dynBGP

dynBGP estimates dynamic single-nucleotide polymorphism (SNP)-heritability over time using Bayesian Gaussian process models to quantify temporal genetic influence on traits.

Key Features:

• Dynamic SNP-Heritability Estimation: Estimates SNP-heritability as a function of time for longitudinal phenotypes.
• Bayesian Gaussian Process Models: Uses Bayesian Gaussian processes to model time-dependent variance components and heritability.
• Tuning-Free Approach: Operates without manual hyperparameter tuning.
• Full Uncertainty Quantification: Employs modern Markov Chain Monte Carlo (MCMC) techniques to provide 95% credible intervals for parameter estimates, with intervals reported as narrower than traditional two-stage methods.
• Performance and Scalability: Demonstrates superior performance relative to random regression models implemented in MTG2 and BLUPF90, and handles datasets with up to 1000 time points and simulations involving tens of thousands of individuals.
• Implementation: Implemented in C++ with integration into R workflows.

Scientific Applications:

• Longitudinal Genetic Studies: Applied to studies of traits that change over time, including growth and development analyses in developmental biology and epidemiological research.
• Quantitative Genetics: Used to characterize the temporal genetic architecture of complex traits by providing dynamic heritability estimates.

Methodology:

Bayesian framework using Gaussian processes to model time-dependent variance components and heritability, with modern MCMC for full uncertainty quantification and borrowing strength across adjacent time points.