Final Report Summary - HYPERGENES (European network for genetic-epidemiological studies: building a method to dissect complex genetic traits, using essential hypertension as a disease model)
The HYPERGENES project's activities are structured in three steps: discovery, validation, and dissemination and results exploitation to be realised in four years.
HYPERGENES discovery phase was performed during the first two years of the project and was focused on building the methodological and technical framework to support the genome wide association analysis, performed on 4000 Caucasian subjects recruited from historical well-characterised European cohorts.
The need of integrating the observations from different studies posed significant challenges which were faced through an integrated epidemiological and bioinformatics approach. The Biomedical information infrastructure (BII) was developed to support the entry, persistency and retrieval of data and knowledge relevant to EH, including clinical, environmental and genotypic data.
Genotyping have been performed on high throughput Illumina technologies, thanks to the coordinated efforts of the laboratories of UNIMI and UNIL.
Both classical and machine learning techniques were used for genetic analysis, to produce an enriched list of SNPs, that resulted associated with EH or TOD, or other endophenotypes relevant to hypertension. The conducted case-control association study lead to hundreds of significant associations, which were only partially overlapping with the results of previous studies.
The best SNPs resulting associated to EH and TODs in the discovery sample together with candidate SNPs well known as being associated to the phenotypic trait of interest were used to build a custom Illumina iSelect HD chip, including 15 000 SNPs. Such tool was used to validate the results obtained in the discovery phase in an additional independent sample of 8000 subjects.
Regions showing most promising results were sequenced in order to obtain a more detailed comprehension of the nucleotide sequence of each region. Sequencing was performed on 92 subjects.
A risk prediction algorithm for EH and TOD was then developed.
A Lab-on-chip (LoC) meant to test the most promising SNPs found associated with microalbuminuria was developed. The LoC was therefore cross-validated on 100 samples.
The developed BII manifests the commonalities among the dozens of HYPERGENES cohorts for multi-cohort analysis, while preserving the disparities and allowing researchers to access the original cohort data. Atop of the BII two machine learning tools were developed: the 'bioclinical data mining tool', which is a general purpose algorithms built into a generic framework for streaming data, and the 'SNP weighting tool', an algorithm specifically developed for utilising the existing knowledgebase for SNP association analysis. These tools are available online for HYPERGENES partners.
These tools concurred to the development of a risk prediction disease model. Given a set of known EH and TODs risk factors for which progressive data are available, the model allows to predict the future clinical range of each parameter, exploring also the contribution of the genomic data to the prediction. To further integrate the model with genomic information, a previously developed gene network was used. Each gene in every pathway was represented by its strongest associated SNP. A good prediction performance was demonstrated for measures having sufficient samples (over thousand).
Moreover, HYPERGENES performed a pathway analysis using the 29 variants identified in ICPB-GWAS, a recent genome-wide association study involving 200 000 individuals, and HYPERGENES results with the aim to verify a common pathway between the two studies. HYPERGENES added a further relevant player, NOS3, within the vasodilator pathway for blood pressure regulation.
HYPERGENES discovery phase was performed during the first two years of the project and was focused on building the methodological and technical framework to support the genome wide association analysis, performed on 4000 Caucasian subjects recruited from historical well-characterised European cohorts.
The need of integrating the observations from different studies posed significant challenges which were faced through an integrated epidemiological and bioinformatics approach. The Biomedical information infrastructure (BII) was developed to support the entry, persistency and retrieval of data and knowledge relevant to EH, including clinical, environmental and genotypic data.
Genotyping have been performed on high throughput Illumina technologies, thanks to the coordinated efforts of the laboratories of UNIMI and UNIL.
Both classical and machine learning techniques were used for genetic analysis, to produce an enriched list of SNPs, that resulted associated with EH or TOD, or other endophenotypes relevant to hypertension. The conducted case-control association study lead to hundreds of significant associations, which were only partially overlapping with the results of previous studies.
The best SNPs resulting associated to EH and TODs in the discovery sample together with candidate SNPs well known as being associated to the phenotypic trait of interest were used to build a custom Illumina iSelect HD chip, including 15 000 SNPs. Such tool was used to validate the results obtained in the discovery phase in an additional independent sample of 8000 subjects.
Regions showing most promising results were sequenced in order to obtain a more detailed comprehension of the nucleotide sequence of each region. Sequencing was performed on 92 subjects.
A risk prediction algorithm for EH and TOD was then developed.
A Lab-on-chip (LoC) meant to test the most promising SNPs found associated with microalbuminuria was developed. The LoC was therefore cross-validated on 100 samples.
The developed BII manifests the commonalities among the dozens of HYPERGENES cohorts for multi-cohort analysis, while preserving the disparities and allowing researchers to access the original cohort data. Atop of the BII two machine learning tools were developed: the 'bioclinical data mining tool', which is a general purpose algorithms built into a generic framework for streaming data, and the 'SNP weighting tool', an algorithm specifically developed for utilising the existing knowledgebase for SNP association analysis. These tools are available online for HYPERGENES partners.
These tools concurred to the development of a risk prediction disease model. Given a set of known EH and TODs risk factors for which progressive data are available, the model allows to predict the future clinical range of each parameter, exploring also the contribution of the genomic data to the prediction. To further integrate the model with genomic information, a previously developed gene network was used. Each gene in every pathway was represented by its strongest associated SNP. A good prediction performance was demonstrated for measures having sufficient samples (over thousand).
Moreover, HYPERGENES performed a pathway analysis using the 29 variants identified in ICPB-GWAS, a recent genome-wide association study involving 200 000 individuals, and HYPERGENES results with the aim to verify a common pathway between the two studies. HYPERGENES added a further relevant player, NOS3, within the vasodilator pathway for blood pressure regulation.