BACKGROUND: In the modern era, the growth of scientific literature presents a daunting challenge for researchers to keep informed of advancements across multiple disciplines. OBJECTIVE: We apply natural language processing (NLP) and embedding learning concepts to design PubDigest, a tool that combs PubMed literature, aiming to pinpoint potential drugs that could be repurposed. METHODS: Using NLP, especially term associations through word embeddings, we explored unrecognized relationships between drugs and diseases. To illustrate the utility of PubDigest, we focused on chronic thromboembolic pulmonary hypertension (CTEPH), a rare disease with an overall limited number of scientific publications. RESULTS: Our literature analysis identified key clinical features linked to CTEPH by applying term frequency-inverse document frequency (TF-IDF) scoring, a technique measuring a term's significance in a text corpus. This allowed us to map related diseases. One standout was venous thrombosis (VT), which showed strong semantic links with CTEPH. Looking deeper, we discovered potential repurposing candidates for CTEPH through large-scale neural network-based contextualization of literature and predictive modeling on both the CTEPH and the VT literature corpora to find novel, yet unrecognized associations between the two diseases. Alongside the anti-thrombotic agent caplacizumab, benzofuran derivatives were an intriguing find. In particular, the benzofuran derivative amiodarone displayed potential anti-thrombotic properties in the literature. Our in vitro tests confirmed amiodarone's ability to reduce platelet aggregation significantly by 68% (p = 0.02). However, real-world clinical data indicated that CTEPH patients receiving amiodarone treatment faced a significant 15.9% higher mortality risk (p<0.001). CONCLUSIONS: While NLP offers an innovative approach to interpreting scientific literature, especially for drug repurposing, it is crucial to combine it with complementary methods like in vitro testing and real-world evidence. Our exploration with benzofuran derivatives and CTEPH underscores this point. Thus, blending NLP with hands-on experiments and real-world clinical data can pave the way for faster and safer drug repurposing approaches, especially for rare diseases like CTEPH.

BACKGROUND: Searching for two-dimensional (2D) structural similarities is a useful tool to identify new active compounds in drug-discovery programs. However, as 2D similarity measures neglect important structural and functional features, similarity by 2D might be underestimated. In the present study, we used combined 2D and three-dimensional (3D) similarity comparisons to reveal possible new functions and/or side-effects of known bioactive compounds. RESULTS: We utilised more than 10,000 compounds from the SuperTarget database with known inhibition values for twelve different anti-cancer targets. We performed all-against-all comparisons resulting in 2D similarity landscapes. Among the regions with low 2D similarity scores are inhibitors of vascular endothelial growth factor receptor (VEGFR) and inhibitors of poly ADP-ribose polymerase (PARP). To demonstrate that 3D landscape comparison can identify similarities, which are untraceable in 2D similarity comparisons, we analysed this region in more detail. This 3D analysis showed the unexpected structural similarity between inhibitors of VEGFR and inhibitors of PARP. Among the VEGFR inhibitors that show similarities to PARP inhibitors was Vatalanib, an oral "multi-targeted" small molecule protein kinase inhibitor being studied in phase-III clinical trials in cancer therapy. An in silico docking simulation and an in vitro HT universal colorimetric PARP assay confirmed that the VEGFR inhibitor Vatalanib exhibits off-target activity as a PARP inhibitor, broadening its mode of action. CONCLUSION: In contrast to the 2D-similarity search, the 3D-similarity landscape comparison identifies new functions and side effects of the known VEGFR inhibitor Vatalanib.

BACKGROUND: Despite progress in conventional cancer therapies, cancer is still one of the leading causes of death in industrial nations. Therefore, an urgent need of progress in fighting cancer remains. A promising alternative to conventional methods is immune therapy. This relies on the fact that low-immunogenic tumours can be eradicated if an immune response against them is induced. Peptide vaccination is carried out by injecting tumour peptides into a patient to trigger a specific immune response against the tumour in its entirety. However, peptide vaccination is a highly complicated treatment and currently many factors like the optimal number of epitopes are not known precisely. Therefore, it is necessary to evaluate how certain parameters influence the therapy. RESULTS: We present the VaccImm Server that allows users to simulate peptide vaccination in cancer therapy. It uses an agent-based model that simulates peptide vaccination by explicitly modelling the involved cells (immune system and cancer) as well as molecules (antibodies, antigens and semiochemicals). As a new feature, our model uses real amino acid sequences to represent molecular binding sites of relevant immune cells. The model is used to generate detailed statistics of the population sizes and states of the single cell types over time. This makes the VaccImm web server well suited to examine the parameter space of peptide vaccination in silico. VaccImm is publicly available without registration on the web at http://bioinformatics.charite.de/vaccimm; all major browsers are supported. CONCLUSIONS: The VaccImm Server provides a convenient way to analyze properties of peptide vaccination in cancer therapy. Using the server, we could gain interesting insights into peptide vaccination that reveal the complex and patient-specific nature of peptide vaccination.

Abstract Summary: Cobweb is a Java applet for real-time network visualization; its strength lies in enabling the interactive exploration of networks. Therefore, it allows new nodes to be interactively added to a network by querying a database on a server. The network constantly rearranges to provide the most meaningful topological view. Availability: Cobweb is available under the GPLv3 and may be freely downloaded at http://bioinformatics.charite.de/cobweb. Contact: [email protected] Supplementary Information: Supplementary data are available at Bioinformatics online.

BACKGROUND: Various experimental techniques yield peptides that are biologically active but have unfavourable pharmacological properties. The design of structurally similar organic compounds, i.e. peptide mimetics, is a challenging field in medicinal chemistry. RESULTS: SuperMimic identifies compounds that mimic parts of a protein, or positions in proteins that are suitable for inserting mimetics. The application provides libraries that contain peptidomimetic building blocks on the one hand and protein structures on the other. The search for promising peptidomimetic linkers for a given peptide is based on the superposition of the peptide with several conformers of the mimetic. New synthetic elements or proteins can be imported and used for searching. CONCLUSION: We present a graphical user interface for finding peptide mimetics that can be inserted into a protein or for fitting small molecules into a protein. Using SuperMimic, promising locations in proteins for the insertion of mimetics can be found quickly and conveniently.

BACKGROUND: The increasing number of known protein structures provides valuable information about pharmaceutical targets. Drug binding sites are identifiable and suitable lead compounds can be proposed. The flexibility of ligands is a critical point for the selection of potential drugs. Since computed 3D structures of millions of compounds are available, the knowledge of their binding conformations would be a great benefit for the development of efficient screening methods. RESULTS: Integration of two public databases allowed superposition of conformers for 193 approved drugs with 5507 crystallised target-bound counterparts. The generation of 9600 drug conformers using an atomic force field was carried out to obtain an optimal coverage of the conformational space. Bioactive conformations are best described by a conformational ensemble: half of all drugs exhibit multiple active states, distributed over the entire range of the reachable energy and conformational space.A number of up to 100 conformers per drug enabled us to reproduce the bound states within a similarity threshold of 1.0 angstroms in 70% of all cases. This fraction rises to about 90% for smaller or average sized drugs. CONCLUSION: Single drugs adopt multiple bioactive conformations if they interact with different target proteins. Due to the structural diversity of binding sites they adopt conformations that are distributed over a broad conformational space and wide energy range. Since the majority of drugs is well represented by a predefined low number of conformers (up to 100) this procedure is a valuable method to compare compounds by three-dimensional features or for fast similarity searches starting with pharmacophores. The underlying 9600 generated drug conformers are downloadable from the Super Drug Web site 1. All superpositions are visualised at the same source. Additional conformers (110,000) of 2400 classified WHO-drugs are also available.

MOTIVATION: Different resources exist for experimentally determined and computed three-dimensional (3D)-structures of low molecular weight structures but for approved drugs, no free, publicly accessible source of 3D-structures and conformers is available. Furthermore, for selection purposes or for correlation of structural similarity with medical application, the assignment of the Anatomical Therapeutic Chemical (ATC) classification codes to each structure according to the WHO-scheme would be desirable. RESULTS: The database contains approximately 2500 3D-structures of active ingredients of essential marketed drugs. To account for structural flexibility they are represented by 10(5) structural conformers. Here we present a web-query system enabling searches for drug name, synonyms, trade name, trivial name, formula, CAS-number, ATC-code etc. 2D-similarity screening (Tanimoto coefficients) and an automatic 3D-superposition procedure based on conformational representation are implemented. Drug structures above a similarity threshold as well as superimposed conformers can be retrieved in the mol- file format via a graphical interface. AVAILABILITY: For academic use the system is accessible at http://bioinf.charite.de/superdrug. The retrieval system requires the free browser-plugin 'chime' from MDL for visualization.

BACKGROUND: The determination of protein surfaces and the detection of binding sites are essential to our understanding of protein-protein interactions. Such binding sites can be characterised as linear and non-linear, the non-linear sites being prevailant. Conventional mapping techniques with arrays of synthetic peptides have limitations with regard to the location of discontinuous or non-linear binding sites of proteins. RESULTS: We present a structure-based approach to the design of peptide libraries that mimic the whole surface or a particular region of a protein. Neighbouring sequence segments are linked by short spacers to conserve local conformation. To this end, we have developed SUPERFICIAL, a program that uses protein structures as input and generates library proposals consisting of linear and non-linear peptides. This process can be influenced by a graphical user interface at different stages, from the surface computation up to the definition of spatial regions. CONCLUSION: Based on 3D structures, SUPERFICIAL may help to negotiate some of the existing limitations, since binding sites consisting of several linear pieces can now be detected.

BACKGROUND: Currently, the PDB contains approximately 29,000 protein structures comprising over 70,000 experimentally determined three-dimensional structures of over 5,000 different low molecular weight compounds. Information about these PDB ligands can be very helpful in the field of molecular modelling and prediction, particularly for the prediction of protein binding sites and function. DESCRIPTION: Here we present an Internet accessible database delivering PDB ligands in the MDL Mol file format which, in contrast to the PDB format, includes information about bond types. Structural similarity of the compounds can be detected by calculation of Tanimoto coefficients and by three-dimensional superposition. Topological similarity of PDB ligands to known drugs can be assessed via Tanimoto coefficients. CONCLUSION: SuperLigands supplements the set of existing resources of information about small molecules bound to PDB structures. Allowing for three-dimensional comparison of the compounds as a novel feature, this database represents a valuable means of analysis and prediction in the field of biological and medical research.

BACKGROUND: Structural and functional research often requires the computation of sets of protein structures based on certain properties of the proteins, such as sequence features, fold classification, or functional annotation. Compiling such sets using current web resources is tedious because the necessary data are spread over many different databases. To facilitate this task, we have created COLUMBA, an integrated database of annotations of protein structures. DESCRIPTION: COLUMBA currently integrates twelve different databases, including PDB, KEGG, Swiss-Prot, CATH, SCOP, the Gene Ontology, and ENZYME. The database can be searched using either keyword search or data source-specific web forms. Users can thus quickly select and download PDB entries that, for instance, participate in a particular pathway, are classified as containing a certain CATH architecture, are annotated as having a certain molecular function in the Gene Ontology, and whose structures have a resolution under a defined threshold. The results of queries are provided in both machine-readable extensible markup language and human-readable format. The structures themselves can be viewed interactively on the web. CONCLUSION: The COLUMBA database facilitates the creation of protein structure data sets for many structure-based studies. It allows to combine queries on a number of structure-related databases not covered by other projects at present. Thus, information on both many and few protein structures can be used efficiently. The web interface for COLUMBA is available at http://www.columba-db.de.

MOTIVATION: The Dictionary of Interfaces in Proteins (DIP) is a database collecting the 3D structure of interacting parts of proteins that are called patches. It serves as a repository, in which patches similar to given query patches can be found. The computation of the similarity of two patches is time consuming and traversing the entire DIP requires some hours. In this work we address the question of how the patches similar to a given query can be identified by scanning only a small part of DIP. The answer to this question requires the investigation of the distribution of the similarity of patches. RESULTS: The score values describing the similarity of two patches can roughly be divided into three ranges that correspond to different levels of spatial similarity. Interestingly, the two iso-score lines separating the three classes can be determined by two different approaches. Applying a concept of the theory of random graphs reveals significant structural properties of the data in DIP. These can be used to accelerate scanning the DIP for patches similar to a given query. Searches for very similar patches could be accelerated by a factor of more than 25. Patches with a medium similarity could be found 10 times faster than by brute-force search.

MOTIVATION: There is no consensus in the literature about how the deepest portions of protein structures are packed. Using an improved Voronoi procedure, we calculate reference packing densities for different regions in the protein interior. Furthermore, we want to clarify where cavities are located. RESULTS: Sets of reference packing densities are provided for regions in proteins that differ in their distance to the surface and to internal cavities, supplementing previous data. Packing in the protein interior is tight but generally inhomogeneous. There are about 4.4 cavities per 100 amino acids in protein structures, they occur in all regions, most frequently in a depth of 2.5-3.6 A underneath the Connolly surface. However, the deepest protein regions have a lower mean packing density than circumjacent regions, because more contacts to cavities occur in the core. AVAILABILITY/SUPPLEMENTARY INFORMATION: Calculation software and detailed packing data are available on request.

MOTIVATION: Should reports on molecular mimicry in particular cases, e.g. responsible for cross-reactivity, be considered as accidental or as a general principle in protein evolution? To answer this question, two types of similarity have to be considered: those in homologues (synonyms) and resemblance between patches from unrelated proteins (homonyms). RESULTS: All interfaces from known protein structures were collected in a comprehensive data bank [Dictionary of Interfaces in Proteins (DIP)]. A fast, sequence-independent, three-dimensional superposition procedure was developed to search automatically for geometrically similar surface areas. Surprisingly, we found a large number of structurally similar interfaces on the surface of unrelated proteins. Even patches from different types of secondary structure were found resembling each other. The putative functional meaning of homonyms is demonstrated with striking examples.