morphospace

The goal of morphospace is to enhance representation and heuristic exploration of multivariate ordinations of shape data. This package can handle the most common types of shape data working in integration with other widely used R packages such as Morpho (Schlager 2017), geomorph (Adams et al. 2021), shapes (Dryden 2019), and Momocs (Bonhome et al. 2014), which cover other more essential steps in the geometric morphometrics pipeline (e.g. importation, normalization, statistical analysis).

Installation

You can install the development version of morphospace from GitHub with:

# install.packages("devtools")
devtools::install_github("millacarmona/morphospace")

Concept

The basic idea behind morphospace is to build empirical morphospaces using multivariate ordination methods, then use the resulting ordination as a reference frame in which elements representing different aspects of morphometric variation are projected. These elements are added to both graphic representations and objects as consecutive ‘layers’ and list slots, respectively, using the %>% pipe operator from magrittr (Bache & Wickham 2022).

The starting point of the morphospace workflow is a set of shapes (i.e. morphometric data that is already free of variation due to differences in orientation, position and scale). These are fed to the mspace function, which generates a morphospace using a variety of multivariate methods related to Principal Component Analysis. This general workflow is broadly outlined below using the tails data set from Fasanelli et al. (2022), which contains tail shapes from 281 specimens belonging to 13 species of the genus Tyrannus.

# Load tail data
data("tails")

shapes <- tails$shapes
spp <- tails$data$species
wf <- tails$links
phy <- tails$tree

# Generate morphospace
mspace(shapes, links = wf, cex.ldm = 5)

The ordination produced by mspace is used as a reference frame in which scatter points, convex hulls / confidence ellipses, a phylogeny, a set of morphometric axes or a landscape surface can be projected using the proj_* functions:

# Get mean shapes of each species
spp_shapes <- expected_shapes(shapes = tails$shapes, x = tails$data$species)

# Generate morphospace and project:
msp <- mspace(shapes = shapes, links = wf, cex.ldm = 5) %>% 
  # scatter points
  proj_shapes(shapes = shapes, col = spp) %>% 
  # convex hulls enclosing groups
  proj_groups(shapes = shapes, groups = spp, alpha = 0.5) %>% 
  # phylogenetic relationships
  proj_phylogeny(shapes = spp_shapes, tree = phy, lwd = 1.5, 
                 col.tips = match(phy$tip.label, levels(spp)))

Once the "mspace" object has been created, the plot_mspace function can be used to either regenerate/modify the plot, add a legend, or to combine morphometric axes with other non-shape variables to produce ‘hybrid’ morphospaces. For example, PC1 can be plotted against size to explore allometric patterns.

# Plot PC1 against log-size, add legend
plot_mspace(msp, x = tails$sizes, axes = 1, nh = 6, nv = 6, cex.ldm = 4, 
            alpha.groups = 0.5, col.points = spp, col.groups = 1:nlevels(spp), 
            phylo = FALSE, xlab = "Log-size", legend = TRUE)

Or ordination axes could be combined with a phylogenetic tree to create a phenogram:

# Plot vertical phenogram using PC1, add a legend
plot_mspace(msp, y = phy, axes = 1, nh = 6, nv = 6, cex.ldm = 4, 
            col.groups = 1:nlevels(spp), ylab = "Time", legend = TRUE)

morphospace can also handle closed outlines (in the form of elliptic Fourier coefficients) and 3D landmark data, as shown below briefly using the shells and shells3D data sets:

# Load data
data("shells")

shapes <- shells$shapes
spp <- shells$data$species

# Generate morphospace
mspace(shapes, mag = 1, nh = 5, nv = 4, bg.model = "light gray") %>%
  proj_shapes(shapes = shapes, col = spp) %>%
  proj_groups(shapes = shapes, groups = spp, alpha = 0.5, ellipse = TRUE)

# Load data
data("shells3D")

shapes <- shells3D$shapes
spp <- shells3D$data$species
mesh_meanspec <- shells3D$mesh_meanspec

# Generate surface mesh template
meanspec_shape <- shapes[,,findMeanSpec(shapes)]
meanmesh <- tps3d(x = mesh_meanspec, 
                  refmat = meanspec_shape, 
                  tarmat = expected_shapes(shapes))

# Generate morphospace
mspace(shapes, mag = 1, bg.model = "gray", cex.ldm = 0, template = meanmesh, 
       adj_frame = c(0.9, 0.85)) %>%
  proj_shapes(shapes = shapes, col = spp, pch = 16) %>%
  proj_groups(shapes = shapes, groups = spp, alpha = 0.3)
#> Preparing for snapshot: rotate mean shape to the desired orientation
#>  (don't close or minimize the rgl device).Press <Enter> in the console to continue:
#> This can take a few seconds...
#> DONE.

Aside from working with these types of morphometric data, morphospace provides functions to perform some useful shape operations, use TPS interpolation of curves/meshes to improve visualizations, and supports a variety of multivariate methods (bgPCA, phylogenetic PCA, PLS, phylogenetic PLS) to produce ordinations. For these and other options and details, go to General usage and Worked examples.

Update 1 (August 2022)

  • Different behavior for proj_shapes (now replaces mspace$x with the actual scores being projected) and proj_axis (now adds one or more axes into an mspace$shape_axis).

  • New ellipses_by_groups_2D (uses car::ellipse) function as an option for proj_groups and plot_mspace.

  • Morphospaces without background shape models are now an option (for both mspace and plot_mspace).

  • plot_mspace now regenerates the original mspace plot by default (proj_* functions were modified such that all the relevant graphical parameters are inherited downstream to plot_mspace), has further flexibility regarding hybrid morphospaces (plot_phenogram has been updated) and allows adding a legend (and some various bugs were fixed as well).

  • Univariate morphospaces and associated density distributions are now an option (all the mspace workflow functions have been modified accordingly, especially proj_shapes and proj_groups).

  • consensus and expected_shapes have been merged in a single function (the name expected_shapes was retained as the former was clashing with ape::consensus), which can handle both factors and numerics.

  • Both detrend_shapes and expected_shapes can now calculate phylogenetically-corrected coefficients for interspecific data sets (Revell 2009).

Update 2 (August 2023)

  • The structure of "mspace" objects has been reorganized and now contain 3 main slots: $ordination (multivariate ordination details), $projected (elements added using proj_* functions) and $plotinfo (used for regeneration using plot_mspace). This has been complemented with a print method for the "mspace" class.

  • New proj_landscape function has been added to represent adaptive surfaces interpolated from functional or performance indices (although can be used for any numerical variable).

  • proj_consensus has been removed.

  • New extract_shapes function for extracting synthetic shapes from "mspace" objects (background shape models, shapes along ordination axes, or specific coordinates selected interactively).

  • New burnaby function, implementing Burnaby’s approach for standardization of morphometric data by computing a shape subspace orthogonal to an arbitrary vector or variable

  • New phyalign_comp function, implementing Phylogenetically aligned component analysis, which finds the linear combination of variables maximizing covariation between trait variation and phylogenetic structure (Collyer & Adams 2021). Still a work in progress.

  • Several internal adjustments have been introduced to the mspace, proj_* and plot_mspace functions in order to improve visualization and make the workflow more flexible.

  • Legends created using plot_mspace have been improved, and scale bars for interpreting landscapes have also been made available.

If you find any bugs please send me an email at pablomillac@gmail.com. Thanks!!

References

Adams D.C., Collyer M.L., Kaliontzopoulou A., & Baken E.K. (2021). geomorph: Software for geometric morphometric analyses. R package version 4.0.2. https://cran.r-project.org/package=geomorph.

Bache S.F., & Wickham H. (2022). magrittr: A Forward-Pipe Operator for R. R package version 2.0.3. https://CRAN.R-project.org/package=magrittr.

Bonhomme V., Picq S., Gaucherel C., & Claude J. (2014). Momocs: Outline Analysis Using R. Journal of Statistical Software, 56(13), 1-24. http://www.jstatsoft.org/v56/i13/.

Collyer, M. L., & Adams, D. (2021). Phylogenetically aligned component analysis. Methods in Ecology and Evolution, 12(2), 359-372. https://doi.org/10.1111/2041-210X.13515.

Dryden, I.L. (2019). shapes: statistical shape analysis. R package version 1.2.5. https://CRAN.R-project.org/package=shapes.

Fasanelli M.N., Milla Carmona P.S., Soto I.M., & Tuero, D.T. (2022). Allometry, sexual selection and evolutionary lines of least resistance shaped the evolution of exaggerated sexual traits within the genus Tyrannus. Journal of Evolutionary Biology, in press. https://doi.org/10.1111/jeb.14000.

Revell, L.J. (2009). Size-correction and principal components for interspecific comparative studies. Evolution, 63, 3258-3268 https://doi.org/10.1111/j.1558-5646.2009.00804.x.

Schlager S. (2017). Morpho and Rvcg - Shape Analysis in R. In Zheng G., Li S., Szekely G. (eds.), Statistical Shape and Deformation Analysis, 217-256. Academic Press. https://doi.org/10.1016/B978-0-12-810493-4.00011-0.