Third party funded individual grant
Start date : 01.02.2027
End date : 31.01.2030
Website: https://www.dyhealthnet.ai/
The field of genome-wide association studies (GWAS) and population cohort studies has benefited from a growing ecosystem of platforms for explorative data analysis. However, existing solutions largely focus on visualization, single-trait post-GWAS analyses, or cohort-specific data access, and lack support for advanced downstream analyses, cross-phenotype investigations, and open-source deployment on user-provided data. In the first funding period, we started to address this gap by developing two complementary platforms: GNExT, tailored to GWAS summary statistics, and DyHealthNet, designed for individual-level cohort data. GNExT goes beyond established frameworks such as PheWeb by providing standardized Nextflow-based data processing pipelines and by enabling downstream systems- and network-medicine analyses, including in silico disease mechanism mining and drug repurposing. DyHealthNet builds upon this functionality to support static and context-specific dynamic association networks derived from raw genomic data, allowing stratified analyses with respect to covariates such as age, sex, or disease status. While both platforms were developed with generalizability in mind, they have so far been primarily tested on data from the Collaborative Health Research in South Tyrol (CHRIS) study. In the second funding phase, our overarching objective is to establish GNExT and DyHealthNet as mature and scalable data analysis platforms that provide low-effort access to state-of-the-art network medicine and machine learning methods for GWAS and population cohort studies and are systematically tested and validated for a wide usage beyond the CHRIS study data. To this end, the work packages WP1–WP4 will deliver methodological advances, including scalable pipelines for polygenic risk score computation and context-specific evaluation, differential analysis of association networks, representation learning models for semantic similarity search across participants, variants, and traits, and functional variant-to-gene mapping to enable gene-based downstream analyses also for variants that fall into non-coding regions of the DNA. These methods will be integrated into both platforms. Moreover, they will be disseminated independently as FAIR-compliant pipelines or software packages. Complementing the methodological work, WP5–WP7 will focus on software engineering and dissemination. This includes optimizing platform backends for scalability and maintainability, developing demonstrator use cases on independent cohorts to showcase analytic value and performance on large datasets beyond the CHRIS study data, and promoting adoption through hands-on training workshops. Together, these efforts will democratize access to advanced GWAS and cohort data analysis, particularly benefiting smaller-scale studies lacking resources to develop dedicated infrastructures.