Automated recognition of functional compound-protein relationships in literature.

MOTIVATION: Much effort has been invested in the identification of protein-protein interactions using text mining and machine learning methods. The extraction of functional relationships between chemical compounds and proteins from literature has received much less attention, and no ready-to-use open-source software is so far available for this task. METHOD: We created a new benchmark dataset of 2,613 sentences from abstracts containing annotations of proteins, small molecules, and their relationships. Two kernel methods were applied to classify these relationships as functional or non-functional, named shallow linguistic and all-paths graph kernel. Furthermore, the benefit of interaction verbs in sentences was evaluated. RESULTS: The cross-validation of the all-paths graph kernel (AUC value: 84.6%, F1 score: 79.0%) shows slightly better results than the shallow linguistic kernel (AUC value: 82.5%, F1 score: 77.2%) on our benchmark dataset. Both models achieve state-of-the-art performance in the research area of relation extraction. Furthermore, the combination of shallow linguistic and all-paths graph kernel could further increase the overall performance slightly. We used each of the two kernels to identify functional relationships in all PubMed abstracts (29 million) and provide the results, including recorded processing time. AVAILABILITY: The software for the tested kernels, the benchmark, the processed 29 million PubMed abstracts, all evaluation scripts, as well as the scripts for processing the complete PubMed database are freely available at https://github.com/KerstenDoering/CPI-Pipeline.

NANPDB: A Resource for Natural Products from Northern African Sources.

Natural products (NPs) are often regarded as sources of drugs or drug leads or simply as a "source of inspiration" for the discovery of novel drugs. We have built the Northern African Natural Products Database (NANPDB) by collecting information on ∼4500 NPs, covering literature data for the period from 1962 to 2016. The data cover compounds isolated mainly from plants, with contributions from some endophyte, animal (e.g., coral), fungal, and bacterial sources. The compounds were identified from 617 source species, belonging to 146 families. Computed physicochemical properties, often used to predict drug metabolism and pharmacokinetics, as well as predicted toxicity information, have been included for each compound in the data set. This is the largest collection of annotated natural compounds produced by native organisms from Northern Africa. While the database includes well-known drugs and drug leads, the medical potential of a majority of the molecules is yet to be investigated. The database could be useful for drug discovery efforts, analysis of the bioactivity of selected compounds, or the discovery of synthesis routes toward secondary metabolites. The current version of NANPDB is available at http://african-compounds.org/nanpdb/ .

PubMedPortable: A Framework for Supporting the Development of Text Mining Applications.

Information extraction from biomedical literature is continuously growing in scope and importance. Many tools exist that perform named entity recognition, e.g. of proteins, chemical compounds, and diseases. Furthermore, several approaches deal with the extraction of relations between identified entities. The BioCreative community supports these developments with yearly open challenges, which led to a standardised XML text annotation format called BioC. PubMed provides access to the largest open biomedical literature repository, but there is no unified way of connecting its data to natural language processing tools. Therefore, an appropriate data environment is needed as a basis to combine different software solutions and to develop customised text mining applications. PubMedPortable builds a relational database and a full text index on PubMed citations. It can be applied either to the complete PubMed data set or an arbitrary subset of downloaded PubMed XML files. The software provides the infrastructure to combine stand-alone applications by exporting different data formats, e.g. BioC. The presented workflows show how to use PubMedPortable to retrieve, store, and analyse a disease-specific data set. The provided use cases are well documented in the PubMedPortable wiki. The open-source software library is small, easy to use, and scalable to the user's system requirements. It is freely available for Linux on the web at https://github.com/KerstenDoering/PubMedPortable and for other operating systems as a virtual container. The approach was tested extensively and applied successfully in several projects.

StreptomeDB 2.0--an extended resource of natural products produced by streptomycetes.

Over the last decades, the genus Streptomyces has stirred huge interest in the scientific community as a source of bioactive compounds. The majority of all known antibiotics is isolated from these bacterial strains, as well as a variety of other drugs such as antitumor agents, immunosuppressants and antifungals. To the best of our knowledge, StreptomeDB was the first database focusing on compounds produced by streptomycetes. The new version presented herein represents a major step forward: its content has been increased to over 4000 compounds and more than 2500 host organisms. In addition, we have extended the background information and included hundreds of new manually curated references to literature. The latest update features a unique scaffold-based navigation system, which enables the exploration of the chemical diversity of StreptomeDB on a structural basis. We have included a phylogenetic tree, based on 16S rRNA sequences, which comprises more than two-thirds of the included host organisms. It enables visualizing the frequency, appearance, and persistence of compounds and scaffolds in an evolutionary context. Additionally, we have included predicted MS- and NMR-spectra of thousands of compounds for assignment of experimental data. The database is freely accessible via http://www.pharmaceutical-bioinformatics.org/streptomedb.

StreptomeDB: a resource for natural compounds isolated from Streptomyces species.

Bacteria from the genus Streptomyces are very important for the production of natural bioactive compounds such as antibiotic, antitumour or immunosuppressant drugs. Around two-thirds of all known natural antibiotics are produced by these bacteria. An enormous quantity of crucial data related to this genus has been generated and published, but so far no freely available and comprehensive database exists. Here, we present StreptomeDB (http://www.pharmaceutical-bioinformatics.de/streptomedb/). To the best of our knowledge, this is the largest database of natural products isolated from Streptomyces. It contains >2400 unique and diverse compounds from >1900 different Streptomyces strains and substrains. In addition to names and molecular structures of the compounds, information about source organisms, references, biological role, activities and synthesis routes (e.g. polyketide synthase derived and non-ribosomal peptides derived) is included. Data can be accessed through queries on compound names, chemical structures or organisms. Extraction from the literature was performed through automatic text mining of thousands of articles from PubMed, followed by manual curation. All annotated compound structures can be downloaded from the website and applied for in silico screenings for identifying new active molecules with undiscovered properties.

Mining and evaluation of molecular relationships in literature.

MOTIVATION: Specific information on newly discovered proteins is often difficult to find in literature. Particularly if only sequences and no common names of proteins or genes are available, preceding sequence similarity searches can be crucial for the process of information collection. In drug research, it is important to know whether a small molecule targets only one specific protein or whether similar or homologous proteins are also influenced that may account for possible side effects. RESULTS: prolific (protein-literature investigation for interacting compounds) provides a one-step solution to investigate available information on given protein names, sequences, similar proteins or sequences on the gene level. Co-occurrences of UniProtKB/Swiss-Prot proteins and PubChem compounds in all PubMed abstracts are retrievable. Concise 'heat-maps' and tables display frequencies of co-occurrences. They provide links to processed literature with highlighted found protein and compound synonyms. Evaluation with manually curated drug-protein relationships showed that up to 69% could be discovered by automatic text-processing. Examples are presented to demonstrate the capabilities of prolific. AVAILABILITY: The web-application is available at http://prolific.pharmaceutical-bioinformatics.de and a web service at http://www.pharmaceutical-bioinformatics.de/prolific/soap/prolific.wsdl. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.