Biology subjects

Antal, P.

Publications and source records attributed to Antal, P..

6 recordsLinked to original sources

bioRxiv2024

GNN4DM: A Graph Neural Network-based method to identify overlapping functional disease modules

MotivationIdentifying disease modules within molecular interaction networks is an essential exploratory step in computational biology, offering insights into disease mechanisms and potential therapeutic targets. Traditional methods often struggle with the inherent complexity and overlapping nature of biological networks, and they are limited in effectively leveraging the vast amount of available genomic data and biological knowledge. This limitation underscores the need for more effective, automated approaches to integrate these rich data sources. ResultsIn this work, we propose GNN4DM, a novel graph neural network-based structured model that automates the discovery of overlapping functional disease modules. GNN4DM effectively integrates network topology with genomic data to learn the representations of the genes corresponding to functional modules and align these with known biological pathways for enhanced interpretability. Following the DREAM benchmark evaluation setting and extending with three independent data sources (GWAS Atlas, FinnGen, and DisGeNET), we show that GNN4DM performs better than several state-of-the-art methods in detecting biologically meaningful modules. Furthermore, we highlight the methods applicability through its discovery of two novel multimorbidity-modules significantly enriched in a wide array of apparently unrelated diseases. Availability and ImplementationSource code, all training data, and all identified disease modules are freely available for download at https://github.com/gezsi/gnn4dm. GNN4DM was implemented in Python. Contactgezsi.andras@vik.bme.hu

systems biology↗

bioRxiv2023

Towards explainable interaction prediction: Embedding biological hierarchies into hyperbolic interaction space

Given the prolonged timelines and high costs associated with traditional approaches, accelerating drug development is crucial. Computational methods, particularly drug-target interaction prediction, have emerged as efficient tools, yet the explainability of machine learning models remains a challenge. Our work aims to provide more interpretable interaction prediction models using similarity-based prediction in a latent space aligned to biological hierarchies. We investigated integrating drug and protein hierarchies into a joint-embedding drug-target latent space via embedding regularization by conducting a comparative analysis between models employing traditional flat Euclidean vector spaces and those utilizing hyperbolic embeddings. Besides, we provided a latent space analysis as an example to show how we can gain visual insights into the trained model with the help of dimensionality reduction. Our results demonstrate that hierarchy regularization improves interpretability without compromising predictive performance. Furthermore, integrating hyperbolic embeddings, coupled with regularization, enhances the quality of the embedded hierarchy trees. Our approach enables a more informed and insightful application of interaction prediction models in drug discovery by constructing an interpretable hyperbolic latent space, simultaneously incorporating drug and target hierarchies and pairing them with available interaction information. Moreover, compatible with pairwise methods, the approach allows for additional transparency through existing explainable AI solutions.

bioinformatics↗

bioRxiv2023

Transfer learning in heterogeneous drug-target interaction predictions using federated boosting

In multitask federated learning, when small amounts of data are available, it can be harder to achieve proper predictive performance, especially if the clients tasks are different. However, task heterogeneity is common in modern Drug-Target interaction (DTI) prediction problems. As the data available for DTI tasks are sparse, it can be challenging for clients to synchronize the tasks used for training. In our method, we used boosting to enhance transfer in the multitask scenario and adapted it to a federated environment, allowing clients to train models without having to agree on the output dimensions. Boosting uses adaptive weighting of the data to train an ensemble of predictors. Weighting data boosting can induce the selection of important tasks when shaping a models latent representation. This way boosting contributes to the weighting of tasks on a client level and enhances transfer, while traditional federated algorithms can be used on a global level. We evaluate our results extensively on the tyrosine kinase assays of the KIBA data set to get a clear picture of connections between boosting federated learning and transfer learning.

bioinformatics↗

bioRxiv2021

Gene co-expression analyses of health(span) across multiple species

Health(span)-related gene clusters/modules were recently identified based on knowledge about the cross-species genetic basis of health, to interpret transcriptomic datasets describing health-related interventions. However, the cross-species comparison of health-related observations reveals a lot of heterogeneity, not least due to widely varying health(span) definitions and study designs, posing a challenge for the exploration of conserved healthspan modules and, specifically, their transfer across species. To improve the identification and exploration of conserved/transferable healthspan modules, here we apply an established workflow based on gene co-expression network analyses employing GEO/ArrayExpress data for human and animal models, and perform a comprehensive meta-analysis of the resulting modules related to health(span), yielding a small set of health(span) candidate genes, backed by the literature. For each experiment, WGCNA (weighted gene correlation network analysis) was thus used to infer modules of genes which correlate in their expression with a "health phenotype score" and to determine the most-connected (hub) genes for each such module, and their interactions. After mapping these hub genes to their human orthologs, 12 health(span) genes were identified in at least two species (ACTN3, ANK1, MRPL18, MYL1, PAXIP1, PPP1CA, SCN3B, SDCBP, SKIV2L, TUBG1, TYROBP, WIPF1), for which enrichment analysis by g:profiler finds an association with actin filament-based movement and associated organelles as well as muscular structures. We conclude that a meta-study of hub genes from co-expression network analyses for the complex phenotype health(span), across multiple species, can yield molecular-mechanistic insights and can direct experimentalists to further investigate the contribution of individual genes and their interactions to health(span).

genomics↗

bioRxiv2020

Educational attainment polygenic scores in Hungary: evidence for validity and a historical gene-environment interaction

Educational attainment is a substantially heritable trait, and it has recently been linked to specific genetic variants by genome-wide association studies (GWASs). However, the effects of such genetic variants are expected to vary across environments, including countries and historical eras. We used polygenic scores (PGSs) to assess molecular genetic effects on educational attainment in Hungary, a country in the Central Eastern European region where behavioral genetic studies are in general scarce and molecular genetic studies of educational attainment have not been previously published. We found that the PGS is significantly associated with highest educational level attained as well as the number of years in education in a sample of Hungarian volunteers (N=829). In an English (N=976) comparison sample with identical measurement protocols the same PGS had a stronger association with educational level, but not with years in education. In line with previous Estonian findings, we found higher PGS effect sizes in Hungarian, but not in English participants who attended higher education after the fall of Communism, although we lacked statistical power for this effect to reach significance. Our results provide evidence that polygenic scores for educational attainment are valid in diverse European populations.

genetics↗

bioRxiv2018

Healthspan pathway maps in C. elegans and humans highlight transcription, prolifera-tion/biosynthesis and lipids

The molecular basis of aging and of aging-associated diseases is being unraveled at an increasing pace. An extended healthspan, and not merely an extension of lifespan, has become the aim of medical practice. However, a precise definition of health and healthspan is not straightforward, and the causal molecular basis of health \"per se\" is largely unknown. Here, we define health based on the absence of diseases and dysfunctions. Based on an extensive review of the literature, in particular for humans and C. elegans, we compile a list of features of health and of the genes associated with them. Clusters of these genes based on molecular interaction data give rise to maps of healthspan pathways for humans, featuring the themes transcription initiation, proliferation and cholesterol/lipid processing, and for C. elegans, featuring the themes immune response, mitochondrion and biosynthesis based on genetic and compound intervention data, and lipids, biosynthesis and transcription based on WormBase compound intervention data. Overlaying healthspan-related gene expression data (describing effects of metabolic intervention associated with improvements in health) onto the aforementioned healthspan pathway maps, we observe the downregulation of Notch signalling in humans and of proliferation/cell-cycle in C. elegans. The former reflects the proinflammatory role of the Notch pathway. We identify transcription, proliferation/biosynthesis and lipids as a common theme on the annotation level, and proliferation-related kinases on the gene/protein level. Our literature-based data corpus, including visualization, is available as a reference for future investigations, at http://www.h2020awe.eu/index.php/pathways/.

bioinformatics↗