support vector machine that combines results by six independent state-of-the-art predictors. Significant improvement in prediction performance is observed, when performed on Saccharomyces cerevisiae and Helicobacter pylori datasets. In addition, we used the final prediction model trained on the PPIs dataset of S. cerevisiae to predict interactions in other species. The results reveal that our meta model is also capable of performing cross-species predictions. The source code and the datasets are available at     

PROTS-RF: A Robust Model for Predicting Mutation-Induced Protein Stability Changes

The ability to improve protein thermostability via protein engineering is of great scientific interest and also has significant practical value. In this report we present PROTS-RF, a robust model based on the Random Forest algorithm capable of predicting thermostability changes induced by not only single-, but also double- or multiple-point mutations. The model is built using 41 features including evolutionary information, secondary structure, solvent accessibility and a set of fragment-based features. It achieves accuracies of 0.799,0.782, 0.787, and areas under receiver operating characteristic (ROC) curves of 0.873, 0.868 and 0.862 for single-, double- and multiple- point mutation datasets, respectively. Contrary to previous suggestions, our results clearly demonstrate that a robust predictive model trained for predicting single point mutation induced thermostability changes can be capable of predicting double and multiple point mutations. It also shows high levels of robustness in the tests using hypothetical reverse mutations. We demonstrate that testing datasets created based on physical principles can be highly useful for testing the robustness of predictive models.     

Predicting co-complexed protein pairs from heterogeneous data

Proteins do not carry out their functions alone. Instead, they often act by participating in macromolecular complexes and play different functional roles depending on the other members of the complex. It is therefore interesting to identify co-complex relationships. Although protein complexes can be identified in a high-throughput manner by experimental technologies such as affinity purification coupled with mass spectrometry (APMS), these large-scale datasets often suffer from high false positive and false negative rates. Here, we present a computational method that predicts co-complexed protein pair (CCPP) relationships using kernel methods from heterogeneous data sources. We show that a diffusion kernel based on random walks on the full network topology yields good performance in predicting CCPPs from protein interaction networks. In the setting of direct ranking, a diffusion kernel performs much better than the mutual clustering coefficient. In the setting of SVM classifiers, a diffusion kernel performs much better than a linear kernel. We also show that combination of complementary information improves the performance of our CCPP recognizer. A summation of three diffusion kernels based on two-hybrid, APMS, and genetic interaction networks and three sequence kernels achieves better performance than the sequence kernels or diffusion kernels alone. Inclusion of additional features achieves a still better ROC(50) of 0.937. Assuming a negative-to-positive ratio of 600ratio1, the final classifier achieves 89.3% coverage at an estimated false discovery rate of 10%. Finally, we applied our prediction method to two recently described APMS datasets. We find that our predicted positives are highly enriched with CCPPs that are identified by both datasets, suggesting that our method successfully identifies true CCPPs. An SVM classifier trained from heterogeneous data sources provides accurate predictions of CCPPs in yeast. This computational method thereby provides an inexpensive method for identifying protein complexes that extends and complements high-throughput experimental data.     

Automated Identification of Subcellular Organelles by Coherent Anti-Stokes Raman Scattering

Coherent anti-Stokes Raman scattering (CARS) is an emerging tool for label-free characterization of living cells. Here, unsupervised multivariate analysis of CARS datasets was used to visualize the subcellular compartments. In addition, a supervised learning algorithm based on the “random forest” ensemble learning method as a classifier, was trained with CARS spectra using immunofluorescence images as a reference. The supervised classifier was then used, to our knowledge for the first time, to automatically identify lipid droplets, nucleus, nucleoli, and endoplasmic reticulum in datasets that are not used for training. These four subcellular components were simultaneously and label-free monitored instead of using several fluorescent labels. These results open new avenues for label-free time-resolved investigation of subcellular components in different cells, especially cancer cells.     

Automated Identification of Subcellular Organelles by Coherent Anti-Stokes Raman Scattering

Coherent anti-Stokes Raman scattering (CARS) is an emerging tool for label-free characterization of living cells. Here, unsupervised multivariate analysis of CARS datasets was used to visualize the subcellular compartments. In addition, a supervised learning algorithm based on the “random forest” ensemble learning method as a classifier, was trained with CARS spectra using immunofluorescence images as a reference. The supervised classifier was then used, to our knowledge for the first time, to automatically identify lipid droplets, nucleus, nucleoli, and endoplasmic reticulum in datasets that are not used for training. These four subcellular components were simultaneously and label-free monitored instead of using several fluorescent labels. These results open new avenues for label-free time-resolved investigation of subcellular components in different cells, especially cancer cells.     

BIPSPI+: Mining Type-Specific Datasets of Protein Complexes to Improve Protein Binding Site Prediction

Computational approaches for predicting protein-protein interfaces are extremely useful for understanding and modelling the quaternary structure of protein assemblies. In particular, partner-specific binding site prediction methods allow delineating the specific residues that compose the interface of protein complexes. In recent years, new machine learning and other algorithmic approaches have been proposed to solve this problem. However, little effort has been made in finding better training datasets to improve the performance of these methods. With the aim of vindicating the importance of the training set compilation procedure, in this work we present BIPSPI+, a new version of our original server trained on carefully curated datasets that outperforms our original predictor. We show how prediction performance can be improved by selecting specific datasets that better describe particular types of protein interactions and interfaces (e.g. homo/hetero). In addition, our upgraded web server offers a new set of functionalities such as the sequence-structure prediction mode, hetero- or homo-complex specialization and the guided docking tool that allows to compute 3D quaternary structure poses using the predicted interfaces. BIPSPI+ is freely available at https://bipspi.cnb.csic.es.     

A bayesian model for exploiting application constraints to enable unsupervised training of a P300-based BCI

This work introduces a novel classifier for a P300-based speller, which, contrary to common methods, can be trained entirely unsupervisedly using an Expectation Maximization approach, eliminating the need for costly dataset collection or tedious calibration sessions. We use publicly available datasets for validation of our method and show that our unsupervised classifier performs competitively with supervised state-of-the-art spellers. Finally, we demonstrate the added value of our method in different experimental settings which reflect realistic usage situations of increasing difficulty and which would be difficult or impossible to tackle with existing supervised or adaptive methods.     

A probabilistic model for mining implicit 'chemical compound-gene' relations from literature

MOTIVATION: The importance of chemical compounds has been emphasized more in molecular biology, and 'chemical genomics' has attracted a great deal of attention in recent years. Thus an important issue in current molecular biology is to identify biological-related chemical compounds (more specifically, drugs) and genes. Co-occurrence of biological entities in the literature is a simple, comprehensive and popular technique to find the association of these entities. Our focus is to mine implicit 'chemical compound and gene' relations from the co-occurrence in the literature. RESULTS: We propose a probabilistic model, called the mixture aspect model (MAM), and an algorithm for estimating its parameters to efficiently handle different types of co-occurrence datasets at once. We examined the performance of our approach not only by a cross-validation using the data generated from the MEDLINE records but also by a test using an independent human-curated dataset of the relationships between chemical compounds and genes in the ChEBI database. We performed experimentation on three different types of co-occurrence datasets (i.e. compound-gene, gene-gene and compound-compound co-occurrences) in both cases. Experimental results have shown that MAM trained by all datasets outperformed any simple model trained by other combinations of datasets with the difference being statistically significant in all cases. In particular, we found that incorporating compound-compound co-occurrences is the most effective in improving the predictive performance. We finally computed the likelihoods of all unknown compound-gene (more specifically, drug-gene) pairs using our approach and selected the top 20 pairs according to the likelihoods. We validated them from biological, medical and pharmaceutical viewpoints.     

Video Sequence-Based Iris Recognition Inspired by Human Cognition Manner

In video sequence-based iris recognition system, the problem of making full use of relationship and correlation among frames still remains to be solved. A brand new template level multimodal fusion algorithm inspired by human cognition manner is proposed. In that a non-isolated geometrical manifold, named Hyper Sausage Chain due to its sausage shape, is trained using the frames from a pattern class for representing an iris class in feature space. We can classify any input iris by observing which manifold it locates in. This process is closer to the function of human being, which takes ＇matter cognition＇ instead of ＇matter classification＇ as its basic principle. The experiments on self-developed JLUBR-IRIS dataset with several video sequences per person demonstrate the effectiveness and usability of the proposed algorithm for video sequence-based iris recognition. Fur- thermore, the comparative experiments on public CASIA-I and CASIA-V4-Interval datasets show that our method can also achieve improved performance of image-based iris recognition system, provided enough samples are involved in training stage.     

In silico Platform for Prediction of N-, O- and C-Glycosites in Eukaryotic Protein Sequences

Glycosylation is one of the most abundant and an important post-translational modification of proteins. Glycosylated proteins (glycoproteins) are involved in various cellular biological functions like protein folding, cell-cell interactions, cell recognition and host-pathogen interactions. A large number of eukaryotic glycoproteins also have therapeutic and potential technology applications. Therefore, characterization and analysis of glycosites (glycosylated residues) in these proteins is of great interest to biologists. In order to cater these needs a number of in silico tools have been developed over the years, however, a need to get even better prediction tools remains. Therefore, in this study we have developed a new webserver GlycoEP for more accurate prediction of N-linked, O-linked and C-linked glycosites in eukaryotic glycoproteins using two larger datasets, namely, standard and advanced datasets. In case of standard datasets no two glycosylated proteins are more similar than 40%; advanced datasets are highly non-redundant where no two glycosites’ patterns (as defined in methods) have more than 60% similarity. Further, based on our results with several algorihtms developed using different machine-learning techniques, we found Support Vector Machine (SVM) as optimum tool to develop glycosite prediction models. Accordingly, using our more stringent and non-redundant advanced datasets, the SVM based models developed in this study achieved a prediction accuracy of 84.26%, 86.87% and 91.43% with corresponding MCC of 0.54, 0.20 and 0.78, for N-, O- and C-linked glycosites, respectively. The best performing models trained on advanced datasets were then implemented as a user-friendly web server GlycoEP (http://www.imtech.res.in/raghava/glycoep/). Additionally, this server provides prediction models developed on standard datasets and allows users to scan sequons in input protein sequences.     

Comparative microbial modules resource: generation and visualization of multi-species biclusters

The increasing abundance of large-scale, high-throughput datasets for many closely related organisms provides opportunities for comparative analysis via the simultaneous biclustering of datasets from multiple species. These analyses require a reformulation of how to organize multi-species datasets and visualize comparative genomics data analyses results. Recently, we developed a method, multi-species cMonkey, which integrates heterogeneous high-throughput datatypes from multiple species to identify conserved regulatory modules. Here we present an integrated data visualization system, built upon the Gaggle, enabling exploration of our method's results (available at http://meatwad.bio.nyu.edu/cmmr.html). The system can also be used to explore other comparative genomics datasets and outputs from other data analysis procedures - results from other multiple-species clustering programs or from independent clustering of different single-species datasets. We provide an example use of our system for two bacteria, Escherichia coli and Salmonella Typhimurium. We illustrate the use of our system by exploring conserved biclusters involved in nitrogen metabolism, uncovering a putative function for yjjI, a currently uncharacterized gene that we predict to be involved in nitrogen assimilation.     

iDrug-Target: predicting the interactions between drug compounds and target proteins in cellular networking via benchmark dataset optimization approach

Information about the interactions of drug compounds with proteins in cellular networking is very important for drug development. Unfortunately, all the existing predictors for identifying drug–protein interactions were trained by a skewed benchmark data-set where the number of non-interactive drug–protein pairs is overwhelmingly larger than that of the interactive ones. Using this kind of highly unbalanced benchmark data-set to train predictors would lead to the outcome that many interactive drug–protein pairs might be mispredicted as non-interactive. Since the minority interactive pairs often contain the most important information for drug design, it is necessary to minimize this kind of misprediction. In this study, we adopted the neighborhood cleaning rule and synthetic minority over-sampling technique to treat the skewed benchmark datasets and balance the positive and negative subsets. The new benchmark datasets thus obtained are called the optimized benchmark datasets, based on which a new predictor called iDrug-Target was developed that contains four sub-predictors: iDrug-GPCR, iDrug-Chl, iDrug-Ezy, and iDrug-NR, specialized for identifying the interactions of drug compounds with GPCRs (G-protein-coupled receptors), ion channels, enzymes, and NR (nuclear receptors), respectively. Rigorous cross-validations on a set of experiment-confirmed datasets have indicated that these new predictors remarkably outperformed the existing ones for the same purpose. To maximize users’ convenience, a public accessible Web server for iDrug-Target has been established at http://www.jci-bioinfo.cn/iDrug-Target/, by which users can easily get their desired results. It has not escaped our notice that the aforementioned strategy can be widely used in many other areas as well.     

Differentially private distributed logistic regression using private and public data

Background

Privacy protecting is an important issue in medical informatics and differential privacy is a state-of-the-art framework for data privacy research. Differential privacy offers provable privacy against attackers who have auxiliary information, and can be applied to data mining models (for example, logistic regression). However, differentially private methods sometimes introduce too much noise and make outputs less useful. Given available public data in medical research (e.g. from patients who sign open-consent agreements), we can design algorithms that use both public and private data sets to decrease the amount of noise that is introduced.

Methodology

In this paper, we modify the update step in Newton-Raphson method to propose a differentially private distributed logistic regression model based on both public and private data.

Experiments and results

We try our algorithm on three different data sets, and show its advantage over: (1) a logistic regression model based solely on public data, and (2) a differentially private distributed logistic regression model based on private data under various scenarios.

Conclusion

Logistic regression models built with our new algorithm based on both private and public datasets demonstrate better utility than models that trained on private or public datasets alone without sacrificing the rigorous privacy guarantee.

     

A Collection of Amino Acid Replacement Matrices Derived from Clusters of Orthologs

Sequence divergence among orthologous proteins was characterized with 34 amino acid replacement matrices, sequence context analysis, and a phylogenetic tree. The model was trained on very large datasets of aligned protein sequences drawn from 15 organisms including protists, plants, Dictyostelium, fungi, and animals. Comparative tests with models currently used in phylogeny, i.e., with JTT+Γ±F and WAG+Γ±F, made on a test dataset of 380 multiple alignments containing protein sequences from all five of the major taxonomic groups mentioned, indicate that our model should be preferred over the JTT+Γ±F and WAG+Γ±F models on datasets similar to the test dataset. The strong performance of our model of orthologous protein sequence divergence can be attributed to its ability to better approximate amino acid equilibrium frequencies to compositions found in alignment columns. Electronic Supplementary Material Electronic Supplementary material is available for this article at and accessible for authorised users. [Reviewing Editor : Dr. Martin Kreitman]     

Building pathway clusters from Random Forests classification using class votes

Background

As one of the most common protein post-translational modifications, glycosylation is involved in a variety of important biological processes. Computational identification of glycosylation sites in protein sequences becomes increasingly important in the post-genomic era. A new encoding scheme was employed to improve the prediction of mucin-type O-glycosylation sites in mammalian proteins.

Results

A new protein bioinformatics tool, CKSAAP_OGlySite, was developed to predict mucin-type O-glycosylation serine/threonine (S/T) sites in mammalian proteins. Using the composition of k-spaced amino acid pairs (CKSAAP) based encoding scheme, the proposed method was trained and tested in a new and stringent O-glycosylation dataset with the assistance of Support Vector Machine (SVM). When the ratio of O-glycosylation to non-glycosylation sites in training datasets was set as 1:1, 10-fold cross-validation tests showed that the proposed method yielded a high accuracy of 83.1% and 81.4% in predicting O-glycosylated S and T sites, respectively. Based on the same datasets, CKSAAP_OGlySite resulted in a higher accuracy than the conventional binary encoding based method (about +5.0%). When trained and tested in 1:5 datasets, the CKSAAP encoding showed a more significant improvement than the binary encoding. We also merged the training datasets of S and T sites and integrated the prediction of S and T sites into one single predictor (i.e. S+T predictor). Either in 1:1 or 1:5 datasets, the performance of this S+T predictor was always slightly better than those predictors where S and T sites were independently predicted, suggesting that the molecular recognition of O-glycosylated S/T sites seems to be similar and the increase of the S+T predictor's accuracy may be a result of expanded training datasets. Moreover, CKSAAP_OGlySite was also shown to have better performance when benchmarked against two existing predictors.

Conclusion

Because of CKSAAP encoding's ability of reflecting characteristics of the sequences surrounding mucin-type O-glycosylation sites, CKSAAP_ OGlySite has been proved more powerful than the conventional binary encoding based method. This suggests that it can be used as a competitive mucin-type O-glycosylation site predictor to the biological community. CKSAAP_OGlySite is now available at http://bioinformatics.cau.edu.cn/zzd_lab/CKSAAP_OGlySite/.     

Global discriminative learning for higher-accuracy computational gene prediction

Most ab initio gene predictors use a probabilistic sequence model, typically a hidden Markov model, to combine separately trained models of genomic signals and content. By combining separate models of relevant genomic features, such gene predictors can exploit small training sets and incomplete annotations, and can be trained fairly efficiently. However, that type of piecewise training does not optimize prediction accuracy and has difficulty in accounting for statistical dependencies among different parts of the gene model. With genomic information being created at an ever-increasing rate, it is worth investigating alternative approaches in which many different types of genomic evidence, with complex statistical dependencies, can be integrated by discriminative learning to maximize annotation accuracy. Among discriminative learning methods, large-margin classifiers have become prominent because of the success of support vector machines (SVM) in many classification tasks. We describe CRAIG, a new program for ab initio gene prediction based on a conditional random field model with semi-Markov structure that is trained with an online large-margin algorithm related to multiclass SVMs. Our experiments on benchmark vertebrate datasets and on regions from the ENCODE project show significant improvements in prediction accuracy over published gene predictors that use intrinsic features only, particularly at the gene level and on genes with long introns.     

A hidden Markov model-based approach for identifying timing differences in gene expression under different experimental factors

MOTIVATION: Time series experiments of cDNA microarrays have been commonly used in various biological studies and conducted under a lot of experimental factors. A popular approach of time series microarray analysis is to compare one gene with another in their expression profiles, and clustering expression sequences is a typical example. On the other hand, a practically important issue in gene expression is to identify the general timing difference that is caused by experimental factors. This type of difference can be extracted by comparing a set of time series expression profiles under a factor with those under another factor, and so it would be difficult to tackle this issue by using only a current approach for time series microarray analysis. RESULTS: We have developed a systematic method to capture the timing difference in gene expression under different experimental factors, based on hidden Markov models. Our model outputs a real-valued vector at each state and has a unique state transition diagram. The parameters of our model are trained from a given set of pairwise (generally multiplewise) expression sequences. We evaluated our model using synthetic as well as real microarray datasets. The results of our experiment indicate that our method worked favourably to identify the timing ordering under different experimental factors, such as that gene expression under heat shock tended to start earlier than that under oxidative stress. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.     

On biometric systems: electrocardiogram Gaussianity and data synthesis

Electrocardiogram is a slow signal to acquire, and it is prone to noise. It can be inconvenient to collect large number of ECG heartbeats in order to train a reliable biometric system; hence, this issue might result in a small sample size phenomenon which occurs when the number of samples is much smaller than the number of observations to model. In this paper, we study ECG heartbeat Gaussianity and we generate synthesized data to increase the number of observations. Data synthesis, in this paper, is based on our hypothesis, which we support, that ECG heartbeats exhibit a multivariate normal distribution; therefore, one can generate ECG heartbeats from such distribution. This distribution is deviated from Gaussianity due to internal and external factors that change ECG morphology such as noise, diet, physical and psychological changes, and other factors, but we attempt to capture the underlying Gaussianity of the heartbeats. When this method was implemented for a biometric system and was examined on the University of Toronto database of 1012 subjects, an equal error rate (EER) of 6.71% was achieved in comparison to 9.35% to the same system but without data synthesis. Dimensionality reduction is widely examined in the problem of small sample size; however, our results suggest that using the proposed data synthesis outperformed several dimensionality reduction techniques by at least 3.21% in EER. With small sample size, classifier instability becomes a bigger issue and we used a parallel classifier scheme to reduce it. Each classifier in the parallel classifier is trained with the same genuine dataset but different imposter datasets. The parallel classifier has reduced predictors’ true acceptance rate instability from 6.52% standard deviation to 1.94% standard deviation.