Combination of text-mining algorithms increases the performance

Automatic extraction of protein-protein interactions using grammatical relationship graph

Background

Relationships between bio-entities (genes, proteins, diseases, etc.) constitute a significant part of our knowledge. Most of this information is documented as unstructured text in different forms, such as books, articles and on-line pages. Automatic extraction of such information and storing it in structured form could help researchers more easily access such information and also make it possible to incorporate it in advanced integrative analysis. In this study, we developed a novel approach to extract bio-entity relationships information using Nature Language Processing (NLP) and a graph-theoretic algorithm.

Methods

Our method, called GRGT (Grammatical Relationship Graph for Triplets), not only extracts the pairs of terms that have certain relationships, but also extracts the type of relationship (the word describing the relationships). In addition, the directionality of the relationship can also be extracted. Our method is based on the assumption that a triplet exists for a pair of interactions. A triplet is defined as two terms (entities) and an interaction word describing the relationship of the two terms in a sentence. We first use a sentence parsing tool to obtain the sentence structure represented as a dependency graph where words are nodes and edges are typed dependencies. The shortest paths among the pairs of words in the triplet are then extracted, which form the basis for our information extraction method. Flexible pattern matching scheme was then used to match a triplet graph with unknown relationship to those triplet graphs with labels (True or False) in the database.

Results

We applied the method on three benchmark datasets to extract the protein-protein-interactions (PPIs), and obtained better precision than the top performing methods in literature.

Conclusions

We have developed a method to extract the protein-protein interactions from biomedical literature. PPIs extracted by our method have higher precision among other methods, suggesting that our method can be used to effectively extract PPIs and deposit them into databases. Beyond extracting PPIs, our method could be easily extended to extracting relationship information between other bio-entities.

Survey of Natural Language Processing Techniques in Bioinformatics

Informatics methods, such as text mining and natural language processing, are always involved in bioinformatics research. In this study, we discuss text mining and natural language processing methods in bioinformatics from two perspectives. First, we aim to search for knowledge on biology, retrieve references using text mining methods, and reconstruct databases. For example, protein-protein interactions and gene-disease relationship can be mined from PubMed. Then, we analyze the applications of text mining and natural language processing techniques in bioinformatics, including predicting protein structure and function, detecting noncoding RNA. Finally, numerous methods and applications, as well as their contributions to bioinformatics, are discussed for future use by text mining and natural language processing researchers.

The Growing Importance of CNVs: New Insights for Detection and Clinical Interpretation

Differences between genomes can be due to single nucleotide variants, translocations, inversions, and copy number variants (CNVs, gain or loss of DNA). The latter can range from sub-microscopic events to complete chromosomal aneuploidies. Small CNVs are often benign but those larger than 500 kb are strongly associated with morbid consequences such as developmental disorders and cancer. Detecting CNVs within and between populations is essential to better understand the plasticity of our genome and to elucidate its possible contribution to disease. Hence there is a need for better-tailored and more robust tools for the detection and genome-wide analyses of CNVs. While a link between a given CNV and a disease may have often been established, the relative CNV contribution to disease progression and impact on drug response is not necessarily understood. In this review we discuss the progress, challenges, and limitations that occur at different stages of CNV analysis from the detection (using DNA microarrays and next-generation sequencing) and identification of recurrent CNVs to the association with phenotypes. We emphasize the importance of germline CNVs and propose strategies to aid clinicians to better interpret structural variations and assess their clinical implications.

DESTAF: a database of text-mined associations for reproductive toxins potentially affecting human fertility

The Dragon Exploration System for Toxicants and Fertility (DESTAF) is a publicly available resource which enables researchers to efficiently explore both known and potentially novel information and associations in the field of reproductive toxicology. To create DESTAF we used data from the literature (including over 10500 PubMed abstracts), several publicly available biomedical repositories, and specialized, curated dictionaries. DESTAF has an interface designed to facilitate rapid assessment of the key associations between relevant concepts, allowing for a more in-depth exploration of information based on different gene/protein-, enzyme/metabolite-, toxin/chemical-, disease- or anatomically centric perspectives. As a special feature, DESTAF allows for the creation and initial testing of potentially new association hypotheses that suggest links between biological entities identified through the database. DESTAF, along with a PDF manual, can be found at http://cbrc.kaust.edu.sa/destaf. It is free to academic and non-commercial users and will be updated quarterly.

Bayesian inference for genomic data integration reduces misclassification rate in predicting protein-protein interactions

Protein-protein interactions (PPIs) are essential to most fundamental cellular processes. There has been increasing interest in reconstructing PPIs networks. However, several critical difficulties exist in obtaining reliable predictions. Noticeably, false positive rates can be as high as >80%. Error correction from each generating source can be both time-consuming and inefficient due to the difficulty of covering the errors from multiple levels of data processing procedures within a single test. We propose a novel Bayesian integration method, deemed nonparametric Bayes ensemble learning (NBEL), to lower the misclassification rate (both false positives and negatives) through automatically up-weighting data sources that are most informative, while down-weighting less informative and biased sources. Extensive studies indicate that NBEL is significantly more robust than the classic naïve Bayes to unreliable, error-prone and contaminated data. On a large human data set our NBEL approach predicts many more PPIs than naïve Bayes. This suggests that previous studies may have large numbers of not only false positives but also false negatives. The validation on two human PPIs datasets having high quality supports our observations. Our experiments demonstrate that it is feasible to predict high-throughput PPIs computationally with substantially reduced false positives and false negatives. The ability of predicting large numbers of PPIs both reliably and automatically may inspire people to use computational approaches to correct data errors in general, and may speed up PPIs prediction with high quality. Such a reliable prediction may provide a solid platform to other studies such as protein functions prediction and roles of PPIs in disease susceptibility.

Using information mining of the medical literature to improve drug safety

OBJECTIVE

Prescription drugs can be associated with adverse effects (AEs) that are unrecognized despite evidence in the medical literature, as shown by rofecoxib's late recall in 2004. We assessed whether applying information mining to PubMed could reveal major drug-AE associations if articles testing whether drugs cause AEs are over-represented in the literature.

DESIGN

MEDLINE citations published between 1949 and September 2009 were retrieved if they mentioned one of 38 drugs and one of 55 AEs. A statistical document classifier (using MeSH index terms) was constructed to remove irrelevant articles unlikely to test whether a drug caused an AE. The remaining relevant articles were analyzed using a disproportionality analysis that identified drug-AE associations (signals of disproportionate reporting) using step-up procedures developed to control the familywise type I error rate.

MEASUREMENTS

Sensitivity and positive predictive value (PPV) for empirical drug-AE associations as judged against drug-AE associations subject to FDA warnings.

RESULTS

In testing, the statistical document classifier identified relevant articles with 81% sensitivity and 87% PPV. Using data filtered by the statistical document classifier, base-case models showed 64.9% sensitivity and 42.4% PPV for detecting FDA warnings. Base-case models discovered 54% of all detected FDA warnings using literature published before warnings. For example, the rofecoxib-heart disease association was evident using literature published before 2002. Analyses incorporating literature mentioning AEs common to the drug class of interest yielded 71.4% sensitivity and 40.7% PPV.

CONCLUSIONS

Results from large-scale literature retrieval and analysis (literature mining) compared favorably with and could complement current drug safety methods.

Extracting causal relations on HIV drug resistance from literature

BACKGROUND

In HIV treatment it is critical to have up-to-date resistance data of applicable drugs since HIV has a very high rate of mutation. These data are made available through scientific publications and must be extracted manually by experts in order to be used by virologists and medical doctors. Therefore there is an urgent need for a tool that partially automates this process and is able to retrieve relations between drugs and virus mutations from literature.

RESULTS

In this work we present a novel method to extract and combine relationships between HIV drugs and mutations in viral genomes. Our extraction method is based on natural language processing (NLP) which produces grammatical relations and applies a set of rules to these relations. We applied our method to a relevant set of PubMed abstracts and obtained 2,434 extracted relations with an estimated performance of 84% for F-score. We then combined the extracted relations using logistic regression to generate resistance values for each <drug, mutation> pair. The results of this relation combination show more than 85% agreement with the Stanford HIVDB for the ten most frequently occurring mutations. The system is used in 5 hospitals from the Virolab project http://www.virolab.org to preselect the most relevant novel resistance data from literature and present those to virologists and medical doctors for further evaluation.

CONCLUSIONS

The proposed relation extraction and combination method has a good performance on extracting HIV drug resistance data. It can be used in large-scale relation extraction experiments. The developed methods can also be applied to extract other type of relations such as gene-protein, gene-disease, and disease-mutation.

Bayesian inference of protein-protein interactions from biological literature

MOTIVATION

Protein-protein interaction (PPI) extraction from published biological articles has attracted much attention because of the importance of protein interactions in biological processes. Despite significant progress, mining PPIs from literatures still rely heavily on time- and resource-consuming manual annotations.

RESULTS

In this study, we developed a novel methodology based on Bayesian networks (BNs) for extracting PPI triplets (a PPI triplet consists of two protein names and the corresponding interaction word) from unstructured text. The method achieved an overall accuracy of 87% on a cross-validation test using manually annotated dataset. We also showed, through extracting PPI triplets from a large number of PubMed abstracts, that our method was able to complement human annotations to extract large number of new PPIs from literature.

AVAILABILITY

Programs/scripts we developed/used in the study are available at http://stat.fsu.edu/~jinfeng/datasets/Bio-SI-programs-Bayesian-chowdhary-zhang-liu.zip.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

DDESC: Dragon database for exploration of sodium channels in human

Background

Sodium channels are heteromultimeric, integral membrane proteins that belong to a superfamily of ion channels. The mutations in genes encoding for sodium channel proteins have been linked with several inherited genetic disorders such as febrile epilepsy, Brugada syndrome, ventricular fibrillation, long QT syndrome, or channelopathy associated insensitivity to pain. In spite of these significant effects that sodium channel proteins/genes could have on human health, there is no publicly available resource focused on sodium channels that would support exploration of the sodium channel related information.

Results

We report here Dragon Database for Exploration of Sodium Channels in Human (DDESC), which provides comprehensive information related to sodium channels regarding different entities, such as "genes and proteins", "metabolites and enzymes", "toxins", "chemicals with pharmacological effects", "disease concepts", "human anatomy", "pathways and pathway reactions" and their potential links. DDESC is compiled based on text- and data-mining. It allows users to explore potential associations between different entities related to sodium channels in human, as well as to automatically generate novel hypotheses.

Conclusion

DDESC is first publicly available resource where the information related to sodium channels in human can be explored at different levels. This database is freely accessible for academic and non-profit users via the worldwide web .

Evading the annotation bottleneck: using sequence similarity to search non-sequence gene data

Background

Non-sequence gene data (images, literature, etc.) can be found in many different public databases. Access to these data is mostly by text based methods using gene names; however, gene annotation is neither complete, nor fully systematic between organisms, and is also not generally stable over time. This provides some challenges for text based access, especially for cross-species searches. We propose a method for non-sequence data retrieval based on sequence similarity, which removes dependence on annotation and text searches. This work was motivated by the need to provide better access to large numbers of in situ images, and the observation that such image data were usually associated with a specific gene sequence. Sequence similarity searches are found in existing gene oriented databases, but mostly give indirect access to non-sequence data via navigational links.

Results

Three applications were built to explore the proposed method: accessing image data, literature and gene names. Searches are initiated with the sequence of the user's gene of interest, which is searched against a database of sequences associated with the target data. The matching (non-sequence) target data are returned directly to the user's browser, organised by sequence similarity. The method worked well for the intended application in image data management. Comparison with text based searches of the image data set showed the accuracy of the method. Applied to literature searches it facilitated retrieval of mostly high relevance references. Applied to gene name data it provided a useful analysis of name variation of related genes within and between species.

Conclusion

This method makes a powerful and useful addition to existing methods for searching gene data based on text retrieval or curated gene lists. In particular the method facilitates cross-species comparisons, and enables the handling of novel or otherwise un-annotated genes. Applications using the method are quick and easy to build, and the data require little maintenance. This approach largely circumvents the need for annotation, which can be a major obstacle to the development of genomic scale data resources.

Integrating protein-protein interactions and text mining for protein function prediction

Background

Functional annotation of proteins remains a challenging task. Currently the scientific literature serves as the main source for yet uncurated functional annotations, but curation work is slow and expensive. Automatic techniques that support this work are still lacking reliability. We developed a method to identify conserved protein interaction graphs and to predict missing protein functions from orthologs in these graphs. To enhance the precision of the results, we furthermore implemented a procedure that validates all predictions based on findings reported in the literature.

Results

Using this procedure, more than 80% of the GO annotations for proteins with highly conserved orthologs that are available in UniProtKb/Swiss-Prot could be verified automatically. For a subset of proteins we predicted new GO annotations that were not available in UniProtKb/Swiss-Prot. All predictions were correct (100% precision) according to the verifications from a trained curator.

Conclusion

Our method of integrating CCSs and literature mining is thus a highly reliable approach to predict GO annotations for weakly characterized proteins with orthologs.