DISNET: a framework for extracting phenotypic disease information from public sources

Identifying symptom etiologies using syntactic patterns and large language models

Differential diagnosis is a crucial aspect of medical practice, as it guides clinicians to accurate diagnoses and effective treatment plans. Traditional resources, such as medical books and services like UpToDate, are constrained by manual curation, potentially missing out on novel or less common findings. This paper introduces and analyzes two novel methods to mine etiologies from scientific literature. The first method employs a traditional Natural Language Processing (NLP) approach based on syntactic patterns. By using a novel application of human-guided pattern bootstrapping patterns are derived quickly, and symptom etiologies are extracted with significant coverage. The second method utilizes generative models, specifically GPT-4, coupled with a fact verification pipeline, marking a pioneering application of generative techniques in etiology extraction. Analyzing this second method shows that while it is highly precise, it offers lesser coverage compared to the syntactic approach. Importantly, combining both methodologies yields synergistic outcomes, enhancing the depth and reliability of etiology mining.

Protein sequence analysis in the context of drug repurposing

MOTIVATION

Drug repurposing speeds up the development of new treatments, being less costly, risky, and time consuming than de novo drug discovery. There are numerous biological elements that contribute to the development of diseases and, as a result, to the repurposing of drugs.

METHODS

In this article, we analysed the potential role of protein sequences in drug repurposing scenarios. For this purpose, we embedded the protein sequences by performing four state of the art methods and validated their capacity to encapsulate essential biological information through visualization. Then, we compared the differences in sequence distance between protein-drug target pairs of drug repurposing and non - drug repurposing data. Thus, we were able to uncover patterns that define protein sequences in repurposing cases.

RESULTS

We found statistically significant sequence distance differences between protein pairs in the repurposing data and the rest of protein pairs in non-repurposing data. In this manner, we verified the potential of using numerical representations of sequences to generate repurposing hypotheses in the future.

Exploring NCATS in-house biomedical data for evidence-based drug repurposing

Drug repurposing is a strategy for identifying new uses of approved or investigational drugs that are outside the scope of the original medical indication. Even though many repurposed drugs have been found serendipitously in the past, the increasing availability of large volumes of biomedical data has enabled more systemic, data-driven approaches for drug candidate identification. At National Center of Advancing Translational Sciences (NCATS), we invent new methods to generate new data and information publicly available to spur innovation and scientific discovery. In this study, we aimed to explore and demonstrate biomedical data generated and collected via two NCATS research programs, the Toxicology in the 21st Century program (Tox21) and the Biomedical Data Translator (Translator) for the application of drug repurposing. These two programs provide complementary types of biomedical data from uncovering underlying biological mechanisms with bioassay screening data from Tox21 for chemical clustering, to enrich clustered chemicals with scientific evidence mined from the Translator towards drug repurposing. 129 chemical clusters have been generated and three of them have been further investigated for drug repurposing candidate identification, which is detailed as case studies.

Identifying patterns to uncover the importance of biological pathways on known drug repurposing scenarios

BACKGROUND

Drug repurposing plays a significant role in providing effective treatments for certain diseases faster and more cost-effectively. Successful repurposing cases are mostly supported by a classical paradigm that stems from de novo drug development. This paradigm is based on the "one-drug-one-target-one-disease" idea. It consists of designing drugs specifically for a single disease and its drug's gene target. In this article, we investigated the use of biological pathways as potential elements to achieve effective drug repurposing.

METHODS

Considering a total of 4214 successful cases of drug repurposing, we identified cases in which biological pathways serve as the underlying basis for successful repurposing, referred to as DREBIOP. Once the repurposing cases based on pathways were identified, we studied their inherent patterns by considering the different biological elements associated with this dataset, as well as the pathways involved in these cases. Furthermore, we obtained gene-disease association values to demonstrate the diminished significance of the drug's gene target in these repurposing cases. To achieve this, we compared the values obtained for the DREBIOP set with the overall association values found in DISNET, as well as with the drug's target gene (DREGE) based repurposing cases using the Mann-Whitney U Test.

RESULTS

A collection of drug repurposing cases, known as DREBIOP, was identified as a result. DREBIOP cases exhibit distinct characteristics compared with DREGE cases. Notably, DREBIOP cases are associated with a higher number of biological pathways, with Vitamin D Metabolism and ACE inhibitors being the most prominent pathways. Additionally, it was observed that the association values of GDAs in DREBIOP cases were significantly lower than those in DREGE cases (p-value < 0.05).

CONCLUSIONS

Biological pathways assume a pivotal role in drug repurposing cases. This investigation successfully revealed patterns that distinguish drug repurposing instances associated with biological pathways. These identified patterns can be applied to any known repurposing case, enabling the detection of pathway-based repurposing scenarios or the classical paradigm.

Uncovering hidden therapeutic indications through drug repurposing with graph neural networks and heterogeneous data

Drug repurposing has gained the attention of many in the recent years. The practice of repurposing existing drugs for new therapeutic uses helps to simplify the drug discovery process, which in turn reduces the costs and risks that are associated with de novo development. Representing biomedical data in the form of a graph is a simple and effective method to depict the underlying structure of the information. Using deep neural networks in combination with this data represents a promising approach to address drug repurposing. This paper presents BEHOR a more comprehensive version of the REDIRECTION model, which was previously presented. Both versions utilize the DISNET biomedical graph as the primary source of information, providing the model with extensive and intricate data to tackle the drug repurposing challenge. This new version's results for the reported metrics in the RepoDB test are 0.9604 for AUROC and 0.9518 for AUPRC. Additionally, a discussion is provided regarding some of the novel predictions to demonstrate the reliability of the model. The authors believe that BEHOR holds promise for generating drug repurposing hypotheses and could greatly benefit the field.

Exploring NCATS In-House Biomedical Data for Evidence-based Drug Repurposing

Drug repurposing is a strategy for identifying new uses of approved or investigational drugs that are outside the scope of the original medical indication. Even though many repurposed drugs have been found serendipitously in the past, the increasing availability of large volumes of biomedical data has enabled more systemic, data-driven approaches for drug candidate identification. At National Center of Advancing Translational Sciences (NCATS), we invent new methods to generate new data and information publicly available to spur innovation and scientific discovery. In this study, we aimed to explore and demonstrate biomedical data generated and collected via two NCATS research programs, the Toxicology in the 21st Century program (Tox21) and the Biomedical Data Translator (Translator) for the application of drug repurposing. These two programs provide complementary types of biomedical data from uncovering underlying biological mechanisms with bioassay screening data from Tox21 for chemical clustering, to enrich clustered chemicals with scientific evidence mined from the Translator towards drug repurposing. 129 chemical clusters have been generated and three of them have been further investigated for drug repurposing candidate identification, which is detailed as case studies.

Repositioning Drugs for Rare Diseases Based on Biological Features and Computational Approaches

Rare diseases are a group of uncommon diseases in the world population. To date, about 7000 rare diseases have been documented. However, most of them do not have a known treatment. As a result of the relatively low demand for their treatments caused by their scarce prevalence, the pharmaceutical industry has not sufficiently encouraged the research to develop drugs to treat them. This work aims to analyse potential drug-repositioning strategies for this kind of disease. Drug repositioning seeks to find new uses for existing drugs. In this context, it seeks to discover if rare diseases could be treated with medicines previously indicated to heal other diseases. Our approaches tackle the problem by employing computational methods that calculate similarities between rare and non-rare diseases, considering biological features such as genes, proteins, and symptoms. Drug candidates for repositioning will be checked against clinical trials found in the scientific literature. In this study, 13 different rare diseases have been selected for which potential drugs could be repositioned. By verifying these drugs in the scientific literature, successful cases were found for 75% of the rare diseases studied. The genetic associations and phenotypical features of the rare diseases were examined. In addition, the verified drugs were classified according to the anatomical therapeutic chemical (ATC) code to highlight the types with a higher predisposition to be repositioned. These promising results open the door for further research in this field of study.

Biomedical data, computational methods and tools for evaluating disease-disease associations

In recent decades, exploring potential relationships between diseases has been an active research field. With the rapid accumulation of disease-related biomedical data, a lot of computational methods and tools/platforms have been developed to reveal intrinsic relationship between diseases, which can provide useful insights to the study of complex diseases, e.g. understanding molecular mechanisms of diseases and discovering new treatment of diseases. Human complex diseases involve both external phenotypic abnormalities and complex internal molecular mechanisms in organisms. Computational methods with different types of biomedical data from phenotype to genotype can evaluate disease-disease associations at different levels, providing a comprehensive perspective for understanding diseases. In this review, available biomedical data and databases for evaluating disease-disease associations are first summarized. Then, existing computational methods for disease-disease associations are reviewed and classified into five groups in terms of the usages of biomedical data, including disease semantic-based, phenotype-based, function-based, representation learning-based and text mining-based methods. Further, we summarize software tools/platforms for computation and analysis of disease-disease associations. Finally, we give a discussion and summary on the research of disease-disease associations. This review provides a systematic overview for current disease association research, which could promote the development and applications of computational methods and tools/platforms for disease-disease associations.

Classifying diseases by using biological features to identify potential nosological models

Established nosological models have provided physicians an adequate enough classification of diseases so far. Such systems are important to correctly identify diseases and treat them successfully. However, these taxonomies tend to be based on phenotypical observations, lacking a molecular or biological foundation. Therefore, there is an urgent need to modernize them in order to include the heterogeneous information that is produced in the present, as could be genomic, proteomic, transcriptomic and metabolic data, leading this way to more comprehensive and robust structures. For that purpose, we have developed an extensive methodology to analyse the possibilities when it comes to generate new nosological models from biological features. Different datasets of diseases have been considered, and distinct features related to diseases, namely genes, proteins, metabolic pathways and genetical variants, have been represented as binary and numerical vectors. From those vectors, diseases distances have been computed on the basis of several metrics. Clustering algorithms have been implemented to group diseases, generating different models, each of them corresponding to the distinct combinations of the previous parameters. They have been evaluated by means of intrinsic metrics, proving that some of them are highly suitable to cover new nosologies. One of the clustering configurations has been deeply analysed, demonstrating its quality and validity in the research context, and further biological interpretations have been made. Such model was particularly generated by OPTICS clustering algorithm, by studying the distance between diseases based on gene sharedness and following cosine index metric. 729 clusters were formed in this model, which obtained a Silhouette coefficient of 0.43.

Integrating heterogeneous data to facilitate COVID-19 drug repurposing

In the COVID-19 pandemic, drug repositioning has presented itself as an alternative to the time-consuming process of generating new drugs. This review describes a drug repurposing process that is based on a new data-driven approach: we put forward five information paths that associate COVID-19-related genes and COVID-19 symptoms with drugs that directly target these gene products, that target the symptoms or that treat diseases that are symptomatically or genetically similar to COVID-19. The intersection of the five information paths results in a list of 13 drugs that we suggest as potential candidates against COVID-19. In addition, we have found information in published studies and in clinical trials that support the therapeutic potential of the drugs in our final list.

A data-driven methodology towards evaluating the potential of drug repurposing hypotheses

Drug repurposing has become a widely used strategy to accelerate the process of finding treatments. While classical de novo drug development involves high costs, risks, and time-consuming paths, drug repurposing allows to reuse already-existing and approved drugs for new indications. Numerous research has been carried out in this field, both in vitro and in silico. Computational drug repurposing methods make use of modern heterogeneous biomedical data to identify and prioritize new indications for old drugs. In the current paper, we present a new complete methodology to evaluate new potentially repurposable drugs based on disease-gene and disease-phenotype associations, identifying significant differences between repurposing and non-repurposing data. We have collected a set of known successful drug repurposing case studies from the literature and we have analysed their dissimilarities with other biomedical data not necessarily participating in repurposing processes. The information used has been obtained from the DISNET platform. We have performed three analyses (at the genetical, phenotypical, and categorization levels), to conclude that there is a statistically significant difference between actual repurposing-related information and non-repurposing data. The insights obtained could be relevant when suggesting new potential drug repurposing hypotheses.

DisMaNET: A network-based tool to cross map disease vocabularies

BACKGROUND AND OBJECTIVES

The growing integration of healthcare sources is improving our understanding of diseases. Cross-mapping resources such as UMLS play a very important role in this area, but their coverage is still incomplete. With the aim to facilitate the integration and interoperability of biological, clinical and literary sources in studies of diseases, we built DisMaNET, a system to cross-map terms from disease vocabularies by leveraging the power and interpretability of network analysis.

METHODS

First, we collected and normalized data from 8 disease vocabularies and mapping sources to generate our datasets. Next, we built DisMaNET by integrating the generated datasets into a Neo4j graph database. Then we exploited the query mechanisms of Neo4j to cross-map disease terms of different vocabularies with a relevance score metric and contrasted the results with some state-of-the-art solutions. Finally, we made our system publicly available for its exploitation and evaluation both through a graphical user interface and REST APIs.

RESULTS

DisMaNET contains almost half a million nodes and near nine hundred thousand edges, including hierarchical and mapping relationships. Its query capabilities enabled the detection of connections between disease vocabularies that are not present in major mapping sources such as UMLS and the Disease Ontology, even for rare diseases. Furthermore, DisMaNET was capable of obtaining more than 80% of the mappings with UMLS reported in MonDO and DisGeNET, and it was successfully exploited to resolve the missing mappings in the DISNET project.

CONCLUSIONS

DisMaNET is a powerful, intuitive and publicly available system to cross-map terms from different disease vocabularies. Our study proves that it is a competitive alternative to existing mapping systems, incorporating the potential of network analysis and the interpretability of the results through a visual interface as its main advantages. Expansion with new sources, versioning and the improvement of the search and scoring algorithms are envisioned as future lines of work.

Leveraging network analysis to evaluate biomedical named entity recognition tools

The ever-growing availability of biomedical text sources has resulted in a boost in clinical studies based on their exploitation. Biomedical named-entity recognition (bio-NER) techniques have evolved remarkably in recent years and their application in research is increasingly successful. Still, the disparity of tools and the limited available validation resources are barriers preventing a wider diffusion, especially within clinical practice. We here propose the use of omics data and network analysis as an alternative for the assessment of bio-NER tools. Specifically, our method introduces quality criteria based on edge overlap and community detection. The application of these criteria to four bio-NER solutions yielded comparable results to strategies based on annotated corpora, without suffering from their limitations. Our approach can constitute a guide both for the selection of the best bio-NER tool given a specific task, and for the creation and validation of novel approaches.

Developing an automated mechanism to identify medical articles from wikipedia for knowledge extraction

Wikipedia contains rich biomedical information that can support medical informatics studies and applications. Identifying the subset of medical articles of Wikipedia has many benefits, such as facilitating medical knowledge extraction, serving as a corpus for language modeling, or simply making the size of data easy to work with. However, due to the extremely low prevalence of medical articles in the entire Wikipedia, articles identified by generic text classifiers would be bloated by irrelevant pages. To control the false discovery rate while maintaining a high recall, we developed a mechanism that leverages the rich page elements and the connected nature of Wikipedia and uses a crawling classification strategy to achieve accurate classification. Structured assertional knowledge in Infoboxes and Wikidata items associated with the identified medical articles were also extracted. This automatic mechanism is aimed to run periodically to update the results and share them with the informatics community.