Data and Code for: Convergence or divergence? A three-layer quantitative synthesis tests a Scale–Spectrum–Stage framework of plant microbiome responses to stress

Abstract

This data and code package accompanies the manuscript “Convergence or divergence? A three-layer quantitative synthesis tests a Scale–Spectrum–Stage framework of plant microbiome responses to stress,” by Lele Liu, Wenyi Sheng, Yuhui Wang, Huijia Song, Yaolin Guo, and Weihua Guo. It is organized into data/, scripts/, d_tier/, and si/, and is intended to make the three-layer quantitative synthesis reproducible. The archive uses its own layout rather than the upstream repository structure: files that were under upstream config/, integration/, meta_analysis/, public_reanalysis/, or main_study/ are bundled here under data/ or scripts/. Version 1.0.0 is dated 2026-09-12 and matches the submission to Global Change Biology. The data/ directory contains the unified comparison tables and result files. unified_comparisons.csv holds 151 stress-vs-control comparison units, with key columns including layer, study_id, comparison_id, row_type, kingdom, organ, stage, stress_type, threeS_cell, yi, vi, and in_cell_model. extraction_data.csv is the 59-row three-tier literature extraction table, comprising 55 rows from 17 studies plus 4 focal-study rows. three_layer_table.csv gives M2 per-cell × per-layer estimates, and sensitivity_leave_layer.csv gives M5 leave-one-layer-out refits. three_layer_results.txt is the full M1–M6 console report and is the only copy for six of the nine rows of SI Table S4. The directory also includes predictions_3S.csv, the pre-registered 3S prediction matrix, figure_source/ with per-panel source tables for Figures 2–5, and si/ with regenerated Supporting Information figures. The scripts/ directory contains the full analysis pipeline. build_unified_table.R converts the three layers into the unified 151-unit table; three_layer_analysis.R runs the joint models M1–M6 using metafor and produces all per-cell estimates; 00_fetch_sra.py downloads SRA/ENA data; 02_dispersion_lnrr.R provides shared PERMDISP and lnRRβ bootstrap functions; 03_integrate.R integrates the public reanalysis layer into comparison units; and make_figures_gcb.R reproduces Figures 2–5 and Figures S1–S2 from data/. The d_tier/ directory contains D-tier recomputation inputs for three of five studies: Ning 2020, Wang 2024, and Halo 2026, each with a treatment coding table and VSEARCH script, or the script and its input table where processed data were provided. The remaining D-tier inputs are Bintarti 2022 and Edwards 2025, which must be obtained from their stated external sources. The reproduction chain is designed to run from the distributed package. Running Rscript scripts/three_layer_analysis.R --unified data/unified_comparisons.csv --outdir . reruns the joint models M1–M6 and rewrites three_layer_table.csv, sensitivity_leave_layer.csv, three_layer_results.txt, and the support figures forest_three_layer.png and heatmap_3S_layer.png. Running Rscript scripts/make_figures_gcb.R rebuilds Figures 2–5 and Figures S1–S2. build_unified_table.R rebuilds the 151-unit table from the three layer outputs. D-tier recomputations require downloading FASTQ files, running bash run_vsearch.sh in each study directory, and then running 02_dispersion_lnrr.R over the transposed ASV table; the WangY 2024 entry needs --spectrum F. Primary sequencing data for the common-garden experiment are available at GSA BioProject PRJCA050192. Software requirements are specified by step. Steps 1 and 3 need R with metafor; the models were run under R 4.5.0 with metafor 5.0.1. Step 2 additionally needs ggplot2 4.0.3, patchwork, R’s cairo device, the Liberation Sans font, and pdftoppm from poppler-utils for the 450 dpi JPG previews; the script stops with an explicit error if pdftoppm is missing. Step 4 needs vsearch 2.30.5, python3, and, for the reanalysis layer, vegan. The package is released under CC-BY-4.0; the full text is in LICENSE, the machine-readable citation is in CITATION.cff, and copyright belongs to Lele Liu, Wenyi Sheng, Yuhui Wang, Huijia Song, Yaolin Guo, and Weihua Guo.