Similarity-based Prediction Research Articles

GalaxySagittarius: Structure- and Similarity-Based Prediction of Protein Targets for Druglike Compounds

Computational techniques for predicting interactions of proteins and druglike molecules have often been used to search for compounds that bind a given protein with high affinity. More recently, such tools have also been applied to the reverse procedure of searching protein targets for a given compound. Among methods for predicting protein-ligand interactions, ligand-based methods relying on similarity to ligands of known interactions are effective only when similar protein-ligand interactions are known. Receptor-based methods predicting protein-ligand interactions by molecular docking are effective only when high-accuracy receptor structures and binding sites are available. Moreover, the computational cost of molecular docking tends to be too high to be applied to the entire protein structure database. In this paper, an effective target prediction method, which combines ligand similarity-based and receptor structure-based approaches, is introduced. In this method, protein-ligand docking is performed after efficient structure- and similarity-based screening. The enriched protein target database by predicted binding ligands and sites allows detection of protein targets with previously unknown ligand interactions. The method, called GalaxySagittarius, is freely available as a web server at http://galaxy.seoklab.org/sagittarius.

Similarity-based deep learning approach for remaining useful life prediction

Abstract Remaining useful life (RUL) prediction has been a difficult problem in industry and academia. Deep learning techniques based on supervised learning have made great progress compared to traditional prediction algorithms. However, due to the random nature of system degradation behavior, deep supervised learning cannot accurately adapt the model obtained based on the training set to the systems in the testing set. The similarity-based prediction method has achieved accurate approximate remaining life, which called similar RUL. Therefore, the combination of supervised learning and similarity-based prediction has the potential to take advantages of the learning ability of the network model and similar RUL to improve the prediction accuracy. This paper investigates the effect of using similarity to improve the prediction accuracy of deep supervised learning. In addition, since health indicator (HI) and target labels are the essential factors of supervised learning, this paper proposes a method for constructing HI based on unsupervised learning. Then according to the constructed HI, a difference-based method proposed to generate target labels. To verify the effectiveness of the proposed method, this paper conducted experimental verification on the C-MAPSS engine dataset and XJTU-SY bearing dataset and made related comparative experiments. Experimental results show that the proposed method obtains favorable results against the traditional methods.

Improved trajectory similarity-based approach for turbofan engine prognostics

Trajectory similarity-based prediction (TSBP) is an emerging real-time remaining useful life (RUL) prediction method that has drawn considerable attention in the field of data-driven prognostics. TSBP is fast, and the corresponding model is easy to train. However, TBSP only provides a point estimation of RUL, which is insufficient for some specific prognostic applications. Hence, this study introduces an improved TSBP method to handle the issue of prognostic uncertainty. On the basis of an adaptive kernel density estimation technique and β-criterion, the improved TSBP method not only provides an accurate and precise point prediction of RUL but also specifies the confidence interval of RUL prediction. The capability of obtaining the confidence interval of RUL can enhance the TSBP method for uncertainty management. The effectiveness of the proposed method is validated through two cases studies, which are related to turbofan engine prognostics.

Counts-of-counts similarity for prediction and search in relational data

Defining appropriate distance functions is a crucial aspect of effective and efficient similarity-based prediction and retrieval. Relational data are especially challenging in this regard. By viewing relational data as multi-relational graphs, one can easily see that a distance between a pair of nodes can be defined in terms of a virtually unlimited class of features, including node attributes, attributes of node neighbors, structural aspects of the node neighborhood and arbitrary combinations of these properties. In this paper we propose a rich and flexible class of metrics on graph entities based on earth mover’s distance applied to a hierarchy of complex counts-of-counts statistics. We further propose an approximate version of the distance using sums of marginal earth mover’s distances. We show that the approximation is correct for many cases of practical interest and allows efficient nearest-neighbor retrieval when combined with a simple metric tree data structure. An experimental evaluation on two real-world scenarios highlights the flexibility of our framework for designing metrics representing different notions of similarity. Substantial improvements in similarity-based prediction are reported when compared to solutions based on state-of-the-art graph kernels.

OPA2Vec: combining formal and informal content of biomedical ontologies to improve similarity-based prediction.

Ontologies are widely used in biology for data annotation, integration and analysis. In addition to formally structured axioms, ontologies contain meta-data in the form of annotation axioms which provide valuable pieces of information that characterize ontology classes. Annotation axioms commonly used in ontologies include class labels, descriptions or synonyms. Despite being a rich source of semantic information, the ontology meta-data are generally unexploited by ontology-based analysis methods such as semantic similarity measures. We propose a novel method, OPA2Vec, to generate vector representations of biological entities in ontologies by combining formal ontology axioms and annotation axioms from the ontology meta-data. We apply a Word2Vec model that has been pre-trained on either a corpus or abstracts or full-text articles to produce feature vectors from our collected data. We validate our method in two different ways: first, we use the obtained vector representations of proteins in a similarity measure to predict protein-protein interaction on two different datasets. Second, we evaluate our method on predicting gene-disease associations based on phenotype similarity by generating vector representations of genes and diseases using a phenotype ontology, and applying the obtained vectors to predict gene-disease associations using mouse model phenotypes. We demonstrate that OPA2Vec significantly outperforms existing methods for predicting gene-disease associations. Using evidence from mouse models, we apply OPA2Vec to identify candidate genes for several thousand rare and orphan diseases. OPA2Vec can be used to produce vector representations of any biomedical entity given any type of biomedical ontology. https://github.com/bio-ontology-research-group/opa2vec. Supplementary data are available at Bioinformatics online.

Survey of Similarity-based Prediction of Drug-protein Interactions.

Survey of Similarity-based Prediction of Drug-protein Interactions.

Onto2Vec: joint vector-based representation of biological entities and their ontology-based annotations.

MotivationBiological knowledge is widely represented in the form of ontology-based annotations: ontologies describe the phenomena assumed to exist within a domain, and the annotations associate a (kind of) biological entity with a set of phenomena within the domain. The structure and information contained in ontologies and their annotations make them valuable for developing machine learning, data analysis and knowledge extraction algorithms; notably, semantic similarity is widely used to identify relations between biological entities, and ontology-based annotations are frequently used as features in machine learning applications.ResultsWe propose the Onto2Vec method, an approach to learn feature vectors for biological entities based on their annotations to biomedical ontologies. Our method can be applied to a wide range of bioinformatics research problems such as similarity-based prediction of interactions between proteins, classification of interaction types using supervised learning, or clustering. To evaluate Onto2Vec, we use the gene ontology (GO) and jointly produce dense vector representations of proteins, the GO classes to which they are annotated, and the axioms in GO that constrain these classes. First, we demonstrate that Onto2Vec-generated feature vectors can significantly improve prediction of protein–protein interactions in human and yeast. We then illustrate how Onto2Vec representations provide the means for constructing data-driven, trainable semantic similarity measures that can be used to identify particular relations between proteins. Finally, we use an unsupervised clustering approach to identify protein families based on their Enzyme Commission numbers. Our results demonstrate that Onto2Vec can generate high quality feature vectors from biological entities and ontologies. Onto2Vec has the potential to significantly outperform the state-of-the-art in several predictive applications in which ontologies are involved.Availability and implementation https://github.com/bio-ontology-research-group/onto2vec Supplementary information Supplementary data are available at Bioinformatics online.

Accurate prediction of orthologs in the presence of divergence after duplication.

MotivationWhen gene duplication occurs, one of the copies may become free of selective pressure and evolve at an accelerated pace. This has important consequences on the prediction of orthology relationships, since two orthologous genes separated by divergence after duplication may differ in both sequence and function. In this work, we make the distinction between the primary orthologs, which have not been affected by accelerated mutation rates on their evolutionary path, and the secondary orthologs, which have. Similarity-based prediction methods will tend to miss secondary orthologs, whereas phylogeny-based methods cannot separate primary and secondary orthologs. However, both types of orthology have applications in important areas such as gene function prediction and phylogenetic reconstruction, motivating the need for methods that can distinguish the two types.ResultsWe formalize the notion of divergence after duplication and provide a theoretical basis for the inference of primary and secondary orthologs. We then put these ideas to practice with the Hybrid Prediction of Paralogs and Orthologs (HyPPO) framework, which combines ideas from both similarity and phylogeny approaches. We apply our method to simulated and empirical datasets and show that we achieve superior accuracy in predicting primary orthologs, secondary orthologs and paralogs.Availability and implementationHyPPO is a modular framework with a core developed in Python and is provided with a variety of C++ modules. The source code is available at https://github.com/manuellafond/HyPPO.Supplementary information Supplementary data are available at Bioinformatics online.

A novel link prediction method for supervising transitivity process

Link prediction has become an important area in network analysis in recent years due to its theoretical and practical significance. In this paper, we present a similarity-based prediction method under simultaneous consideration of multiple information sources and the corresponding discrimination ability. We first propose a novel supervised transitivity similarity index (STSI), in which the likelihood ratio in the Bayesian theory is employed to supervise the transitivity process. Then, based on the proposed STSI, we design a supervised transitivity similarity algorithm (STSA) for predicting missing links. Finally, empirical experiments are conducted to demonstrate the effectiveness of the proposed method. The experimental results show that our method can achieve a good performance, compared with other mainstream baselines.

Structural similarity-based prediction of the potential active ingredients and mechanism of action of traditional Chinese medicine formulations used to anti-aging

ObjectiveTo predict the potential active ingredients (PAIs) and mechanism of action of traditional Chinese medicine formulations used to delay aging. MethodsWe incorporated the use of quantum-chemistry calculations and machine learning to predict the active ingredients of some Chinese herbal medicines used to delay aging. Then, a network-pharmacology approach was used to uncover how these PAIs delayed aging. ResultsTwelve PAIs with anti-aging effects were discovered: androsterone, MHP, cortisone, propyl methyl trisulfide, retinol, retinal, cortisol, 11-cis-Retinol (2R, 3R)-3-hydroxyproline, 4,5alpha-Dihydrocortisone (2S)-2-ammonio-6-ureidohexanoate and17alpha, 21-Dihydroxy-5beta-pregnane-3,11,20-trione. Enrichment analyses indicated that a putative compound target and aging target were significantly associated with: regulation of the immune system; insulin receptor signaling pathway; regulation of the mitotic cell cycle; response to nutrient levels; response to oxidative stress; release of cytochrome c from mitochondria; learning or memory; inflammatory response. ConclusionsA novel method was proposed to predict the PAIs of anti-aging herbal medicines by incorporating quantum-chemistry calculations and machine learning. Then, a network-pharmacology approach was used to uncover how these PAIs delay aging. The information provided by our study on PAIs may aid the discovery of anti-aging drugs.

The Bi-Direction Similarity Integration Method for Predicting Microbe-Disease Associations

Identification of microbe-disease associations provides insight into the mechanism that microbes cause diseases at the molecular level. Existing microbe–disease association prediction methods mainly utilize microbe–disease association profiles to calculate microbe–microbe similarities and disease–disease similarities, and then build similarity-based prediction models. However, they ignore important biological knowledge, e.g., disease Medical Subject Headings (MeSH), and do not consider unequal contributions of microbe information and disease information. In this paper, we propose the bi-direction similarity integration label propagation (BDSILP) method for predicting microbe–disease associations. First, BDSILP introduces disease MeSH to calculate the disease–disease semantic similarity and the microbe–microbe functional similarity. Although MeSH is not available for all diseases, BDSILP presents a strategy for integrating multiple similarities for microbes and diseases. Second, two graphs are constructed by using integrated disease similarity and integrated microbe similarity, and BDSILP implements the label propagation on the graphs to score microbe–disease pairs. Third, BDSILP adopts the weighted averages of their scores as final predictions. BDSILP produces better performances than existing state-of-the-art methods, achieving the AUC of 0.9131 and the AUPR of 0.5343 in leave-one-out cross validation, and achieving the AUC of 0.9051 and the AUPR of 0.3037 in five-fold cross validation. Moreover, case studies and discussion demonstrate that BDSILP is promising for predicting novel microbe–disease associations.

Long-term prediction of system degradation with similarity analysis of multivariate patterns

To forecast the long-term degradation behavior of mechanical systems, a method named Distance Based Sequential Aggregation with Gaussian Mixture Model (DBSA-GMM) that completes predictions with two steps was proposed: first, it calculates the statistical distances (SD-s) of the objective degradation signature pattern to historical precursors, then, it uses the SD-S to generate hypothetical Gaussian estimations of the objective features, and synthesizes these Gaussians to build up Gaussian Mixture Model (GMM) approximations of feature Probability Density Functions (PDF-s) with a newly proposed algorithm called Descending Order Aggregation (DOA). DBSA-GMM was applied in the condition prediction of a fleet of commercial aero-engines and showed advantageous prediction precision over Auto-Regressive Moving Average (ARMA), Back Propagation Artificial Neural Network (BP-ANN), and former similarity based prediction (SBP) methods. Meanwhile, DOA was also validated to be with higher generalization ability with additional tests on outlier samples against Kernel Density Estimation (KDE) method.

A hybrid quantum-induced swarm intelligence clustering for the urban trip recommendation in smart city

The development of internet technologies has brought digital services to the hands of common man. In the selection process of relevant digital services to the active target user, recommender systems have proved its efficiency as a successful decision support tool. Among many successful techniques incorporated to generate recommendations, collaborative filtering has been widely used to make similarity-based predictions for the recommendation of the relevant list of items to the users. As an advancement, utilizing clustering mechanisms with collaborative filtering for grouping similar users as clusters can enhance the efficiency of the recommendation generated. Though many clustering mechanisms have been employed to group similar users in the existing works, incorporation of bio-inspired clustering has yet to be explored for the generation of optimal recommendations. In this paper, a novel user clustering approach based on Quantum-behaved Particle Swarm Optimization (QPSO) has been proposed for the collaborative filtering based recommender system. The proposed recommendation approach has been evaluated on real-world large-scale datasets of Yelp and TripAdvisor for hit-rate, precision, recall, f-measure, and accuracy. The obtained results illustrate the advantageous performance of proposed approach over its peer works of recent times. We have also developed a new mobile recommendation framework XplorerVU for the urban trip recommendation in smart cities, to evaluate the proposed recommendation approach and the real-time implementation details of the mobile application in the smart-cities are also presented. The evaluation results prove the usefulness of the generated recommendations and depict the users’ satisfaction on the proposed recommendation approach.

Shrubby cinquefoil (Dasiphora fruticosa (L.) Rydb.) mapping in Northwestern Estonia based upon site similarities

BackgroundDifferent methods have been used to map species and habitat distributions. In this paper, similarity-based reasoning—a methodological approach that has received less attention—was applied to estimate the distribution and coverage of Dasiphora fruticosa for the region in the Baltic states where grows the most abundant population of this species.MethodsField observations, after thinning to at least 50 m interval, included 1480 coverage estimations in the species presence locations and 8317 absence locations. Species coverage for the 750 km2 of directly unobserved area was calculated using machine learning in the similarity-based prediction system Constud. Separate predictive sets of site features (e.g. land cover, soil type) and exemplar weights were calibrated for spatial partitions of the study area (probable presence region, unclear region, proved absence region). A modified version of the Gower’s distance metric, as used in Constud, is described.ResultsThe resulting maps depicted the predicted coverage, the certainty of decision when predicting presence or absence, and the mean similarity to the exemplar locations used while predicting. Coverage prediction errors were smaller in the unclear partition—where the species was mostly absent—than in the probable presence partition, where coverage ranged from 0 to 90%.ConclusionsWe call for methodological comparisons using the same data set.

Estimating multilateral trade behaviors on the world trade web with limited information

The privacy and the confidentiality limit the amount of the accessible information in investigating socioeconomic systems. Reconstructing network with limited information is of great application importance. Most existing network reconstruction methods concentrate on completely connected networks, undirected or unweighed networks. In the paper, we present a novel approach to reconstruct the topological structure of directed weighted network (DWN) for estimating multilateral trade behaviors on the World Trade Web using limited information: the values of out- and in-strengths of nodes, and the total number of links and nodes. Our approach uses first out- and in-strength fitnesses to estimate the unknown out- and in-degrees. Then we propose a directed enhanced configuration model (DECM) and reconstruct the topological structure of real DWN. This approach is experimentally verified on real-world networks, using the significantly topological properties: assortativity and clustering. The results show a good agreement between these quantities calculated on real DWN and its DECM-induced ensemble averages. Furthermore, we compare DECM algorithm with weighted node similarity algorithms for link prediction, demonstrating that DECM outperforms similarity-based link prediction algorithms. Finally, the comparison between DECM and directed weighted configuration model (DWCM) using Information-theoretic criteria rigorously confirms that DECM evidently outperforms DWCM for reconstructing network topological structure. Accordingly, this approach can be employed as an important tool for providing deeper insights into the privacy protection of socioeconomic networks.

Soil polygon disaggregation through similarity-based prediction with legacy pedons

Conventional soil maps generally contain one or more soil types within a single soil polygon. But their geographic locations within the polygon are not specified. This restricts current applications of the maps in site-specific agricultural management and environmental modelling. We examined the utility of legacy pedon data for disaggregating soil polygons and the effectiveness of similarity-based prediction for making use of the under- or over-sampled legacy pedon data for the disaggregation. The method consisted of three steps. First, environmental similarities between the pedon sites and each location were computed based on soil formative environmental factors. Second, according to soil types of the pedon sites, the similarities were aggregated to derive similarity distribution for each soil type. Third, a hardening process was performed on the maps to allocate candidate soil types within the polygons. The study was conducted at the soil subgroup level in a semi-arid area situated in Manitoba, Canada. Based on 186 independent pedon sites, the evaluation of the disaggregated map of soil subgroups showed an overall accuracy of 67% and a Kappa statistic of 0.62. The map represented a better spatial pattern of soil subgroups in both detail and accuracy compared to a dominant soil subgroup map, which was commonly used in practice. Incorrect predictions mainly occurred in the agricultural plain area and the soil subgroups that are very similar in taxonomy, indicating that new environmental covariates need to be developed. We concluded that the combination of legacy pedon data with similarity-based prediction is an effective solution for soil polygon disaggregation.

A link prediction algorithm based on label propagation

The study of link prediction in graph theory has received more and more attention in recent years. Considering the attributes of nodes in online social networks are generally inaccurate, it is very important and efficient to use the network structure characteristics rather than nodes’ information to predict edges in networks. In this paper, we present a simple but effective similarity-based prediction strategy based on label propagation, which mimics the communication between people naturally. We perform an experimental comparison of the proposed method against four classic local similarity-based link prediction algorithms using real-world networks. The experimental results show that our method offers higher precision than these well-known approaches. Hence, we can provide more accurate friend recommendations for online social networks and reduce experimental costs in the fields of biology, and better understand the evolution mechanism of complex networks.

Molecular similarity-based predictions of the Tox21 screening outcome

To assess the toxicity of new chemicals and drugs, regulatory agencies require in vivo testing for many toxic endpoints, resulting in millions of animal experiments conducted each year. However, following the Replace, Reduce, Refine (3R) principle, the development and optimization of alternative methods, in particular in silico methods, has been put into focus in the recent years. It is generally acknowledged that the more complex a toxic endpoint, the more difficult it is to model. Therefore, computational toxicology is shifting from modelling general and complex endpoints to the investigation and modelling of pathways of toxicity and the underlying molecular effects. The U.S. Toxicology in the 21st Century (Tox21) initiative has screened a large library of compounds, including approximately 10K environmental chemicals and drugs, for different mechanisms responsible for eliciting toxic effects, and made the results publicly available. Through the Tox21 Data Challenge, the consortium has established a platform for computational toxicologists to develop and validate their predictive models. Here, we present a fast and successful method for the prediction of different outcomes of the nuclear receptor and stress response pathway screening from the Tox21 Data Challenge 2014. The method is based on the combination of molecular similarity calculations and a naïve Bayes machine learning algorithm and has been implemented as a KNIME pipeline. Molecules are represented as binary vectors consisting of a concatenation of common two-dimensional molecular fingerprint types with topological compound properties. The prediction method has been optimized individually for each modelled target and evaluated in a cross-validation as well as with the independent Tox21 validation set. Our results show that the method can achieve good prediction accuracies and rank among the top algorithms submitted to the prediction challenge, indicating its broad applicability in toxicity prediction.

Similarity-based prediction for Anatomical Therapeutic Chemical classification of drugs by integrating multiple data sources.

Anatomical Therapeutic Chemical (ATC) classification system, widely applied in almost all drug utilization studies, is currently the most widely recognized classification system for drugs. Currently, new drug entries are added into the system only on users' requests, which leads to seriously incomplete drug coverage of the system, and bioinformatics prediction is helpful during this process. Here we propose a novel prediction model of drug-ATC code associations, using logistic regression to integrate multiple heterogeneous data sources including chemical structures, target proteins, gene expression, side-effects and chemical-chemical associations. The model obtains good performance for the prediction not only on ATC codes of unclassified drugs but also on new ATC codes of classified drugs assessed by cross-validation and independent test sets, and its efficacy exceeds previous methods. Further to facilitate the use, the model is developed into a user-friendly web service SPACE ( S: imilarity-based P: redictor of A: TC C: od E: ), which for each submitted compound, will give candidate ATC codes (ranked according to the decreasing probability_score predicted by the model) together with corresponding supporting evidence. This work not only contributes to knowing drugs' therapeutic, pharmacological and chemical properties, but also provides clues for drug repositioning and side-effect discovery. In addition, the construction of the prediction model also provides a general framework for similarity-based data integration which is suitable for other drug-related studies such as target, side-effect prediction etc. The web service SPACE is available at http://www.bprc.ac.cn/space.

Mapping soil salinity using a similarity-based prediction approach: A case study in Huanghe River Delta, China

Spatial distribution of soil salinity can be estimated based on its environmental factors because soil salinity is strongly affected and indicated by environmental factors. Different with other properties such as soil texture, soil salinity varies with short-term time. Thus, how to choose powerful environmental predictors is especially important for soil salinity. This paper presents a similarity-based prediction approach to map soil salinity and detects powerful environmental predictors for the Huanghe (Yellow) River Delta area in China. The similarity-based approach predicts the soil salinities of unsampled locations based on the environmental similarity between unsampled and sampled locations. A dataset of 92 points with salt data at depth of 30–40 cm was divided into two subsets for prediction and validation. Topographical parameters, soil textures, distances to irrigation channels and to the coastline, land surface temperature from Moderate Resolution Imaging Spectroradiometer (MODIS), Normalized Difference Vegetation Indices (NDVIs) and land surface reflectance data from Landsat Thematic Mapper (TM) imagery were generated. The similarity-based prediction approach was applied on several combinations of different environmental factors. Based on three evaluation indices including the correlation coefficient (CC) between observed and predicted values, the mean absolute error and the root mean squared error we found that elevation, distance to irrigation channels, soil texture, night land surface temperature, NDVI, and land surface reflectance Band 5 are the optimal combination for mapping soil salinity at the 30–40 cm depth in the study area (with a CC value of 0.69 and a root mean squared error value of 0.38). Our results indicated that the similarity-based prediction approach could be a vital alternative to other methods for mapping soil salinity, especially for area with limited observation data and could be used to monitor soil salinity distributions in the future.