mobest - Spatiotemporal Ancestry Interpolation and Search

(latest version)

This R package provides types and functions for spatiotemporal interpolation of human genetic ancestry components, probabilistic similarity search and the calculation of a derived measure for mobility estimation. The workflow in mobest version 1.0.0 was specifically developed to support this research compendium: https://github.com/nevrome/mobest.analysis.2022, which in turn underpins [Schmid and Schiffels, 2023].

• mobest assumes you have a set of genetic samples with spatial (two coordinates in a projected reference system) and temporal positions (years BC/AD) for which you calculated a derived, numeric measure of genetic ancestry (e.g. coordinates in a PCA or MDS space).

• mobest provides a framework to perform spatiotemporal interpolation using Gaussian process regression (kriging) with the laGP package to reconstruct an ancestry field based on the ancestry measure you provided.

• mobest finally allows to derive a similarity probability for samples of interest within the interpolated field, which – under certain circumstances – can be interpreted as an origin probability.

Contents:

• 1. Install mobest
  • 1.1. Install mobest directly as an R package
  • 1.2. Create an apptainer image to run mobest
• 2. Basic similarity search
  • 2.1. Preparing the computational environment
  • 2.2. Preparing the input data
  • 2.3. Specifying the search sample
  • 2.4. Running mobest’s interpolation and search function
  • 2.5. Inspecting the computed results
• 3. Better map plots
• 4. Multiple samples and time slices
  • 4.1. Multiple search time slices
  • 4.2. Multiple search samples
  • 4.3. Summarizing multiple searches in one figure
• 5. Interpolation parameter estimation
  • 5.1. Preparing the computational environment
  • 5.2. Using a subset of the variogram to estimate the nugget parameter
  • 5.3. Finding optimal lengthscale parameters with crossvalidation
  • 5.4. An HPC crossvalidation setup for large lengthscale parameter spaces
• 6. Advanced mobest features
  • 6.1. Gaussian process regression on top of a linear model
  • 6.2. Spatiotemporal interpolation permutations in a model grid
  • 6.3. Similarity search with permutations
• 7. Input data types
  • 7.1. Basic data types
  • 7.2. Permutation data types
• 8. Citation
• 9. Bibliography