Meso‑structural domains, not aggregated networks, reveal the action of structural selection

This repository contains all data, code, and figure outputs required to fully reproduce the analyses presented in the manuscript "Meso‑structural domains, not aggregated networks, reveal the action of structural selection". The dataset includes the raw trophic network, cleaned edge lists, meso‑structural metrics, structural‑selection outputs, figure‑level CSVs, summary tables, and the complete analysis pipeline used to generate all results.

The project is organized into the following components:

Manuscript files

• Miranda-Pérez and Mendoza-Reyes Manuscript.pdf (main text).
• Miranda-Pérez and Mendoza-Reyes Supplementary Material.pdf (figure and tables).

/data/

Raw and processed datasets used throughout the pipeline:

• FW_008.csv — raw adjacency matrix of the food web.
• fw_cleaned_edges.csv — cleaned directed edge list (from, to, weight).
• metrics_global_local.csv — global and local meso‑structural metrics per species.
• metrics_with_domains.csv — same as above, plus k‑means meso‑structural domains.
• df_full.csv — full meso‑structural dataset including domains and (when available) structural‑selection outputs.
• df_slopes.csv — per‑egonet structural‑selection slopes and p‑values.

FW_008.csv — Raw adjacency matrix of the Caribbean marine food web

Source: Bascompte, J., Melián, C. J., & Sala, E. (2005). Interaction strength combinations and the overfishing of a marine food web. Proceedings of the National Academy of Sciences USA, 102: 5443–5447. Original URL: http://www.pnas.org/content/102/15/5443

Description:

• 249 species/groups (rows = prey, columns = predators)
• 3313 trophic interactions
• Connectance = 0.107
• Single connected component
• Cell values represent the relative frequency of prey items in predator diets
• Cannibalism documented in 11 species/groups

Redistribution: This dataset is publicly available from the original source and is included here solely for reproducibility, following standard scientific practice.

/data_figures/

One CSV per figure panel, containing the exact data used to generate each plot. These files ensure full reproducibility of Figures 1–4 and Supplementary Figure S1.

/tables/

Summary tables corresponding to Tables 1–4 in the manuscript and supplement:

• table1_global_metrics.csv
• table2_mesostructural_correlations.csv
• table3_mesostructural_domains.csv
• table4_structural_selection.csv

/code/

All scripts required to reproduce the full analysis pipeline:

• run_all.R — master pipeline.
• 00_clean_data.R — cleaning of the raw adjacency matrix.
• 01_fig1.R — code for Figure 1.
• 02_mesostructural_mosaic.R — code for Figure 2.
• 03_fig3.R — code for Figure 3.
• 04_fig4.R — code for Figure 4.
• 05_supp_figure.R — code for Supplementary Figure S1.
• 06_tables.R — code for summary tables.
• R/ — utility functions used by the pipeline.

/figures_main/

PDF versions of Figures 1–4.

/figures_supp/

PDF version of Supplementary Figure S1.

All analyses were performed in R using a fully scripted and reproducible workflow. This Zenodo deposit provides complete transparency and ensures long‑term accessibility of all materials associated with the study.