Paper: Rumen microbiome composition of beef cows across Córdoba, Colombia characterized by indicator analysis, microbial networks, and predictive modelling
Authors: Vincent P. Reyes et al.
Sequencing: 16S rRNA amplicon sequencing (V3–V4 region)
Samples: 142 rumen samples from 6 regions of Córdoba, Colombia
Pipeline: QIIME2 → R (phyloseq, vegan, mixOmics, randomForest, igraph)
.
├── 01_Preprocessing/
│ ├── Step1_ASV_Filtering.R # Singleton + prevalence/abundance filtering
│ ├── Step2_Sequencing_Depth.R # Depth summary, rarefaction curves
│ ├── Step3_Alpha_Diversity_and_Rarefy.R # Rarefy to min depth → ps_rarefied.rds; alpha diversity
│ ├── Analyze_Sequencing_Depth.R # Extended depth diagnostics
│ └── Plot_Sequencing_Depth.R # Sequencing depth figures
│
├── 02_Beta_Diversity_dbRDA/
│ ├── Beta_Diversity_PCoA_PERMANOVA.R # PCoA (Bray-Curtis + UniFrac), PERMANOVA, betadispersion
│ ├── dbRDA_Analysis.R # Overall and partial dbRDA (rarefied Bray-Curtis)
│ ├── Variance_Partitioning_Rarefied.R # varpart(): pure Geography vs. Environment
│ └── Plot_dbRDA_partial_region.R # Publication figure: partial dbRDA
│
├── 03_Taxonomic_Abundance/
│ ├── Combine_Taxonomy_Plots.R # Phylum and family relative abundance bar plots
│ └── Plot_ASV_Intersections.R # UpSet plot of ASV intersections across regions
│
├── 04_Machine_Learning/
│ └── Machine_Learning_sPLSDA_RF.R # sPLS-DA (tuned) + Random Forest classification by region
│
├── 05_Network_Analysis/
│ ├── Network_Analysis_All_Regions.R # Co-occurrence networks per region (SparCC/Spearman)
│ ├── Network_MultiPanel_Generator.R # Multi-panel network figure generator
│ └── Network_Metrics_Comparison.R # Hub detection and network metric comparison
│
└── 06_Supplementary/
├── Supplementary_Sensitivity_Analysis.R # Pre- vs post-filtering sensitivity PCoA + PERMANOVA
├── Analyze_PERMANOVA_Farm_Level.R # Farm-level PERMANOVA (pseudoreplication check)
└── Analyze_Sample_Loss.R # Sample loss tracking across filtering steps
01_Preprocessing/Step1 → Step2 → Step3 (generates ps_filtered.rds, ps_rarefied.rds)
↓
02_Beta_Diversity_dbRDA/ (all scripts read ps_rarefied.rds)
03_Taxonomic_Abundance/ (reads ps_filtered.rds → relative abundance)
04_Machine_Learning/ (reads ps_filtered.rds → relative abundance)
05_Network_Analysis/ (reads ps_filtered.rds)
06_Supplementary/ (self-contained)
Working directory: Set to the project root containing
ps_filtered.rds,ps_rarefied.rds, andmetadata.csvbefore running any script.
| Analysis | Data used | Rationale |
|---|---|---|
| Alpha diversity | Rarefied (ps_rarefied.rds) |
Observed/Shannon are sensitive to library size |
| Beta diversity, PERMANOVA, dbRDA | Rarefied Bray-Curtis | Bray-Curtis is sensitive to absolute abundance; rarefaction equalises depth |
| Variance partitioning | Rarefied Bray-Curtis | Consistent with beta diversity pipeline |
| Taxonomic abundance | Relative abundance | Compositional; library-size independent |
| sPLS-DA / Random Forest | Family-level relative abundance | Compositional; Hellinger for PCA |
# CRAN
install.packages(c("vegan", "ggplot2", "tidyverse", "randomForest",
"igraph", "pairwiseAdonis", "ggrepel", "patchwork",
"multcompView", "rstatix", "ggpubr"))
# Bioconductor
BiocManager::install(c("phyloseq", "ANCOMBC", "mixOmics"))Raw FASTQ files and the QIIME2-derived ASV table (asv-table-filtered2.tsv), taxonomy (taxonomy.tsv), phylogenetic tree (tree.nwk), and sample metadata (metadata.csv) are available upon reasonable request to the corresponding author.
Reyes VP et al. (2025). Rumen microbiome composition of beef cows across Córdoba, Colombia characterized by indicator analysis, microbial networks, and predictive modelling. [Journal].
MIT