Biology subjects

Queralt-Rosinach, N.

Publications and source records attributed to Queralt-Rosinach, N..

7 recordsLinked to original sources

bioRxiv2024

Knowledge Graphs and Explainable AI for Drug Repurposing on Rare Diseases

Artificial Intelligence (AI)-based drug repurposing is an emerging strategy to identify drug candidates to treat rare diseases. However, cutting-edge algorithms based on Deep Learning (DL) typically dont provide a human understandable explanation supporting their predictions. This is a problem because it hampers the biologists ability to decide which predictions are the most plausible drug candidates to test in costly lab experiments. In this study, we propose rd-explainer a novel AI drug repurposing method for rare diseases which obtains possible drug candidates together with human understandable explanations. The method is based on Graph Neural Network (GNN) technology and explanations were generated as semantic graphs using state-of-the-art eXplainable AI (XAI). The model learns features from current background knowledge on the target rare disease structured as a Knowledge Graph (KG), which integrates curated facts and their evidence on different biomedical entities such as symptoms, drugs, genes and ortholog genes. Our experiments demonstrate that our method has excellent performance that is superior to state-of-the-art models. We investigated the application of XAI on drug repurposing for rare diseases and we prove our method is capable of discovering plausible drug candidates based on testable explanations. The data and code are publicly available at https://github.com/PPerdomoQ/rare-disease-explainer. HighlightsO_LIWe demonstrated the use of graph-based explainable AI for drug repurposing on rare diseases to accelerate sound discovery of new therapies for this underrepresented group. C_LIO_LIWe developed rd-explainer for rare disease specific drug research for faster translation. It predicts drugs to treat symptoms/phenotypes, it is highly performant and novel candidates are plausible according to evidence in the scientific literature and clinical trials. Key is that it learns a GNN model that is trained on a knowledge graph built specifically for a rare disease. We provide rd-explainer code freely available for the community. C_LIO_LIrd-explainer is researcher-centric interpretable ML for hypothesis generation and lab-in-the-loop drug research. Explanations of predictions are semantic graphs in line with human reasoning. C_LIO_LIWe detected an effect of knowledge graph topology on explainability. This highlights the importance of knowledge representation for the drug repurposing task. C_LI

bioinformatics↗

bioRxiv2023

Collaborative network analysis for the interpretation of transcriptomics data in rare diseases, an application to Huntington's disease

BackgroundRare diseases may affect the quality of life of patients and in some cases be life-threatening. Therapeutic opportunities are often limited, in part because of the lack of understanding of the molecular mechanisms that can cause disease. This can be ascribed to the low prevalence of rare diseases and therefore the lower sample sizes available for research. A way to overcome this is to integrate experimental rare disease data with prior knowledge using network-based methods. Taking this one step further, we hypothesized that combining and analyzing the results from multiple network-based methods could provide data-driven hypotheses of pathogenicity mechanisms from multiple perspectives. ResultsWe analyzed a Huntingtons disease (HD) transcriptomics dataset using six network-based methods in a collaborative way. These methods either inherently reported enriched annotation terms or their results were fed into enrichment analyses. The resulting significantly enriched Reactome pathways were then summarized using the ontological hierarchy which allowed the integration and interpretation of outputs from multiple methods. Among the resulting enriched pathways, there are pathways that have been shown previously to be involved in HD and pathways whose direct contribution to disease pathogenesis remains unclear and requires further investigation. ConclusionsIn summary, our study shows that collaborative network analysis approaches are well-suited to study rare diseases, as they provide hypotheses for pathogenic mechanisms from multiple perspectives. Applying different methods to the same case study can uncover different disease mechanisms that would not be apparent with the application of a single method.

bioinformatics↗

bioRxiv2023

Integrative analysis of CAKUT multi-omics data

Congenital Anomalies of the Kidney and Urinary Tract (CAKUT) is the leading cause of childhood end-stage renal disease and a significant cause of chronic kidney disease in adults. Genetic and environmental factors are known to influence CAKUT development, but the currently known disease mechanism remains incomplete. Our goal is to identify affected pathways and networks in CAKUT, and thereby aid in getting a better understanding of its pathophysiology. Multi-omics experiments, including amniotic fluid miRNome, peptidome, and proteome analyses, can shed light on foetal kidney development in non-severe CAKUT patients compared to severe CAKUT cases. We performed FAIRification of these omics data sets to facilitate their integration with external data resources. Furthermore, we analysed and integrated the omics data sets using three different bioinformatics strategies. The three bioinformatics analyses provided complementary features, but all pointed towards an important role for collagen in CAKUT development. We published the three analysis strategies as containerized workflows. These workflows can be applied to other FAIR data sets and help gaining knowledge on other rare diseases.

systems biology↗

bioRxiv2021

Design and application of a knowledge network for automatic prioritization of drug mechanisms

MotivationDrug repositioning is an attractive alternative to de novo drug discovery due to reduced time and costs to bring drugs to market. Computational repositioning methods, particularly non-black-box methods that can account for and predict a drugs mechanism, may provide great benefit for directing future development. By tuning both data and algorithm to utilize relationships important to drug mechanisms, a computational repositioning algorithm can be trained to both predict and explain mechanistically novel indications. ResultsIn this work, we examined the 123 curated drug mechanism paths found in the drug mechanism database (DrugMechDB) and after identifying the most important relationships, we integrated 18 data sources to produce a heterogeneous knowledge graph, MechRepoNet, capable of capturing the information in these paths. We applied the Rephetio repurposing algorithm to MechRepoNet using only a subset of relationships known to be mechanistic in nature and found adequate predictive ability on an evaluation set with AUROC value of 0.83. The resulting repurposing model allowed us to prioritize paths in our knowledge graph to produce a predicted treatment mechanism. We found that DrugMechDB paths, when present in the network were rated highly among predicted mechanisms. We then demonstrated MechRepoNets ability to use mechanistic insight to identify a drugs mechanistic target, with a mean reciprocal rank of .525 on a test set of known drug-target interactions. Finally, we walked through a repurposing example of the anti-cancer drug imantinib for use in the treatment of asthma, to demonstrate this methods utility in providing mechanistic insight into repurposing predictions it provides. Availability and implementationThe Python code to reproduce the entirety of this analysis is available at: https://github.com/SuLab/MechRepoNet Contactasu@scripps.edu Supplementary informationSupplemental information is available at Bioinformatics online.

bioinformatics↗

bioRxiv2019

Wikidata as a FAIR knowledge graph for the life sciences

Wikidata is a community-maintained knowledge base that epitomizes the FAIR principles of Findability, Accessibility, Interoperability, and Reusability. Here, we describe the breadth and depth of biomedical knowledge contained within Wikidata, assembled from primary knowledge repositories on genomics, proteomics, genetic variants, pathways, chemical compounds, and diseases. We built a collection of open-source tools that simplify the addition and synchronization of Wikidata with source databases. We furthermore demonstrate several use cases of how the continuously updated, crowd-contributed knowledge in Wikidata can be mined. These use cases cover a diverse cross section of biomedical analyses, from crowdsourced curation of biomedical ontologies, to phenotype-based diagnosis of disease, to drug repurposing.

bioinformatics↗

bioRxiv2019

Structured Reviews for Data and Knowledge Driven Research

MotivationHypothesis generation is a critical step in research and a cornerstone in the rare disease field. Research is most efficient when those hypotheses are based on the entirety of knowledge known to date. Systematic review articles are commonly used in biomedicine to summarize existing knowledge and contextualize experimental data. But the information contained within review articles is typically only expressed as free-text, which is difficult to use computationally. Researchers struggle to navigate, collect and remix prior knowledge as it is scattered in several silos without seamless integration and access. This lack of a structured information framework hinders research by both experimental and computational scientists.\n\nResultsTo better organize knowledge and data, we built a structured review article that is specifically focused on NGLY1 Deficiency, an ultra-rare genetic disease first reported in 2012. We represented this structured review as a knowledge graph, and then stored this knowledge graph in a Neo4j database to simplify dissemination, querying, and visualization of the network. Relative to free-text, this structured review better promotes the principles of findability, accessibility, interoperability, and reusability (FAIR). In collaboration with domain experts in NGLY1 Deficiency, we demonstrate how this resource can improve the efficiency and comprehensiveness of hypothesis generation. We also developed a read-write interface that allows domain experts to contribute FAIR structured knowledge to this community resource. In contrast to traditional free-text review articles, this structured review exists as a living knowledge graph that is curated by humans and accessible to computational analyses. Finally, we have generalized this workflow into modular and repurposable components that can be applied to other domain areas. This NGLY1 Deficiency-focused network is publicly available at http://ngly1graph.org/.\n\nAvailability and implementationSource code and network data files are at: https://github.com/SuLab/ngly1-graph and https://github.com/SuLab/bioknowledge-reviewer.\n\nContactasu@scripps.edu

bioinformatics↗

bioRxiv2019

Time-resolved compound repositioning predictions on a texted-mined knowledge network

BackgroundComputational compound repositioning has the potential for identifying new uses for existing drugs, and new algorithms and data source aggregation strategies provide ever-improving results via in silico metrics. However, even with these advances, the number of compounds successfully repositioned via computational screening remains low. New strategies for algorithm evaluation that more accurately reflect the repositioning potential of a compound could provide a better target for future optimizations.\n\nResultsUsing a text-mined database, we applied a previously described network-based computational repositioning algorithm, yielding strong results via cross-validation, averaging 0.95 AUROC on test-set indications. The text-mined data was then used to build networks corresponding to different time-points in biomedical knowledge. Training the algorithm on contemporary indications and testing on future showed a marked reduction in performance, peaking in performance metrics with the 1985 network at an AUROC of .797. Examining performance reductions due to removal of specific types of relationships highlighted the importance of drug-drug and disease-disease similarity metrics. Using data from future timepoints, we demonstrate that further acquisition of these kinds of data may help improve computational results.\n\nConclusionsEvaluating a repositioning algorithm using indications unknown to input network better tunes its ability to find emerging drug indications, rather than finding those which have been withheld. Focusing efforts on improving algorithmic performance in a time-resolved paradigm may further improve computational repositioning predictions.

bioinformatics↗