Estimation of dynamic SNP-heritability with Bayesian Gaussian process models

Abstract

Motivation Improved DNA technology has made it practical to estimate single-nucleotide polymorphism (SNP)-heritability among distantly related individuals with unknown relationships. For growth- and development-related traits, it is meaningful to base SNP-heritability estimation on longitudinal data due to the time-dependency of the process. However, only few statistical methods have been developed so far for estimating dynamic SNP-heritability and quantifying its full uncertainty. Results We introduce a completely tuning-free Bayesian Gaussian process (GP)-based approach for estimating dynamic variance components and heritability as their function. For parameter estimation, we use a modern Markov Chain Monte Carlo method which allows full uncertainty quantification. Several datasets are analysed and our results clearly illustrate that the 95% credible intervals of the proposed joint estimation method (which ‘borrows strength’ from adjacent time points) are significantly narrower than of a two-stage baseline method that first estimates the variance components at each time point independently and then performs smoothing. We compare the method with a random regression model using MTG2 and BLUPF90 software and quantitative measures indicate superior performance of our method. Results are presented for simulated and real data with up to 1000 time points. Finally, we demonstrate scalability of the proposed method for simulated data with tens of thousands of individuals. Availability and implementation The C++ implementation dynBGP and simulated data are available in GitHub: https://github.com/aarjas/dynBGP. The programmes can be run in R. Real datasets are available in QTL archive: https://phenome.jax.org/centers/QTLA. Supplementary information Supplementary data are available at Bioinformatics online.

Age-dependent genetic architecture across ontogeny of body size in sticklebacks

“…As such, the method allows to model dynamic variance components and h 2 for longitudinal data. Similarly, to the animal models described above, the dynGP analysis uses the GRM constructed from SNP data to account for relatedness among individuals and estimate V A and h 2 under a linear mixed model framework [62]. We followed the methodology of Arjas et al [62] and ran separate models for each cross for 1 000 000 MCMC-iterations (see electronic supplementary material, methods for details).…”

“…Similarly, to the animal models described above, the dynGP analysis uses the GRM constructed from SNP data to account for relatedness among individuals and estimate V A and h 2 under a linear mixed model framework [62]. We followed the methodology of Arjas et al [62] and ran separate models for each cross for 1 000 000 MCMC-iterations (see electronic supplementary material, methods for details). As the model formulation in dynGP does not allow for fixed effects we used vectors of residuals from linear regressions of age-specific body size on sex as response variables in order to have equivalent structures to the animal models described above.…”

“…We then re-estimated V A and h 2 following the 'functionvalued trait' approach by using the method described in [62] implemented in the dynGP package (https://github.com/ aarjas/dynBGP).…”

Age-dependent genetic architecture underlines similar heritability of body size in sticklebacks

“…We then re-estimated VA and h² following the 'function-valued trait' approach by using the method described in [55] implemented in the dynGP package (https://github.com/aarjas/dynBGP). Contrary to the univariate animal model approach, this method combines all measurements of body size over time and models the dependency among those measurements using a Bayesian Gaussian process.…”

“…As such, the method allows to model dynamic variance components and heritability for longitudinal data using SNP data, and to provide reliable uncertainty estimates around the estimated quantities. In these analyses, we followed the methodology of Arjas et al [55] and estimated the GRMs using the rrBLUP R package [56] and ran the models for 1 000 000 MCMC-iterations. To account for the effect of sex, we used vectors of residuals from linear regressions of age-specific body size on sex as response variables.…”