Mining literature for protein-protein interactions

MOTIVATION: A central problem in bioinformatics is how to capture information from the vast current scientific literature in a form suitable for analysis by computer. We address the special case of information on protein-protein interactions, and show that the frequencies of words in Medline abstracts can be used to determine whether or not a given paper discusses protein-protein interactions. For those papers determined to discuss this topic, the relevant information can be captured for the Database of Interacting PROTEINS: Furthermore, suitable gene annotations can also be captured. RESULTS: Our Bayesian approach scores Medline abstracts for probability of discussing the topic of interest according to the frequencies of discriminating words found in the abstract. More than 80 discriminating words (e.g. complex, interaction, two-hybrid) were determined from a training set of 260 Medline abstracts corresponding to previously validated entries in the Database of Interacting Proteins. Using these words and a log likelihood scoring function, approximately 2000 Medline abstracts were identified as describing interactions between yeast proteins. This approach now forms the basis for the rapid expansion of the Database of Interacting Proteins.     

Automatic extraction of mutations from Medline and cross-validation with OMIM

Mutations help us to understand the molecular origins of diseases. Researchers, therefore, both publish and seek disease-relevant mutations in public databases and in scientific literature, e.g. Medline. The retrieval tends to be time-consuming and incomplete. Automated screening of the literature is more efficient. We developed extraction methods (called MEMA) that scan Medline abstracts for mutations. MEMA identified 24,351 singleton mutations in conjunction with a HUGO gene name out of 16,728 abstracts. From a sample of 100 abstracts we estimated the recall for the identification of mutation-gene pairs to 35% at a precision of 93%. Recall for the mutation detection alone was >67% with a precision rate of >96%. This shows that our system produces reliable data. The subset consisting of protein sequence mutations (PSMs) from MEMA was compared to the entries in OMIM (20,503 entries versus 6699, respectively). We found 1826 PSM-gene pairs to be in common to both datasets (cross-validated). This is 27% of all PSM-gene pairs in OMIM and 91% of those pairs from OMIM which co-occur in at least one Medline abstract. We conclude that Medline covers a large portion of the mutations known to OMIM. Another large portion could be artificially produced mutations from mutagenesis experiments. Access to the database of extracted mutation-gene pairs is available through the web pages of the EBI (refer to http://www.ebi. ac.uk/rebholz/index.html).     

Markov model recognition and classification of DNA/protein sequences within large text databases

MOTIVATION: Short sequence patterns frequently define regions of biological interest (binding sites, immune epitopes, primers, etc.), yet a large fraction of this information exists only within the scientific literature and is thus difficult to locate via conventional means (e.g. keyword queries or manual searches). We describe herein a system to accurately identify and classify sequence patterns from within large corpora using an n-gram Markov model (MM). RESULTS: As expected, on test sets we found that identification of sequences with limited alphabets and/or regular structures such as nucleic acids (non-ambiguous) and peptide abbreviations (3-letter) was highly accurate, whereas classification of symbolic (1-letter) peptide strings with more complex alphabets was more problematic. The MM was used to analyze two very large, sequence-containing corpora: over 7.75 million Medline abstracts and 9000 full-text articles from Journal of Virology. Performance was benchmarked by comparing the results with Journal of Virology entries in two existing manually curated databases: VirOligo and the HLA Ligand Database. Performance estimates were 98 +/- 2% precision/84% recall for primer identification and classification and 67 +/- 6% precision/85% recall for peptide epitopes. We also find a dramatic difference between the amounts of sequence-related data reported in abstracts versus full text. Our results suggest that automated extraction and classification of sequence elements is a promising, low-cost means of sequence database curation and annotation. AVAILABILITY: MM routine and datasets are available upon request.     

Textpresso: an ontology-based information retrieval and extraction system for biological literature

We have developed Textpresso, a new text-mining system for scientific literature whose capabilities go far beyond those of a simple keyword search engine. Textpresso's two major elements are a collection of the full text of scientific articles split into individual sentences, and the implementation of categories of terms for which a database of articles and individual sentences can be searched. The categories are classes of biological concepts (e.g., gene, allele, cell or cell group, phenotype, etc.) and classes that relate two objects (e.g., association, regulation, etc.) or describe one (e.g., biological process, etc.). Together they form a catalog of types of objects and concepts called an ontology. After this ontology is populated with terms, the whole corpus of articles and abstracts is marked up to identify terms of these categories. The current ontology comprises 33 categories of terms. A search engine enables the user to search for one or a combination of these tags and/or keywords within a sentence or document, and as the ontology allows word meaning to be queried, it is possible to formulate semantic queries. Full text access increases recall of biological data types from 45% to 95%. Extraction of particular biological facts, such as gene-gene interactions, can be accelerated significantly by ontologies, with Textpresso automatically performing nearly as well as expert curators to identify sentences; in searches for two uniquely named genes and an interaction term, the ontology confers a 3-fold increase of search efficiency. Textpresso currently focuses on Caenorhabditis elegans literature, with 3,800 full text articles and 16,000 abstracts. The lexicon of the ontology contains 14,500 entries, each of which includes all versions of a specific word or phrase, and it includes all categories of the Gene Ontology database. Textpresso is a useful curation tool, as well as search engine for researchers, and can readily be extended to other organism-specific corpora of text. Textpresso can be accessed at http://www.textpresso.org or via WormBase at http://www.wormbase.org.     

EBIMed--text crunching to gather facts for proteins from Medline

To allow efficient and systematic retrieval of statements from Medline we have developed EBIMed, a service that combines document retrieval with co-occurrence-based analysis of Medline abstracts. Upon keyword query, EBIMed retrieves the abstracts from EMBL-EBI's installation of Medline and filters for sentences that contain biomedical terminology maintained in public bioinformatics resources. The extracted sentences and terminology are used to generate an overview table on proteins, Gene Ontology (GO) annotations, drugs and species used in the same biological context. All terms in retrieved abstracts and extracted sentences are linked to their entries in biomedical databases. We assessed the quality of the identification of terms and relations in the retrieved sentences. More than 90% of the protein names found indeed represented a protein. According to the analysis of four protein-protein pairs from the Wnt pathway we estimated that 37% of the statements containing such a pair mentioned a meaningful interaction and clarified the interaction of Dkk with LRP. We conclude that EBIMed improves access to information where proteins and drugs are involved in the same biological process, e.g. statements with GO annotations of proteins, protein-protein interactions and effects of drugs on proteins. AVAILABILITY: Available at http://www.ebi.ac.uk/Rebholz-srv/ebimed     

miRTex: A Text Mining System for miRNA-Gene Relation Extraction

MicroRNAs (miRNAs) regulate a wide range of cellular and developmental processes through gene expression suppression or mRNA degradation. Experimentally validated miRNA gene targets are often reported in the literature. In this paper, we describe miRTex, a text mining system that extracts miRNA-target relations, as well as miRNA-gene and gene-miRNA regulation relations. The system achieves good precision and recall when evaluated on a literature corpus of 150 abstracts with F-scores close to 0.90 on the three different types of relations. We conducted full-scale text mining using miRTex to process all the Medline abstracts and all the full-length articles in the PubMed Central Open Access Subset. The results for all the Medline abstracts are stored in a database for interactive query and file download via the website at http://proteininformationresource.org/mirtex. Using miRTex, we identified genes potentially regulated by miRNAs in Triple Negative Breast Cancer, as well as miRNA-gene relations that, in conjunction with kinase-substrate relations, regulate the response to abiotic stress in Arabidopsis thaliana. These two use cases demonstrate the usefulness of miRTex text mining in the analysis of miRNA-regulated biological processes.     

BioContrasts: extracting and exploiting protein-protein contrastive relations from biomedical literature

MOTIVATION: Contrasts are useful conceptual vehicles for learning processes and exploratory research of the unknown. For example, contrastive information between proteins can reveal what similarities, divergences and relations there are of the two proteins, leading to invaluable insights for better understanding about the proteins. Such contrastive information are found to be reported in the biomedical literature. However, there have been no reported attempts in current biomedical text mining work that systematically extract and present such useful contrastive information from the literature for exploitation. RESULTS: Our BioContrasts system extracts protein-protein contrastive information from MEDLINE abstracts and presents the information to biologists in a web-application for exploitation. Contrastive information are identified in the text abstracts with contrastive negation patterns such as 'A but not B'. A total of 799 169 pairs of contrastive expressions were successfully extracted from 2.5 million MEDLINE abstracts. Using grounding of contrastive protein names to Swiss-Prot entries, we were able to produce 41 471 pieces of contrasts between Swiss-Prot protein entries. These contrastive pieces of information are then presented via a user-friendly interactive web portal that can be exploited for applications such as the refinement of biological pathways. AVAILABILITY: BioContrasts can be accessed at http://biocontrasts.i2r.a-star.edu.sg. It is also mirrored at http://biocontrasts.biopathway.org. SUPPLEMENTARY INFORMATION: Supplementary materials are available at Bioinformatics online.     

Identification of biological relationships from text documents using efficient computational methods

The biological literature databases continue to grow rapidly with vital information that is important for conducting sound biomedical research and development. The current practices of manually searching for information and extracting pertinent knowledge are tedious, time-consuming tasks even for motivated biological researchers. Accurate and computationally efficient approaches in discovering relationships between biological objects from text documents are important for biologists to develop biological models. The term "object" refers to any biological entity such as a protein, gene, cell cycle, etc. and relationship refers to any dynamic action one object has on another, e.g. protein inhibiting another protein or one object belonging to another object such as, the cells composing an organ. This paper presents a novel approach to extract relationships between multiple biological objects that are present in a text document. The approach involves object identification, reference resolution, ontology and synonym discovery, and extracting object-object relationships. Hidden Markov Models (HMMs), dictionaries, and N-Gram models are used to set the framework to tackle the complex task of extracting object-object relationships. Experiments were carried out using a corpus of one thousand Medline abstracts. Intermediate results were obtained for the object identification process, synonym discovery, and finally the relationship extraction. For the thousand abstracts, 53 relationships were extracted of which 43 were correct, giving a specificity of 81 percent. These results are promising for multi-object identification and relationship finding from biological documents.     

Finding relevant references to genes and proteins in Medline using a Bayesian approach

MOTIVATION: Mining the biomedical literature for references to genes and proteins always involves a tradeoff between high precision with false negatives, and high recall with false positives. Having a reliable method for assessing the relevance of literature mining results is crucial to finding ways to balance precision and recall, and for subsequently building automated systems to analyze these results. We hypothesize that abstracts and titles that discuss the same gene or protein use similar words. To validate this hypothesis, we built a dictionary- and rule-based system to mine Medline for references to genes and proteins, and used a Bayesian metric for scoring the relevance of each reference assignment. RESULTS: We analyzed the entire set of Medline records from 1966 to late 2001, and scored each gene and protein reference using a Bayesian estimated probability (EP) based on word frequency in a training set of 137837 known assignments from 30594 articles to 36197 gene and protein symbols. Two test sets of 148 and 150 randomly chosen assignments, respectively, were hand-validated and categorized as either good or bad. The distributions of EP values, when plotted on a log-scale histogram, are shown to markedly differ between good and bad assignments. Using EP values, recall was 100% at 61% precision (EP=2 x 10(-5)), 63% at 88% precision (EP=0.008), and 10% at 100% precision (EP=0.1). These results show that Medline entries discussing the same gene or protein have similar word usage, and that our method of assessing this similarity using EP values is valid, and enables an EP cutoff value to be determined that accurately and reproducibly balances precision and recall, allowing automated analysis of literature mining results. .     

Identifying biological concepts from a protein-related corpus with a probabilistic topic model

Background  

Biomedical literature, e.g., MEDLINE, contains a wealth of knowledge regarding functions of proteins. Major recurring biological concepts within such text corpora represent the domains of this body of knowledge. The goal of this research is to identify the major biological topics/concepts from a corpus of protein-related MEDLINE^? titles and abstracts by applying a probabilistic topic model.     

Functional cohesion of gene sets determined by latent semantic indexing of PubMed abstracts

High-throughput genomic technologies enable researchers to identify genes that are co-regulated with respect to specific experimental conditions. Numerous statistical approaches have been developed to identify differentially expressed genes. Because each approach can produce distinct gene sets, it is difficult for biologists to determine which statistical approach yields biologically relevant gene sets and is appropriate for their study. To address this issue, we implemented Latent Semantic Indexing (LSI) to determine the functional coherence of gene sets. An LSI model was built using over 1 million Medline abstracts for over 20,000 mouse and human genes annotated in Entrez Gene. The gene-to-gene LSI-derived similarities were used to calculate a literature cohesion p-value (LPv) for a given gene set using a Fisher's exact test. We tested this method against genes in more than 6,000 functional pathways annotated in Gene Ontology (GO) and found that approximately 75% of gene sets in GO biological process category and 90% of the gene sets in GO molecular function and cellular component categories were functionally cohesive (LPv<0.05). These results indicate that the LPv methodology is both robust and accurate. Application of this method to previously published microarray datasets demonstrated that LPv can be helpful in selecting the appropriate feature extraction methods. To enable real-time calculation of LPv for mouse or human gene sets, we developed a web tool called Gene-set Cohesion Analysis Tool (GCAT). GCAT can complement other gene set enrichment approaches by determining the overall functional cohesion of data sets, taking into account both explicit and implicit gene interactions reported in the biomedical literature. Availability: GCAT is freely available at http://binf1.memphis.edu/gcat.     

RelEx--relation extraction using dependency parse trees

MOTIVATION: The discovery of regulatory pathways, signal cascades, metabolic processes or disease models requires knowledge on individual relations like e.g. physical or regulatory interactions between genes and proteins. Most interactions mentioned in the free text of biomedical publications are not yet contained in structured databases. RESULTS: We developed RelEx, an approach for relation extraction from free text. It is based on natural language preprocessing producing dependency parse trees and applying a small number of simple rules to these trees. We applied RelEx on a comprehensive set of one million MEDLINE abstracts dealing with gene and protein relations and extracted approximately 150,000 relations with an estimated performance of both 80% precision and 80% recall. AVAILABILITY: The used natural language preprocessing tools are free for use for academic research. Test sets and relation term lists are available from our website (http://www.bio.ifi.lmu.de/publications/RelEx/).     

基于MySQL数据库的植物编码基因数据挖掘初步分析

利用Linux操作系统的文本过滤技术对通过关键词检索提取NCBI网络数据库的核酸序列和Medline论文摘要快速建立了5种主要植物即拟南芥菜、水稻、玉米、小麦、马铃薯的编码mRNA序列和论文摘要的MySQL数据库,提取的编码序列记录共达到48,900多条,摘要记录达24,000多条,初步分析了这些mRNA序列的主要类型,如包括编码酶蛋白、转录因子等的mRNA序列,并进一步就MySQL数据库数据挖掘可能用于建构基因调控网络进行了初探。     

AliBaba: PubMed as a graph

The biomedical literature contains a wealth of information on associations between many different types of objects, such as protein-protein interactions, gene-disease associations and subcellular locations of proteins. When searching such information using conventional search engines, e.g. PubMed, users see the data only one-abstract at a time and 'hidden' in natural language text. AliBaba is an interactive tool for graphical summarization of search results. It parses the set of abstracts that fit a PubMed query and presents extracted information on biomedical objects and their relationships as a graphical network. AliBaba extracts associations between cells, diseases, drugs, proteins, species and tissues. Several filter options allow for a more focused search. Thus, researchers can grasp complex networks described in various articles at a glance. AVAILABILITY: http://alibaba.informatik.hu-berlin.de/     

Medline search engine for finding genetic markers with biological significance

MOTIVATION: Genome-wide high density SNP association studies are expected to identify various SNP alleles associated with different complex disorders. Understanding the biological significance of these SNP alleles in the context of existing literature is a major challenge since existing search engines are not designed to search literature for SNPs or other genetic markers. The literature mining of gene and protein functions has received significant attention and effort while similar work on genetic markers and their related diseases is still in its infancy. Our goal is to develop a web-based tool that facilitates the mining of Medline literature related to genetic studies and gene/protein function studies. Our solution consists of four main function modules for (1) identification of different types of genetic markers or genetic variations in Medline records (2) distinguishing positive versus negative linkage or association between genetic markers and diseases (3) integrating marker genomic location data from different databases to enable the retrieval of Medline records related to markers in the same linkage disequilibrium region (4) and a web interface called MarkerInfoFinder to search, display, sort and download Medline citation results. Tests using published data suggest MarkerInfoFinder can significantly increase the efficiency of finding genetic disorders and their underlying molecular mechanisms. The functions we developed will also be used to build a knowledge base for genetic markers and diseases. AVAILABILITY: The MarkerInfoFinder is publicly available at: http://brainarray.mbni.med.umich.edu/brainarray/datamining/MarkerInfoFinder.     

Can Bibliographic Pointers for Known Biological Data Be Found Automatically? Protein Interactions as a Case Study

The Dictionary of Interacting Proteins (DIP) (Xenarios et al., 2000) is a large repository of protein interactions: its March 2000 release included 2379 protein pairs whose interactions have been detected by experimental methods. Even if many of these correspond to poorly characterized proteins, the result of massive yeast two-hybrid screenings, as many as 851 correspond to interactions detected using direct biochemical methods.We used information retrieval technology to search automatically for sentences in Medline abstracts that support these 851 DIP interactions. Surprisingly, we found correspondence between DIP protein pairs and Medline sentences describing their interactions in only 30% of the cases. This low coverage has interesting consequences regarding the quality of annotations (references) introduced in the database and the limitations of the application of information extraction (IE) technology to Molecular Biology. It is clear that the limitation of analyzing abstracts rather than full papers and the lack of standard protein names are difficulties of considerably more importance than the limitations of the IE methodology employed. A positive finding is the capacity of the IE system to identify new relations between proteins, even in a set of proteins previously characterized by human experts. These identifications are made with a considerable degree of precision.THIS IS, TO OUR KNOWLEDGE, THE FIRST LARGE SCALE ASSESSMENT OF IE CAPACITY TO DETECT PREVIOUSLY KNOWN INTERACTIONS: we thus propose the use of the DIP data set as a biological reference to benchmark IE systems.     

Weakly supervised learning of information structure of scientific abstracts--is it accurate enough to benefit real-world tasks in biomedicine?

MOTIVATION: Many practical tasks in biomedicine require accessing specific types of information in scientific literature; e.g. information about the methods, results or conclusions of the study in question. Several approaches have been developed to identify such information in scientific journal articles. The best of these have yielded promising results and proved useful for biomedical text mining tasks. However, relying on fully supervised machine learning (ml) and a large body of annotated data, existing approaches are expensive to develop and port to different tasks. A potential solution to this problem is to employ weakly supervised learning instead. In this article, we investigate a weakly supervised approach to identifying information structure according to a scheme called Argumentative Zoning (az). We apply four weakly supervised classifiers to biomedical abstracts and evaluate their performance both directly and in a real-life scenario in the context of cancer risk assessment. RESULTS: Our best weakly supervised classifier (based on the combination of active learning and self-training) performs well on the task, outperforming our best supervised classifier: it yields a high accuracy of 81% when just 10% of the labeled data is used for training. When cancer risk assessors are presented with the resulting annotated abstracts, they find relevant information in them significantly faster than when presented with unannotated abstracts. These results suggest that weakly supervised learning could be used to improve the practical usefulness of information structure for real-life tasks in biomedicine.