OmicKriging

OmicKriging predicts complex traits by integrating genomic, transcriptomic, and epigenomic similarities using Kriging to translate omics-level similarity into phenotypic predictions.

Key Features:

• Integration of Omics-Level Data: Integrates genomic, transcriptomic (mRNA and microRNA), and epigenomic datasets to capture multifaceted contributors to complex traits.
• Use of Kriging Methodology: Applies Kriging from geostatistics to model spatial or temporal correlations in biological datasets and translate omics similarities into phenotypic predictions.
• Efficient Computational Framework: Incorporates prior knowledge about subsets of omics data to represent sparse and highly polygenic components while minimizing computational burden.
• Performance and Scalability: Demonstrated on seven disease datasets from the Wellcome Trust Case Control Consortium (WTCCC), integrating sparse and highly polygenic components and achieving performance comparable to complex Bayesian models at reduced computational cost.
• Enhanced Predictive Power through Data Integration: Combining mRNA and microRNA expression data improves prediction of cellular growth phenotypes and enhances clinical statin response prediction compared with single-dataset approaches.

Scientific Applications:

• Disease Risk Prediction: Integrates genetic and epigenetic information to predict individual susceptibility to diseases.
• Drug Response Profiling: Predicts individual responses to drugs, including clinical statin response, using integrated omics signatures.
• Phenotypic Analysis: Correlates omics-level variation with observable traits such as cellular growth phenotypes.

Methodology:

Integrates comprehensive omics datasets and applies Kriging to model spatial or temporal correlations between genetic, transcriptomic (mRNA and microRNA), and epigenomic similarities and phenotypes; incorporates prior knowledge about omics subsets to model sparse and highly polygenic components and was evaluated on seven WTCCC disease datasets with comparisons to complex Bayesian models.