Global gut content data synthesis and phylogeny delineate reef fish trophic guilds

This repository contains code and data needed to reproduce the figures and results of the manuscript: “Global gut content data synthesis and phylogeny delineate reef fish trophic guilds” (ADD REF)

Instructions

All analyses were done in R. First you need to open an R session with working directory set to the root of the project.

We use a number of packages as listed in the DESCRIPTION, missing packages can be easily installed by devtools:
devtools::install_deps()
(run install.packages("devtools") to install devtools if needed.)

Then the entire project can be reproduced by running the file make.R or scripts can be run seperately.
❗ 💥 CAUTION: The script depends on parallel computation and uses up to 50 cores and 250G of memory, and thus should be run on a supercomputer. It takes about one week to reproduce the entire project.

Content

The directory contains:
- 📄 DESCRIPTION: Contains all packages needed to reproduce the analyses.
- 📄 make.R: Script to reproduce all parts of the project.
- 📁 data: Folder containing all data.
- 📁 scripts: Folder containing scripts to perform multiple parts of the analysis to produce results and figures.
- 📁 R: Folder containing custom functions used to perform phylogenetic regression.
- 📁 output: Folder containing all results and figures.

data

original_expert_classification.csv:
Trophic classifications, originating from 33 publications.
converted_expert_classification.csv:
Modified trophic classifications to compare trophic guilds across publications. We standardized by converting the original trophic categories into five broad trophic guilds: herbivores and detritivores, invertivores, omnivores, planktivores, and piscivores.
data_guts.csv:
Dietary data from five published works: Hiatt & Strasburg (1960) for the Marshall Islands, Randall (1967) for Puerto Rico and the Virgin Islands, Hobson (1974) for Hawaii, Harmelin-Vivien (1979) for Madagascar, and Sano et al. (1984) for Okinawa. In addition, we provide hitherto unpublished data on the gut contents of 3,015 individuals of 111 species collected in New Caledonia from 1984 to 2000. Prey items were grouped into 38 ecologically informative prey groups.
all_reef_fish.csv
Global list of global reef fish species, used for extrapolation.

scripts

01_guilds_compare_experts.R:
We conducted a systematic review in the literature to collect information about how different authors have classified reef fish species into different trophic guilds. We could obtain 30 distinct classifications and this information is contained in the file “data/original_experts_classification.csv”. These data were then made comparable (see “data/converted_experts_classification.csv”. The script compares pairs of experts and quantify the agreement/disagreement among them.
02_global_network_guilds_site_specific.R:
This script defines the diet for each combination of reef fish species and site. Then we used the method introduced by Becket (2016) to identify trophic guilds according to a maximization of the modularity of the weighted bipartite network. Once we identified the trophic guilds for each species x site combination we realized that in the vast majority of cases the same species was classified in the same way regardless the location of sampling.

Beckett, S.J. 2016 Improved community detection in weighted bipartite networks. Royal Society open science 3, 140536.

03_global_network_guilds.R:
Once we realized that, at our resolution, the trophic guilds for the same species in different locations were exactly the same, we decided to perform the same analysis at global level (i.e. combining ditar data for the same species coming from different sites) to define the final trophic guilds. This is what this script does.
04_trophic_interactions_calibration_dataset.R
Once the guilds were defined we tested whether it was possible to directly predict the probability of interaction between reef fishes and the prey categories used. We did that using machine learning to predict trophic interactions according to the phylogenetic position of reef fishes and their body size. This script perform a series of preliminary analyses (i.e. definition of the most likely phylogenetic tree for reef fishes, extraction of phylogenetic eigenvector maps for reef fishes) that are needed to obtain the calibration dataset.
05_h20_ensemble_model_calibration.R
This script sets up a h2o cluster (https://www.h2o.ai) to predict the probability of trophic interaction between reef fish species and our prey categories. Then it uses the calibration dataset to perform an ensemble model composed of: boosted regression trees, extreme gradient boosting, random forest and a glm used as a super-learner.
06_h20_model_check_and_extrapolation.R This script contains the functions needed to check the model performance (e.g. AUC, TSS) and to extrapolate the likely trophic interaction for all reef fish species.
07_phylogenetic_multinomial_model_and_extrapolation.R This script performs a phylogenetic regression to predict the probabilities of trophic guilds for each species, as well as the extrapolation to global reef fishes.
08_network_plot_diets.R
This script contains code to reproduce all elements of figure 3.
09_plot_phylotree.R
This script contains code to reproduce all elements of figure 4.

How to download this project for people not familiar with GitHub:

On the project main page on GitHub, click on the green button clone or download and then click on Download ZIP.

Working environment

sessionInfo()
#> R version 3.6.3 (2020-02-29)
#> Platform: x86_64-pc-linux-gnu (64-bit)
#> Running under: Ubuntu 18.04.5 LTS
#> 
#> Matrix products: default
#> BLAS:   /usr/lib/x86_64-linux-gnu/blas/libblas.so.3.7.1
#> LAPACK: /usr/lib/x86_64-linux-gnu/lapack/liblapack.so.3.7.1
#> 
#> locale:
#>  [1] LC_CTYPE=fr_FR.UTF-8       LC_NUMERIC=C              
#>  [3] LC_TIME=fr_FR.UTF-8        LC_COLLATE=fr_FR.UTF-8    
#>  [5] LC_MONETARY=fr_FR.UTF-8    LC_MESSAGES=fr_FR.UTF-8   
#>  [7] LC_PAPER=fr_FR.UTF-8       LC_NAME=C                 
#>  [9] LC_ADDRESS=C               LC_TELEPHONE=C            
#> [11] LC_MEASUREMENT=fr_FR.UTF-8 LC_IDENTIFICATION=C       
#> 
#> attached base packages:
#> [1] stats     graphics  grDevices utils     datasets  methods   base     
#> 
#> loaded via a namespace (and not attached):
#>  [1] compiler_3.6.3  magrittr_1.5    tools_3.6.3     htmltools_0.4.0
#>  [5] yaml_2.2.0      Rcpp_1.0.5      stringi_1.4.3   rmarkdown_2.1  
#>  [9] knitr_1.25      stringr_1.4.0   xfun_0.10       digest_0.6.22  
#> [13] rlang_0.4.7     evaluate_0.14