dynamics_of_size_and_shape_variation

This repository includes R scripts and data for geometric morphometric analyses of morphological variation during embryonic development in Triturus newts.
We explored two phases of embryonic development separately:

• Tailbud phase: Quantified the shape of the whole embryo
• Prehatching phase: Quantified head shape

---

Input Files

• FactorsGroup1.xlsx – Classifiers for tailbud embryos
• FactorsGroup2.xlsx – Classifiers for prehatching embryos
• g1a.tps – The first replicate of landmark coordinates for tailbud embryos
• g1b.tps – The second replicate of landmark coordinates for tailbud embryos
• g2a.tps – The first replicate of landmark coordinates for prehatching embryos
• g2b.tps – The second replicate of landmark coordinates for prehatching embryos
• slajding2.xlsx – Matrix defining semilandmarks in tailbud embryos
• slajding.xlsx – Matrix defining semilandmarks in prehatching embryos

---

Dependencies

Install the following R libraries before running analyses:

library(RRPP)
library(geomorph)
library(xlsx)
library(ggplot2)
library(abind)
library(car)

---

Workflow Overview

Data Import & Preprocessing

• Read TPS coordinates
• Link with metadata
• Check outliers
• Define stage factors

---

Procrustes Superimposition & Extracting Symmetric Component

• For tailbud embryos:

  procGPA <- gpagen(coords, curves = sliders)

• For prehatching embryos:

  gpag2 <- bilat.symmetry(dfg2$shape, ind = dfg2$ind, object.sym = TRUE, land.pairs = parovi, curves = slide2)

These steps:

• Correct for size, orientation, and translation
• Slide semilandmarks to minimize bending energy

Output:

• Shape (Procrustes coordinates for tailbud stages and symmetric component for prehatching stages)
• Size (centroid size) coordinates

---

Clustering

Gaussian Mixture Model clustering using Expectation-Maximisation:

cluster <- hddc(two.d.array(gpa$coords), model = "ALL", criterion = "bic", itermax = 10000, mc.cores = 30, scaling = TRUE)
predict(cluster) # Calculates posterior probabilities for each embryo

---

Principal Component Analysis (PCA)

• Visualizes group separation

---

Shape Differences

• Test mean shape differences between stages with permutations:

  pairwise.procD.lm(...)
  # 999 permutations for p-values

---

Size Differences

• Test differences in size between stages:

  summary(aov(Csize ~ as.factor(stage)))
  TukeyHSD(aov(Csize ~ as.factor(stage))) # Multiple comparisons of means

---

Allometry

• Test allometric effect:

  procD.lm(shape ~ log(Csize))
  # Produces allometric regression scores for visualization

---

Morphological Disparity

• Compare within-group variance across development:

  morphol.disparity(procD.lm object, ...)
  # Plotted with confidence trends through stages

---

Citation

If using this workflow in your publication, please cite:

Bugarcic, M., Ivanovic, A., Cvijanovic, M., Ajdukovic, M., Drobnjakovic, M., Wielstra, B., Vucic, T. (2025). Dynamics of size and shape variation during embryonic development in the Balkan crested newt (2025), Under Review

---

Contact

For issues within script:
Marko Bugarcic, MSc
PhD Candidate
University of Belgrade – Faculty of Biology
Email: marko.bugarcic@bio.bg.ac.rs

For data requests:
Tijana Vucic
Leiden University – Institute of Biology Leiden, University of Belgrade – Faculty of Biology
Email: tijana.vucic@bio.bg.ac.rs