Search

The InterPro database (https://www.ebi.ac.uk/interpro/) provides an integrative classification of protein sequences into families, and identifies functionally important domains and conserved sites. InterProScan is the underlying software that allows protein and nucleic acid sequences to be searched against InterPro's signatures. Signatures are predictive models which describe protein families, domains or sites, and are provided by multiple databases. InterPro combines signatures representing equivalent families, domains or sites, and provides additional information such as descriptions, literature references and Gene Ontology (GO) terms, to produce a comprehensive resource for protein classification. Founded in 1999, InterPro has become one of the most widely used resources for protein family annotation. Here, we report the status of InterPro (version 81.0) in its 20th year of operation, and its associated software, including updates to database content, the release of a new website and REST API, and performance improvements in InterProScan.

iPTMnet is a bioinformatics resource that integrates protein post-translational modification (PTM) data from text mining and curated databases and ontologies to aid in knowledge discovery and scientific study. The current iPTMnet website can be used for querying and browsing rich PTM information but does not support automated iPTMnet data integration with other tools. Hence, we have developed a RESTful API utilizing the latest developments in cloud technologies to facilitate the integration of iPTMnet into existing tools and pipelines. We have packaged iPTMnet API software in Docker containers and published it on DockerHub for easy redistribution. We have also developed Python and R packages that allow users to integrate iPTMnet for scientific discovery, as demonstrated in a use case that connects PTM sites to kinase signaling pathways.

Significant progress has been made in applying deep learning on natural language processing tasks recently. However, deep learning models typically require a large amount of annotated training data while often only small labeled datasets are available for many natural language processing tasks in biomedical literature. Building large-size datasets for deep learning is expensive since it involves considerable human effort and usually requires domain expertise in specialized fields. In this work, we consider augmenting manually annotated data with large amounts of data using distant supervision. However, data obtained by distant supervision is often noisy, we first apply some heuristics to remove some of the incorrect annotations. Then using methods inspired from transfer learning, we show that the resulting models outperform models trained on the original manually annotated sets.

Open information extraction is an important task in Biomedical domain. The goal of the OpenIE is to automatically extract structured information from unstructured text with no or little supervision. It aims to extract all the relation tuples from the corpus without requiring pre-specified relation types. The existing tools may extract ill-structured or incomplete information, or fail on the Biomedical literature due to the long and complicated sentences. In this paper, we propose a novel pattern-based information extraction method for the wide-window entities (WW-PIE). WW-PIE utilizes dependency parsing to break down the long sentences first and then utilizes frequent textual patterns to extract the high-quality information. The pattern hierarchical grouping organize and structure the extractions to be straightforward and precise. Consequently, comparing with the existing OpenIE tools, WW-PIE produces structured output that can be directly used for downstream applications. The proposed WW-PIE is also capable in extracting n-ary and nested relation structures, which is less studied in the existing methods. Extensive experiments on real-world biomedical corpus from PubMed abstracts demonstrate the power of WW-PIE at extracting precise and well-structured information.

Background The Gene Ontology (GO) is a resource that supplies information about gene product function using ontologies to represent biological knowledge. These ontologies cover three domains: Cellular Component (CC), Molecular Function (MF), and Biological Process (BP). GO annotation is a process which assigns gene functional information using GO terms to relevant genes in the literature. It is a common task among the Model Organism Database (MOD) groups. Manual GO annotation relies on human curators assigning gene functional information using GO terms by reading the biomedical literature. This process is very time-consuming and labor-intensive. As a result, many MODs can afford to curate only a fraction of relevant articles. Methods GO terms from the CC domain can be essentially divided into two sub-hierarchies: subcellular location terms, and protein complex terms. We cast the task of gene annotation using GO terms from the CC domain as relation extraction between gene and other entities: (1) extract cases where a protein is found to be in a subcellular location, and (2) extract cases where a protein is a subunit of a protein complex. For each relation extraction task, we use an approach based on triggers and syntactic dependencies to extract the desired relations among entities. Results We tested our approach on the BC4GO test set, a publicly available corpus for GO annotation. Our approach obtains a F1-score of 71%, a precision of 91% and a recall of 58% for predicting GO terms from CC Domain for given genes. Conclusions We have described a novel approach of treating gene annotation with GO terms from CC domain as two relation extraction subtasks. Evaluation results show that our approach achieves a F1-score of 71% for predicting GO terms for given genes. Thereby our approach can be used to accelerate the process of GO annotation for the bio-annotators.

Many biomedical entity mentions contain other entity mentions nested inside. Most current named entity recognition (NER) systems deal with only flat entities and ignore such nested entities, which may introduce errors to subsequent tasks such as relation extraction and knowledge base completion. Recently, fully supervised methods are proposed for nested named entity recognition. Despite their success on benchmark datasets, supervised methods rely on human annotation and lead to highly specialized systems that cannot be easily adapted to new entity types. In this study, we propose PENNER, a novel and effective pattern-enhanced nested named entity recognition method that relies on massive corpora plus only very weak supervision. We compare PENNER with a state-of-the-art BioNER system, PubTator, and observe great improvement at recognizing genes, chemicals, diseases and species. PENNER can also accurately extract new types of entities, such as biological process and treatment, that are not annotated by PubTator.

We have developed a cloud-based (AWS and IBM SoftLayer) knowledge environment for scalable semantic mining of scientific literature and PTM integrative knowledge discovery in precision medicine, building upon our novel natural language processing (NLP) technologies and bioinformatics infrastructure. We provided semantic integration of full-scale PubMed mining results from disparate text mining tools, along with kinase-substrate data from iPTMnet, and PTM proteoforms and their relations from Protein Ontology (PRO). We shared the digital objects of those applications in multiple interoperable formats and have registered them in bioCADDIE using CEDAR. We experimented with multiple system setups using operating system, programming language, web server, or database server that best fits each application. We evaluated the cost effectiveness of cloud computing by only paying for what we use and readily experimenting with additional services. A web portal is available for accessing our cloud-based knowledge environment at https://proteininformationresource.org/cloud/.

Protein complex plays an essential role in cellular functions and is an important named entity in the biomedical field. Since protein complex –relevant experimental results are usually published in scientific articles, recognizing protein complex mentions from literature is a crucial step of discovering protein complex-related information from existing scientific research studies. In this paper, we propose a method for protein complex mention recognition, which applies knowledge automatically learned from PubMed. Evaluation shows our method achieves a F1-score of 81%, demonstrating its effectiveness in the protein complex recognition task.

Prediction of protein-protein interaction (PPI) remains a central task in systems biology. With more PPIs identified, forming PPI networks, it has become feasible and also imperative to study PPIs at the network level, such as evolutionary analysis of the networks, for better understanding of PPI networks and for more accurate prediction of pairwise PPIs by leveraging the information gained at the network level. In this work we developed a novel method that enables us to incorporate evolutionary information into geometric space to improve PPI prediction, which in turn can be used to select and evaluate various evolutionary models. The method is tested with cross-validation using human PPI network and yeast PPI network data. The results show that the accuracy of PPI prediction measured by ROC score is increased by up to 14.6%, as compared to a baseline without using evolutionary information. The results also indicate that our modified evolutionary model DANEOsf-combining a gene duplication/neofunctionalization model and scale-free model-has a better fitness and prediction efficacy for these two PPI networks. The improved PPI prediction performance may suggest that our DANEOsf evolutionary model can uncover the underlying evolutionary mechanism for these two PPI networks better than other tested models. Consequently, of particular importance is that our method offers an effective way to select evolutionary models that best capture the underlying evolutionary mechanisms, evaluating the fitness of evolutionary models from the perspective of PPI prediction on real PPI networks.

The InterPro database (http://www.ebi.ac.uk/interpro/) integrates together predictive models or 'signatures' representing protein domains, families and functional sites from multiple, diverse source databases: Gene3D, PANTHER, Pfam, PIRSF, PRINTS, ProDom, PROSITE, SMART, SUPERFAMILY and TIGRFAMs. Integration is performed manually and approximately half of the total approximately 58,000 signatures available in the source databases belong to an InterPro entry. Recently, we have started to also display the remaining un-integrated signatures via our web interface. Other developments include the provision of non-signature data, such as structural data, in new XML files on our FTP site, as well as the inclusion of matchless UniProtKB proteins in the existing match XML files. The web interface has been extended and now links out to the ADAN predicted protein-protein interaction database and the SPICE and Dasty viewers. The latest public release (v18.0) covers 79.8% of UniProtKB (v14.1) and consists of 16 549 entries. InterPro data may be accessed either via the web address above, via web services, by downloading files by anonymous FTP or by using the InterProScan search software (http://www.ebi.ac.uk/Tools/InterProScan/).

Distant supervision has been applied to automatically generate labeled data for biomedical relation extraction. Noise exists in both positively and negatively-labeled data and affects the performance of supervised machine learning methods. In this paper, we propose three novel heuristics based on the notion of proximity, trigger word and confidence of patterns to leverage lexical and syntactic information to reduce the level of noise in the distantly labeled data. Experiments on three different tasks, extraction of protein-protein-interaction, miRNA-gene regulation relation and protein-localization event, show that the proposed methods can improve the F-score over the baseline by 6, 10 and 14 points for the three tasks, respectively. We also show that when the models are configured to output high-confidence results, high precisions can be obtained using the proposed methods, making them promising for facilitating manual curation for databases.

Comparison sentences are very commonly used by authors in biomedical literature to report results of experiments. In such comparisons, authors typically make observations under two different scenarios. In this paper, we present a system to automatically identify such comparative sentences and their components i.e. the compared entities, the scale of the comparison and the aspect on which the entities are being compared. Our methodology is based on dependencies obtained by applying a parser to extract a wide range of comparison structures. We evaluated our system for its effectiveness in identifying comparisons and their components. The system achieved a F-score of 0.87 for comparison sentence identification and 0.77-0.81 for identifying its components.

Identifying the residues in a protein that are involved in protein-protein interaction and identifying the contact matrix for a pair of interacting proteins are two computational tasks at different levels of an in-depth analysis of protein-protein interaction. Various methods for solving these two problems have been reported in the literature. However, the interacting residue prediction and contact matrix prediction were handled by and large independently in those existing methods, though intuitively good prediction of interacting residues will help with predicting the contact matrix. In this work, we developed a novel protein interacting residue prediction system, contact matrix-interaction profile hidden Markov model (CM-ipHMM), with the integration of contact matrix prediction and the ipHMM interaction residue prediction. We propose to leverage what is learned from the contact matrix prediction and utilize the predicted contact matrix as “feedback” to enhance the interaction residue prediction. The CM-ipHMM model showed significant improvement over the previous method that uses the ipHMM for predicting interaction residues only. It indicates that the downstream contact matrix prediction could help the interaction site prediction.

BioC is a simple XML format for text, annotations and relations, and was developed to achieve interoperability for biomedical text processing. Following the success of BioC in BioCreative IV, the BioCreative V BioC track addressed a collaborative task to build an assistant system for BioGRID curation. In this paper, we describe the framework of the collaborative BioC task and discuss our findings based on the user survey. This track consisted of eight subtasks including gene/protein/organism named entity recognition, protein-protein/genetic interaction passage identification and annotation visualization. Using BioC as their data-sharing and communication medium, nine teams, world-wide, participated and contributed either new methods or improvements of existing tools to address different subtasks of the BioC track. Results from different teams were shared in BioC and made available to other teams as they addressed different subtasks of the track. In the end, all submitted runs were merged using a machine learning classifier to produce an optimized output. The biocurator assistant system was evaluated by four BioGRID curators in terms of practical usability. The curators' feedback was overall positive and highlighted the user-friendly design and the convenient gene/protein curation tool based on text mining.Database URL: http://www.biocreative.org/tasks/biocreative-v/track-1-bioc/.

Protein-Protein interaction (PPI) prediction is a central task in achieving a better understanding of cellular and intracellular processes. Because high-throughput experimental methods are both expensive and time-consuming, and are also known of suffering from the problems of incompleteness and noise, many computational methods have been developed, with varied degrees of success. However, the inference of PPI network from multiple heterogeneous data sources remains a great challenge. In this work, we developed a novel method based on Approximate Bayesian Computation and modified Differential Evolution sampling (ABC-DEP) and Regularized Laplacian (RL) kernel. The method enables inference of PPI networks from topological properties and multiple heterogeneous features including gene expression and Pfam domain profiles, in forms of weighted kernels. The optimal weights are obtained by ABC-DEP, and the kernel fusion built based on optimal weights serves as input to RL to infer missing or new edges in the PPI network. Detailed comparisons with control methods have been made, and the results show that the accuracy of PPI prediction measured by AUC is increased by up to 23%, as compared to a base-line without using optimal weights. The method can provide insights into the relations between PPIs and various feature kernels and demonstrates strong capability of predicting far-away interactions that cannot be well detected by traditional RL method.

Reconstruction of PPI networks is a central task in systems biology, and inference from multiple heterogeneous data sources offers a promising computational approach to making de novo PPI prediction by leveraging complementary information and the partial network structure. However, how to quickly learn weights for heterogeneous data sources remains a challenge. In this work, we developed a method to infer de novo PPIs by combining multiple data sources represented in kernel format and obtaining optimal weights based on random walk over the existing partial network. Our proposed method utilizes Baker algorithm and the training data to construct a transition matrix which constrains how a random walk would traverse the partial network. Multiple heterogeneous features for the proteins in the network, including gene expression and Pfam domain profiles, are then combined into the form of a weighted kernel, which provides a new “adjacency matrix” for the whole network but is required to comply with the transition matrix on the part of the training subnetwork. This requirement is met by adjusting the weights to minimize the element-wise difference between the transition matrix and the weighted kernel. The minimization problem is solved by linear programming. The weighted kernel is then transformed to regularized Laplacian (RL) kernel to infer missing or new edges in the PPI network. The results on synthetic data and real data from Yeast show that the accuracy of PPI prediction measured by AUC is increased by up to 19% as compared to a control method without using optimal weights. Moreover, the weights learned by our method Weight Optimization by Linear Programming (WOLP) are very consistent with that learned by sampling, and can provide insights into the relations between PPIs and various feature kernels, thereby improving PPI prediction.

Objective: Scientific findings regarding human pathogens and their host responses are buried in the growing volume of biomedical literature and there is an urgent need to mine information pertaining to pathogenesis-related proteins especially host pathogen protein-protein interactions (HP-PPIs) from literature. Methods: In this paper, we report our exploration of developing an automated system to identify MEDLINE abstracts referring to HP-PPIs. An annotated corpus consisting of 1360 MEDLINE abstracts was generated. With this corpus, we developed and evaluated document classification systems using support vector machines (SVMs). We also investigated the effects of three feature selection methods:information gain (IG), @g^2 test, and specific mutual information (SI). The performance was measured using normalized discounted cumulative gain (NDCG) and positive predictive value (PPV) and all measures were obtained through 10-fold cross validation. Results: NDCG measures for classification systems using all features or a subset of features selected using IG and @g^2 test range from 0.83 to 0.89 while classification systems built based on features selected using SI had relatively lower NDCG measures. The classification system achieved a PPV of 50.7% for the top 10% ranked documents comparing to a baseline PPV of 10.0%. Conclusions: Our results indicate that document classification systems can be constructed to efficiently retrieve HP-PPI related documents. Feature selection was effective in reducing the dimensionality of features to build a compact system.

Tumor molecular profiling plays an integral role in identifying genomic anomalies which may help in personalizing cancer treatments, improving patient outcomes and minimizing risks associated with different therapies. However, critical information regarding the evidence of clinical utility of such anomalies is largely buried in biomedical literature. It is becoming prohibitive for biocurators, clinical researchers and oncologists to keep up with the rapidly growing volume and breadth of information, especially those that describe therapeutic implications of biomarkers and therefore relevant for treatment selection. In an effort to improve and speed up the process of manually reviewing and extracting relevant information from literature, we have developed a natural language processing (NLP)-based text mining (TM) system called eGARD (extracting Genomic Anomalies association with Response to Drugs). This system relies on the syntactic nature of sentences coupled with various textual features to extract relations between genomic anomalies and drug response from MEDLINE abstracts. Our system achieved high precision, recall and F-measure of up to 0.95, 0.86 and 0.90, respectively, on annotated evaluation datasets created in-house and obtained externally from PharmGKB. Additionally, the system extracted information that helps determine the confidence level of extraction to support prioritization of curation. Such a system will enable clinical researchers to explore the use of published markers to stratify patients upfront for ‘best-fit’ therapies and readily generate hypotheses for new clinical trials.

Objectives: Biomedical named entity recognition (BNER) is a critical component in automated systems that mine biomedical knowledge in free text. Among different types of entities in the domain, gene/protein would be the most studied one for BNER. Our goal is to develop a gene/protein name recognition system BioTagger-GM that exploits rich information in terminology sources using powerful machine learning frameworks and system combination. Design: BioTagger-GM consists of four main components: (1) dictionary lookup— gene/protein names in BioThesaurus and biomedical terms in UMLS Metathesaurus are tagged in text, (2) machine learning—machine learning systems are trained using dictionary lookup results as one type of feature, (3) post-processing— heuristic rules are used to correct recognition errors, and (4) system combination—a voting scheme is used to combine recognition results from multiple systems. Measurements: The BioCreAtIvE II Gene Mention (GM) corpus was used to evaluate the proposed method. To test its general applicability, the method was also evaluated on the JNLPBA corpus modified for gene/protein name recognition. The performance of the systems was evaluated through cross-validation tests and measured using precision, recall, and F-Measure. Results: BioTagger-GM achieved an F-Measure of 0.8887 on the BioCreAtIvE II GM corpus, which is higher than that of the first-place system in the BioCreAtIvE II challenge. The applicability of the method was also confirmed on the modified JNLPBA corpus. Conclusion: The results suggest that terminology sources, powerful machine learning frameworks, and system combination can be integrated to build an effective BNER system.

InterPro is an integrated resource for protein families, domains and functional sites, which integrates the following protein signature databases: PROSITE, PRINTS, ProDom, Pfam, SMART, TIGRFAMs, PIRSF, SUPERFAMILY, Gene3D and PANTHER. The latter two new member databases have been integrated since the last publication in this journal. There have been several new developments in InterPro, including an additional reading field, new database links, extensions to the web interface and additional match XML files. InterPro has always provided matches to UniProtKB proteins on the website and in the match XML file on the FTP site. Additional matches to proteins in UniParc (UniProt archive) are now available for download in the new match XML files only. The latest InterPro release (13.0) contains more than 13 000 entries, covering over 78% of all proteins in UniProtKB. The database is available for text- and sequence-based searches via a webserver (http://www.ebi.ac.uk/interpro), and for download by anonymous FTP (ftp://ftp.ebi.ac.uk/pub/databases/interpro). The InterProScan search tool is now also available via a web service at http://www.ebi.ac.uk/Tools/webservices/WSInterProScan.html.

We introduce RLIMS-P version 2.0, an enhanced rule-based information extraction (IE) system for mining kinase, substrate, and phosphorylation site information from scientific literature. Consisting of natural language processing and IE modules, the system has integrated several new features, including the capability of processing full-text articles and generalizability towards different post-translational modifications (PTMs). To evaluate the system, sets of abstracts and full-text articles, containing a variety of textual expressions, were annotated. On the abstract corpus, the system achieved F-scores of 0.91, 0.92, and 0.95 for kinases, substrates, and sites, respectively. The corresponding scores on the full-text corpus were 0.88, 0.91, and 0.92. It was additionally evaluated on the corpus of the 2013 BioNLP-ST GE task, and achieved an F-score of 0.87 for the phosphorylation core task, improving upon the results previously reported on the corpus. Full-scale processing of all abstracts in MEDLINE and all articles in PubMed Central Open Access Subset has demonstrated scalability for mining rich information in literature, enabling its adoption for biocuration and for knowledge discovery. The new system is generalizable and it will be adapted to tackle other major PTM types. RLIMS-P 2.0 online system is available online (http://proteininformationresource.org/rlimsp/) and the developed corpora are available from iProLINK (http://proteininformationresource.org/iprolink/).

Background Automatically detecting gene/protein names in the literature and connecting them to databases records, also known as gene normalization, provides a means to structure the information buried in free-text literature. Gene normalization is critical for improving the coverage of annotation in the databases, and is an essential component of many text mining systems and database curation pipelines. Methods In this manuscript, we describe a gene normalization system specifically tailored for plant species, called pGenN (pivot-based Gene Normalization). The system consists of three steps: dictionary-based gene mention detection, species assignment, and intra species normalization. We have developed new heuristics to improve each of these phases. Results We evaluated the performance of pGenN on an in-house expertly annotated corpus consisting of 104 plant relevant abstracts. Our system achieved an F-value of 88.9% (Precision 90.9% and Recall 87.2%) on this corpus, outperforming state-of-art systems presented in BioCreative III. We have processed over 440,000 plant-related Medline abstracts using pGenN. The gene normalization results are stored in a local database for direct query from the pGenN web interface (proteininformationresource.org/pgenn/). The annotated literature corpus is also publicly available through the PIR text mining portal (proteininformationresource.org/iprolink/).

Kernel-based methods are widely used for relation extraction task and obtain good results by leveraging lexical and syntactic information. However, in biomedical domain these methods are limited by the size of dataset and have difficulty in coping with variations in text. To address this problem, we propose Extended Dependency Graph (EDG) by incorporating a few simple linguistic ideas and include information beyond syntax. We believe the use of EDG will enable machine learning methods to generalize more easily. Experiments confirm that EDG provides up to 10% f-value improvement over dependency graph using mainstream kernel methods over five corpora. We conducted additional experiments to provide a more detailed analysis of the contributions of individual modules in EDG construction.

Motivation: Our purpose is to develop a statistical modeling approach for cancer biomarker discovery and provide new insights into early cancer detection. We propose the concept of dependence network, apply it for identifying cancer biomarkers, and study the difference between the protein or gene samples from cancer and non-cancer subjects based on mass-spectrometry (MS) and microarray data.Results: Three MS and two gene microarray datasets are studied. Clear differences are observed in the dependence networks for cancer and non-cancer samples. Protein/gene features are examined three at one time through an exhaustive search. Dependence networks are constructed by binding triples identified by the eigenvalue pattern of the dependence model, and are further compared to identify cancer biomarkers. Such dependence-network-based biomarkers show much greater consistency under 10-fold cross-validation than the classification-performance-based biomarkers. Furthermore, the biological relevance of the dependence-network-based biomarkers using microarray data is discussed. The proposed scheme is shown promising for cancer diagnosis and prediction.Availability: See supplements:Contact: qiupeng@umd.edu

Protein phosphorylation is a reversible post-translational modification where a protein kinase adds a phosphate group to a protein, potentially regulating its function, localization and/or activity. Phosphorylation can affect protein–protein interactions (PPIs), abolishing interaction with previous binding partners or enabling new interactions. Extracting phosphorylation information coupled with PPI information from the scientific literature will facilitate the creation of phosphorylation interaction networks of kinases, substrates and interacting partners, toward knowledge discovery of functional outcomes of protein phosphorylation. Increasingly, PPI databases are interested in capturing the phosphorylation state of interacting partners. We have previously developed the eFIP (Extracting Functional Impact of Phosphorylation) text mining system, which identifies phosphorylated proteins and phosphorylation-dependent PPIs. In this work, we present several enhancements for the eFIP system: (i) text mining for full-length articles from the PubMed Central open-access collection; (ii) the integration of the RLIMS-P 2.0 system for the extraction of phosphorylation events with kinase, substrate and site information; (iii) the extension of the PPI module with new trigger words/phrases describing interactions and (iv) the addition of the iSimp tool for sentence simplification to aid in the matching of syntactic patterns. We enhance the website functionality to: (i) support searches based on protein roles (kinases, substrates, interacting partners) or using keywords; (ii) link protein entities to their corresponding UniProt identifiers if mapped and (iii) support visual exploration of phosphorylation interaction networks using Cytoscape. The evaluation of eFIP on full-length articles achieved 92.4% precision, 76.5% recall and 83.7% F-measure on 100 article sections. To demonstrate eFIP for knowledge extraction and discovery, we constructed phosphorylation-dependent interaction networks involving 14-3-3 proteins identified from cancer-related versus diabetes-related articles. Comparison of the phosphorylation interaction network of kinases, phosphoproteins and interactants obtained from eFIP searches, along with enrichment analysis of the protein set, revealed several shared interactions, highlighting common pathways discussed in the context of both diseases. Database URL: http://proteininformationresource.org/efip

The exponential growth of large-scale molecular sequence data and of the PubMed scientific literature has prompted active research in biological literature mining and information extraction to facilitate genome/proteome annotation and improve the quality of biological databases. Motivated by the promise of text mining methodologies, but at the same time, the lack of adequate curated data for training and benchmarking, the Protein Information Resource (PIR) has developed a resource for protein literature mining--iProLINK (integrated Protein Literature INformation and Knowledge). As PIR focuses its effort on the curation of the UniProt protein sequence database, the goal of iProLINK is to provide curated data sources that can be utilized for text mining research in the areas of bibliography mapping, annotation extraction, protein named entity recognition, and protein ontology development. The data sources for bibliography mapping and annotation extraction include mapped citations (PubMed ID to protein entry and feature line mapping) and annotation-tagged literature corpora. The latter includes several hundred abstracts and full-text articles tagged with experimentally validated post-translational modifications (PTMs) annotated in the PIR protein sequence database. The data sources for entity recognition and ontology development include a protein name dictionary, word token dictionaries, protein name-tagged literature corpora along with tagging guidelines, as well as a protein ontology based on PIRSF protein family names. iProLINK is freely accessible at http://pir.georgetown.edu/iprolink, with hypertext links for all downloadable files.

Increasingly, scientists have begun to tackle gene functions and other complex regulatory processes by studying organisms at the global scales for various levels of biological organization, ranging from genomes to metabolomes and physiomes. Meanwhile, new bioinformatics methods have been developed for inferring protein function using associative analysis of functional properties to complement the traditional sequence homology-based methods. To fully exploit the value of the high-throughput system biology data and to facilitate protein functional studies requires bioinformatics infrastructures that support both data integration and associative analysis. The iProClass database, designed to serve as a framework for data integration in a distributed networking environment, provides comprehensive descriptions of all proteins, with rich links to over 50 databases of protein family, function, pathway, interaction, modification, structure, genome, ontology, literature, and taxonomy. In particular, the database is organized with PIRSF family classification and maps to other family, function, and structure classification schemes. Coupled with the underlying taxonomic information for complete genomes, the iProClass system (http://pir.georgetown.edu/iproclass/) supports associative studies of protein family, domain, function, and structure. A case study of the phosphoglycerate mutases illustrates a systematic approach for protein family and phylogenetic analysis. Such studies may serve as a basis for further analysis of protein functional evolution, and its relationship to the co-evolution of metabolic pathways, cellular networks, and organisms.

With the accelerated accumulation of genomic sequence data, there is a pressing need to develop computational methods and advanced bioinformatics infrastructure for reliable and large-scale protein annotation and biological knowledge discovery. The Protein Information Resource (PIR) provides an integrated public resource of protein informatics to support genomic and proteomic research. PIR produces the Protein Sequence Database of functionally annotated protein sequences. The annotation problems are addressed by a classification-driven and rule-based method with evidence attribution, coupled with an integrated knowledge base system being developed. The approach allows sensitive identification, consistent and rich annotation, and systematic detection of annotation errors, as well as distinction of experimentally verified and computationally predicted features. The knowledge base consists of two new databases, sequence analysis tools, and graphical interfaces. PIR-NREF, a non-redundant reference database, provides a timely and comprehensive collection of all protein sequences, totaling more than 1,000,000 entries. iProClass, an integrated database of protein family, function, and structure information, provides extensive value-added features for about 830,000 proteins with rich links to over 50 molecular databases. This paper describes our approach to protein functional annotation with case studies and examines common identification errors. It also illustrates that data integration in PIR supports exploration of protein relationships and may reveal protein functional associations beyond sequence homology.

We review recent results in literature data mining for biology and discuss the need and the steps for a challenge evaluation for this field. Literature data mining has progressed from simple recognition of terms to extraction of interaction relationships from complex sentences, and has broadened from recognition of protein interactions to a range of problems such as improving homology search, identifying cellular location, and so on. To encourage participation and accelerate progress in this expanding field, we propose creating challenge evaluations, and we describe two specific applications in this context. Contact: wuc@georgetown.edutsujii@is.s.u-tokyo.ac.jppark@nlp.kaist.ac.krlynette@mitre.orglimsoon@lit.org.sg * To whom correspondence should be addressed.

Protein-protein interactions (PPIs) play essential roles in many biological processes. The large gap between the fast growing protein sequence information and the experimentally verified PPI information has motivated the development of computational methods to predict PPIs and residue-residue (RR) contacts. Although many tools are available for PPI predictions, most do not make a distinction between direct PPIs, which refer to physical contacts between two proteins, and indirect PPIs. Furthermore, few tools predict detailed RR contacts between interacting proteins, which are valuable for understanding chemical interaction mechanisms and designing investigations of the interaction. Here we present the web server DDI2PPI to provide PPI and RR contact matrix predictions from query protein sequences, focusing on predicting direct physical interactions. DDI2PPI implements algorithms that predict PPIs and RR contact matrices through domain-domain interaction (DDI) analysis based on interaction profile Hidden Markov Models (ipHMMs) and Support Vector Machines (SVMs). It covers more than 6000 DDI families in 3DID, a database of three-dimensional interacting domains. DDI2PPI (version 1.0) is freely available at http://annotation.dbi.udel.edu/ppi_prediction/.

BioCreative: Critical Assessment of Information Extraction in Biology is an international community-wide effort for evaluating text mining (TM) and information extraction systems applied to the biological domain (http://www.biocreative.org/).The Challenge Evaluations and the accompanying BioCreative Workshops bring together the TM and biology communities to drive the development of practically relevant TM systems. One of the main goals of this initiative is that the resulting systems facilitate a more efficient literature information access to biologists in general, but also provide tools that can be directly integrated into the biocuration workflow and the knowledge discovery process carried out by databases. Beyond addressing the current barriers faced by TM technologies applied to biological literature, BioCreative has further been conducting user requirement analyses, user-based evaluations and fostering standards development for TM tool reuse and integration. This DATABASE virtual issue captures the major results from the Fourth BioCreative Challenge Evaluation Workshop, and is the sixth special issue devoted to BioCreative. Built on the success of the previous Challenge Evaluations and Workshops (BioCreative I, II, II.5, III, 2012) (1–5), the BioCreative IV Workshop was held in Bethesda, MD, on October 7–9, 2013.

When compared to Sanger sequencing technology, next-generation sequencing (NGS) technologies are hindered by shorter sequence read length, higher base-call error rate, non-uniform coverage, and platform-specific sequencing artifacts. These characteristics lower the quality of their downstream analyses, e.g. de novo and reference-based assembly, by introducing sequencing artifacts and errors that may contribute to incorrect interpretation of data. Although many tools have been developed for quality control and pre-processing of NGS data, none of them provide flexible and comprehensive trimming options in conjunction with parallel processing to expedite pre-processing of large NGS datasets. We developed ngsShoRT (next-generation sequencing Short Reads Trimmer), a flexible and comprehensive open-source software package written in Perl that provides a set of algorithms commonly used for pre-processing NGS short read sequences. We compared the features and performance of ngsShoRT with existing tools: CutAdapt, NGS QC Toolkit and Trimmomatic. We also compared the effects of using pre-processed short read sequences generated by different algorithms on de novo and reference-based assembly for three different genomes: Caenorhabditis elegans, Saccharomyces cerevisiae S288c, and Escherichia coli O157 H7. Several combinations of ngsShoRT algorithms were tested on publicly available Illumina GA II, HiSeq 2000, and MiSeq eukaryotic and bacteria genomic short read sequences with the focus on removing sequencing artifacts and low-quality reads and/or bases. Our results show that across three organisms and three sequencing platforms, trimming improved the mean quality scores of trimmed sequences. Using trimmed sequences for de novo and reference-based assembly improved assembly quality as well as assembler performance. In general, ngsShoRT outperformed comparable trimming tools in terms of trimming speed and improvement of de novo and reference-based assembly as measured by assembly contiguity and correctness. Trimming of short read sequences can improve the quality of de novo and reference-based assembly and assembler performance. The parallel processing capability of ngsShoRT reduces trimming time and improves the memory efficiency when dealing with large datasets. We recommend combining sequencing artifacts removal, and quality score based read filtering and base trimming as the most consistent method for improving sequence quality and downstream assemblies. ngsShoRT source code, user guide and tutorial are available at http://research.bioinformatics.udel.edu/genomics/ngsShoRT/ . ngsShoRT can be incorporated as a pre-processing step in genome and transcriptome assembly projects.

With the exponential accumulation of sequence data, continued progress in the Human Genome Project will depend increasingly on advanced computational tools to manage and analyze the data. Utilizing information embedded within families of homologous sequences, a gene family identification approach may facilitate the understanding of gene functions. We have developed a GeneFIND (Gene Family Identification Network Design) system for database searching against gene families. It provides rapid and accurate protein family identification by combining global and motif sequence similarities and incorporating ProClass family information. Multi-level filters are used, starting with the MOTIFIND neural networks and BLAST search, followed by SSEARCH alignment, motif pattern match, hidden Markov modeling of motifs and ClustalW motif alignment. GeneFIND has been implemented as a full-scale system for the classification of more than 1200 ProSite and 6000 PIR families. It has been used to identify thousands of new family members and is well suited for genomic sequence analysis. The system is available for on-line family identification from our WWW server ().

Artificial neural networks provide a unique computing architecture whose potential has attracted interest from researchers across different disciplines. As a technique for computational analysis, neural network technology is very well suited for the analysis of molecular sequence data. It has been applied successfully to a variety of problems, ranging from gene identification, to protein structure prediction and sequence classification. This article provides an overview of major neural network paradigms, discusses design issues, and reviews current applications in DNA/RNA and protein sequence analysis.

Proteins are complex biological polymers that mediate virtually all cellular functions. Typically these functions are modulated by protein-protein interactions (PPI). Tremendous efforts have been made by life scientists to detect PPIs through different experimental approaches and document the results through publications. On the informatics front, however, there lacks an effective means for retrieving PPI information from published literatures. In this work we present a novel framework for identifying experimental methods employed for analyzing PPI from biomedical articles. Different from state-of-the-art approaches based only on text, we explore using the combination of attributes from figures, figure captions, and text within figures for identifying PPI experimental methods. Our work is motivated by the observation that biomedical figures often constitute direct evidence of experimental results and therefore provide complementary information to texts. We start with automatically extracting unimodal panels (subfigures) and their associated subcaptions and then classifying the subfigure into different types using a proposed hierarchical image taxonomy. Next, we combine the subfigure types with text-based features to form a hybrid feature descriptor and use it for PPI method classification. We further construct a dataset starting from a set of 2,256 documents provided by the molecular interaction database MINT. Here we show that our new approach outperforms the text-only solution for associating figures with PPI methods.

Literature-based annotation of protein phosphorylation is the focus of many biological databases, as phosphorylation is a global regulator of cellular activity. To speed up manual curation of phosphorylation information, text mining technology has been utilized. In this paper, we report our ongoing effort to enhance RLIMS-P, a rule-based information extraction (IE) system to identify protein phosphorylation information in scientific literature. Despite the high accuracy attained by RLIMS-P, the use of elaborated patterns and rules resulted in a substantial effort for system development and maintenance. To mitigate this challenge, we redesigned RLIMS-P and integrated new natural language processing (NLP) techniques. It has also been adapted to mine full-text articles and generalized to be able to exploit common features for different post-translational modifications (PTMs). The updated RLIMS-P (version 2.0) was evaluated on abstracts in the publicly available BioNLP GENIA event extraction (GE) corpus, and achieved F-scores of 0.92 and 0.96 for phosphorylation substrate and site, respectively. On a full-text corpus developed in-house, it achieved F-scores of 0.91 and 0.92 for substrate and site, and 0.88 for kinase. The system was applied to the PubMed Central (PMC) Open Access Subset, and promising results have been obtained in mining the full-text articles. RLIMS-P focuses on protein phosphorylation information, but its new design would be generalizable for other PTM types. RLIMS-P version 2.0 is available at: http://proteininformationresource.org/rlimsp/.

Motivation: Prediction of protein–protein interaction has become an important part of systems biology in reverse engineering the biological networks for better understanding the molecular biology of the cell. Although significant progress has been made in terms of prediction accuracy, most computational methods only predict whether two proteins interact but not their interacting residues—the information that can be very valuable for understanding the interaction mechanisms and designing modulation of the interaction. In this work, we developed a computational method to predict the interacting residue pairs—contact matrix for interacting protein domains, whose rows and columns correspond to the residues in the two interacting domains respectively and whose values (1 or 0) indicate whether the corresponding residues (do or do not) interact. Results: Our method is based on supervised learning using support vector machines. For each domain involved in a given domain–domain interaction (DDI), an interaction profile hidden Markov model (ipHMM) is first built for the domain family, and then each residue position for a member domain sequence is represented as a 20-dimension vector of Fisher scores, characterizing how similar it is as compared with the family profile at that position. Each element of the contact matrix for a sequence pair is now represented by a feature vector from concatenating the vectors of the two corresponding residues, and the task is to predict the element value (1 or 0) from the feature vector. A support vector machine is trained for a given DDI, using either a consensus contact matrix or contact matrices for individual sequence pairs, and is tested by leave-one-out cross validation. The performance averaged over a set of 115 DDIs collected from the 3 DID database shows significant improvement (sensitivity up to 85%, and specificity up to 85%), as compared with a multiple sequence alignment-based method (sensitivity 57%, and specificity 78%) previously reported in the literature. Contact: lliao@cis.udel.edu or wuc@cis.udel.edu Supplementary information: Supplementary data are available at Bioinformatics online.

Knowledge representation of the role of phosphorylation is essential for the meaningful understanding of many biological processes. However, such a representation is challenging because proteins can exist in numerous phosphorylated forms with each one having its own characteristic protein–protein interactions (PPIs), functions and subcellular localization. In this article, we evaluate the current state of phosphorylation event curation and then present a bioinformatics framework for the annotation and representation of phosphorylated proteins and construction of phosphorylation networks that addresses some of the gaps in current curation efforts. The integrated approach involves (i) text mining guided by RLIMS-P, a tool that identifies phosphorylation-related information in scientific literature; (ii) data mining from curated PPI databases; (iii) protein form and complex representation using the Protein Ontology (PRO); (iv) functional annotation using the Gene Ontology (GO); and (v) network visualization and analysis with Cytoscape. We use this framework to study the spindle checkpoint, the process that monitors the assembly of the mitotic spindle and blocks cell cycle progression at metaphase until all chromosomes have made bipolar spindle attachments. The phosphorylation networks we construct, centered on the human checkpoint kinase BUB1B (BubR1) and its yeast counterpart MAD3, offer a unique view of the spindle checkpoint that emphasizes biologically relevant phosphorylated forms, phosphorylation-state–specific PPIs and kinase–substrate relationships. Our approach for constructing protein phosphorylation networks can be applied to any biological process that is affected by phosphorylation. Database URL: http://www.yeastgenome.org/

Gallus Reactome provides a manually curated reference database for chicken signaling and metabolic pathways. Providing such a high-quality reference database requires significant time input from both curators and pathway experts. This paper describes the application of orthology prediction, computational linguistic tools including text mining and information extraction applications, along with resources available from the Protein Information Resource, to facilitate pathway curation. In this paper, we describe these methods in the context of curating the chicken TGF beta signaling pathway.

Text mining applications using natural language processing are often confronted with long and complicated sentences. This is observed particularly in the abstracts of scientific articles where authors summarize, in few sentences, the various facts described throughout the manuscript. Being rich in novel and important information, the abstract has been the primary target of biomedicai text mining applications. In this work, we aim to simplify complex sentences in abstracts of biomedicai text so that they can be readily processed by text mining applications. We focus on syntactic constructs that are frequently encountered in the biomedicai literature, such as coordinations, relative clauses, and appositions, with emphasis on their boundary detection. Our approach yielded good detection performance (average F-measure between 86.5% and 92.7%), and aided in improving biomedicai text mining applications, RLIMS-P and Rank Pref .

In this paper, we attempt to build a pipeline that identifies and extracts disease-drug relationships via sentence classification, and demonstrate the feasibility and utility of our approach using tardive dyskinesia as a case study. We manually developed and annotated a biomedicai training corpus for tardive dyskinesia. Using 10-fold cross validation, we tested and trained a naive Bayes classifier to identify sentences pertaining to disease-drug relationships. Our precision, recall, and F-measure were all approximately 66%, and area under the ROC curve was over 80%. Our method helps to elucidate various drug effects on tardive dyskinesia and constitutes an initial effort toward the task of disease-drug relationship extraction.

A neural network classification method has been developed as an alternative approach to the search/organization problem of large molecular databases. Two artificial neural systems have been implemented on a Cray supercomputer for rapid protein/nucleic acid sequence classifications. The neural networks used are three-layered, feed-forward networks that employ back-propagation learning algorithm. The molecular sequences are encoded into neural input vectors by applying an n-gram hashing method or a SVD (singular value decomposition) method. Once trained with known sequences in the molecular databases, the neural system becomes an associative memory capable of classifying unknown sequences based on the class information embedded in its neural interconnections. The protein system, which classifies proteins into PIR (Protein Identification Resource) superfamilies, showed a 82% to a close to 100% sensitivity at a speed that is about an order of magnitude faster than other search methods. The pilot nucleic acid system, which classifies ribosomal RNA sequences according to phylogenetic groups, has achieved a 100% classification accuracy. The system could be used to reduce the database search time and help organize the molecular sequence databases. The tool is generally applicable to any databases that are organized according to family relationships.

A gene classification artificial neural system has been developed for rapid annotation of the molecular sequencing data being generated by the Human Genome Project. Three neural networks have been implemented, one full-scale system to classify protein sequences according to PIR (Protein Identification Resources) superfamilies, one system to classify ribosomal RNA sequences into RDP (Ribosomal Database Project) phylogenetic classes, and one pilot system to classify proteins according to Blocks motifs. The sequence encoding schema involved an n-gram hashing method to convert molecular sequences into neural input vectors, a SVD (singular value decomposition) method to compress vectors, and a term weighting method to extract motif information. The neural networks used were three-layered, feed-forward networks that employed back-propagation or counter-propagation learning paradigms. The system runs faster by one to two orders of magnitude than existing method and has a sensitivity of 85 to 100%. The gene classification artificial neural system is available on the Internet, and may be extended into a gene identification system for classifying indiscriminately sequenced DNA fragments.

Motivation: Identifier (ID) mapping establishes links between various biological databases and is an essential first step for molecular data integration and functional annotation. ID mapping allows diverse molecular data on genes and proteins to be combined and mapped to functional pathways and ontologies. We have developed comprehensive protein-centric ID mapping services providing mappings for 90 IDs derived from databases on genes, proteins, pathways, diseases, structures, protein families, protein interaction, literature, ontologies, etc. The services are widely used and have been regularly updated since 2006. Availability: www.uniprot.org/mappingandproteininformation-resource.org/pirwww/search/idmapping.shtml Contact: huang@dbi.udel.edu

The iProClass database is an integrated resource that provides comprehensive family relationships and structural and functional features of proteins, with rich links to various databases. It is extended from ProClass, a protein family database that integrates PIR superfamilies and PROSITE motifs. The iProClass currently consists of more than 200 000 non-redundant PIR and SWISS-PROT proteins organized with more than 28 000 superfamilies, 2600 domains, 1300 motifs, 280 post-translational modification sites and links to more than 30 databases of protein families, structures, functions, genes, genomes, literature and taxonomy. Protein and family summary reports provide rich annotations, including membership information with length, taxonomy and keyword statistics, full family relationships, comprehensive enzyme and PDB cross-references and graphical feature display. The database facilitates classification-driven annotation for protein sequence databases and complete genomes, and supports structural and functional genomic research. The iProClass is implemented in Oracle 8i object-relational system and available for sequence search and report retrieval at http://pir.georgetow n.edu/iproclass/.

In the past decades, a variety of publicly available data repositories and resources have been developed to support protein related information management, data-driven hypothesis generation and biological knowledge discovery. However, there is also an increasing confusion for the researchers who are trying to quickly find the appropriate resources to help them solve their problems. In this chapter, we present a comprehensive review (with categorization and description) of major protein bioinformatics databases and resources that are relevant to comparative proteomics research. We conclude the chapter by discussing the challenges and opportunities for developing new protein bioinformatics databases.

High-throughput “omics” technologies bring new opportunities for biological and biomedical researchers to ask complex questions and gain new scientific insights. However, the voluminous, complex, and context-dependent data being maintained in heterogeneous and distributed environments plus the lack of well-defined data standard and standardized nomenclature imposes a major challenge which requires advanced computational methods and bioinformatics infrastructures for integration, mining, visualization, and comparative analysis to facilitate data-driven hypothesis generation and biological knowledge discovery. In this paper, we present the challenges in high-throughput “omics” data integration and analysis, introduce a protein-centric approach for systems integration of large and heterogeneous high-throughput “omics” data including microarray, mass spectrometry, protein sequence, protein structure, and protein interaction data, and use scientific case study to illustrate how one can use varied “omics” data from different laboratories to make useful connections that could lead to new biological knowledge.

Identification of protein-protein interactions (PPIs) and downstream functional events is important to biologists because these are key to building protein networks and biological pathways. Of high relevance are the PPIs involving proteins with some post-translational modification (PTM), since PTMs constitute one way of regulating protein function. However, this type of PPIs is not yet well represented in public databases (such as UniProtKB and IntAct), which would link the PPI information to the unmodified forms of proteins. Hence, literature is still the primary source for PPI information linked to modified forms of proteins. In this work, we propose to automate the process that mines and combines PPI information and PTM information from the biomedical literature via a novel, rule-based system. Further, the results are ranked based on the relations between PPI and PTM information. Currently, we limit the scope of our system to phosphorylation, a common PTM.

Protein posttranslational modification (PTM) is a fundamental biological process, and currently few text mining systems focus on PTM information extraction. A rule-based text mining system, RLIMS-P (Rule-based LIterature Mining System for Protein Phosphorylation), was recently developed by our group to extract protein substrate, kinase and phosphorylated residue/sites from MEDLINE abstracts. This chapter covers the evaluation and benchmarking of RLIMS-P and highlights some novel and unique features of the system. The extraction patterns of RLIMS-P capture a range of lexical, syntactic and semantic constraints found in sentences expressing phosphorylation information. RLIMS-P also has a second phase that puts together information extracted from different sentences. This is an important feature since it is not common to find the kinase, substrate and site of phosphorylation to be mentioned in the same sentence. Small modifications to the rules for extraction of phosphorylation information have also allowed us to develop systems for extraction of two other PTMs, acetylation and methylation. A thorough evaluation of these two systems needs to be completed. Finally, an online version of RLIMSP with enhanced functionalities, namely, phosphorylation annotation ranking, evidence tagging, and protein entity mapping, has been developed and is publicly accessible.

ProClass is a protein family database that organizes non-redundant sequence entries into families defined collectively by PROSITE patterns and PIR superfamilies. By combining global similarities and functional motifs into a single classification scheme, ProClass helps to reveal domain and family relationships and classify multi-domain proteins. The database currently consists of more than 120 000 sequence entries, approximately 60% of which is classified into about 3500 families. To maximize family information retrieval, the database provides links to various protein family/domain and structural class databases and contains multiple motif alignments of all PROSITE patterns as well as global alignments of PIR superfamilies. The motif sequences are retrieved from both PIR-International and SWISS-PROT databases, including a large number of new members detected by our GeneFIND family identification system. ProClass can be used to support full-scale genomic annotation, because of its high classification rate. The ProClass database is available for on-line search and record retrieval from our WWW server at http://diana.uthct.edu/proclass.html