An R package for the spatial analysis of population genetic marker data: a
locus data type, genetic diversity, structure, distance and relatedness,
Population Graphs, directional gene flow (genetic gravity), and simulation.
Full documentation: https://dyerlab.github.io/gstudio/
remotes::install_github("dyerlab/gstudio")Four gateway functions cover the analyses, each choosing an estimator with
mode =, so there is one name to remember per question. Every function takes
its data first, so analyses read as pipelines:
library(gstudio)
library(dplyr)
data(arapat)
arapat |>
genetic_diversity(mode = "He") # A, Ae, He, Ho, Hi, Fis, ...
arapat |>
genetic_structure(stratum = "Species", mode = "Gst") # Gst, Gst_prime, Dest, Fst
arapat |>
genetic_distance(stratum = "Species", mode = "cGD") # AMOVA, Nei, cGD, pGD, ...
arapat |>
slice(1:10) |>
genetic_relatedness(mode = "Nason") # Nason, LynchRitland, ...The same calls work without pipes, with every option passed as an argument,
e.g. genetic_distance(arapat, stratum = "Species", mode = "cGD").
Tidy results. Analyses return standard data.frames (or matrices) that you
plot yourself with ggplot2:
library(ggplot2)
arapat |>
frequencies(loci = "LTRS", stratum = "Species") |>
ggplot(aes(x = Allele, y = Frequency, fill = Stratum)) +
geom_col(position = "dodge") +
theme_minimal()The exception is graphs: plot() on a Population Graph or a gravity field
returns a ggplot from one shared backend, which you can extend with +. For
fully custom graph figures, use ggraph.
data(lopho)
lopho |>
plot() # nodes sized by 'size', filled by 'Region'
arapat |>
population_graph() |>
plot(layout = strata_coordinates(arapat))population_graph()/popgraph(): build a conditional-independence graph from multilocus data.decorate_graph(): attach data columns (e.g. coordinates, region) as vertex attributes.plot(): optionslayout,node_size,node_labels,node_fill,edgesandedge_width. Directed graphs (asymmetric_popgraph()) are drawn with arrows.animate_popgraphs(): animate a sequence of graphs on fixed node positions.to_sf(),to_json(),write_popgraph(), ...: export.
Each Population Graph edge is symmetric, but its two populations are compared against different neighbourhoods. Genetic gravity reads the direction of gene flow from that difference.
data(gravity) # simulated genotypes, flow toward Pop25
graph <- gravity |>
population_graph()
deme <- as.integer(sub("Pop", "", igraph::V(graph)$name))
graph |>
source_sink_scores(x = deme) # gravity and source-sink score S
graph |>
source_sink_test(x = deme) # gradient in S along an axis
graph |>
ibgd(x = deme) # isolation by graph distance (Mantel test)
graph |>
ibgd(x = deme, mode = "pgd") # does a directional model fit better?
gravity |>
genetic_distance(mode = "pgd") # partitioned (directed) cGD
graph |>
gravity_field() |>
plot() # map of sources and sinksSee vignette("genetic_gravity") for the model and a full walkthrough.
strata_coordinates() gives population centroids, and addAlleleFrequencies()
adds allele-frequency pie charts to a leaflet map:
library(leaflet)
library(leaflet.minicharts)
freqs <- arapat |>
frequencies(loci = "LTRS", stratum = "Species")
arapat |>
strata_coordinates(stratum = "Species") |>
leaflet() |>
addTiles() |>
addAlleleFrequencies(freqs)make_populations(),migration_matrix(),simulate_pop(): forward-time, individual-based simulation, including directional stepping stones.read_population()/write_population(): read and write genotype files.to_genepop(),to_structure(), ...: export to other programs.
Version 1.15 consolidates the interface, and some changes break code written
for 1.14: the standalone estimators (He(), Fis(), Fst(), Gst(),
dist_nei(), rel_nason(), ...) are reached through the gateway functions, sp-based
exports are replaced by to_sf(), tests return htest objects, and table
columns follow one naming convention (Stratum, from/to, statistic,
p.value). See NEWS for the full list, with old and new names.
Dyer RJ (2009) GStudio: a suite of tools for the spatial analysis of genetic marker data. Molecular Ecology Resources 9: 110--113.
Questions, bug reports and contributions are welcome: open an issue at https://github.com/dyerlab/gstudio/issues, or contact Rodney J. Dyer.