End-to-end pipeline, from raw SRA reads to the differential-interactome / rewired-protein results and every manuscript figure and supplementary table. Scripts run in numeric order. The backbone (01–07) takes SRA → rewired proteins; the downstream scripts (08–13) produce enrichment, modules, figures, and supplementary tables. Figure and supplementary generation are kept in their own scripts, separate from the analytical backbone.

The differential-interactome (Q-value) methodology is adapted from Gulfidan et al. (2020), "Pan-cancer mapping of differential protein–protein interactions," Sci Rep 10:3272, doi:10.1038/s41598-020-60127-x.

Layout

codes/
  config.R                      shared configuration (paths, thresholds, study metadata, seed=42)
  environment.yml               conda environment (CLI tools + R/Bioconductor packages)
  inputs/                       bundled small inputs (sample sheet, metadata, KEGG KO, UniProt GO)
  .gitignore                    excludes the large STRING files from commits

  # ---- backbone: SRA -> rewired proteins ----
  01_download_sra.sh            prefetch + fasterq-dump        -> data/raw/<study>/*.fastq.gz
  02_preprocess.sh              FastQC, fastp, SortMeRNA, BBMap, MultiQC -> data/clean/<study>/clean/
  03_align_quantify.sh          STAR + featureCounts           -> data/counts/<study>.txt
  04_differential_expression.R  DESeq2 (apeglm shrinkage)       -> data/degs/, outputs/differential_expression/
  05_orthology_mapping.R        KO-based FOXG/FPSE -> FGSG       -> resources/
  06_differential_interactome.R Q-value differential interactome-> outputs/differential_interactome/
  07_rewired_analysis.R         rewired-protein classification  -> outputs/rewired_analysis/

  # ---- downstream: enrichment, modules, figures, supplementary ----
  08_enrichment.R               KEGG/GO hypergeometric enrichment (DEGs + rewired)
  09_network_modules.R          aggregated networks + Louvain modules
  10_figures_volcano.R          volcano panel with rewired overlay (Fig 4)
  11_figures_overlaps.R         DEG/rewired Venn + UpSet + classification (Figs 2, 3, 5, 6)
  12_figures_marker.R           marker-gene heatmap (Fig 8)
  13_supplementary_tables.R     Supplementary Tables S1-S9

Inputs

The small inputs are bundled in codes/inputs/; the large STRING network and the genome references are not committed (too large for git) and are downloaded once. config.R reads the small inputs from codes/inputs/ and the large STRING files from ../string_data/.

Bundled in codes/inputs/ (provided):

• study_table.csv, study_metadata.csv — sample sheet / study metadata; the sample sheet drives the download loop and the control/treatment assignment (one row per run).
• fgr_ko.txt, fox_ko.txt, fpu_ko.txt — KEGG Orthology assignments. Re-downloadable from the KEGG REST API, e.g. https://rest.kegg.jp/link/ko/fgr (and /fox, /fpu).
• fgr_uniprot_go.tsv — F. graminearum GO annotations (UniProt, https://www.uniprot.org).

Download once (not in git):

• STRING v12.0 → place in string_data/ at the project root. From https://string-db.org (version 12.0): 229533.protein.links.v12.0.txt + 229533.protein.aliases.v12.0.txt (F. graminearum, taxid 229533) and 426428.protein.aliases.v12.0.txt (F. oxysporum, taxid 426428). The .protein.links file is ~100 MB, which is why it is kept out of the repository.
• Genome references → data/ref/<asm>/ (FASTA *_genomic.fna + GTF *_genomic.gtf). NCBI RefSeq: fgr = GCF_000240135.3, fox = GCF_000149955.1, fpu = GCF_000303195.2 (https://www.ncbi.nlm.nih.gov/datasets/).
• SortMeRNA default DB smr_v4.3_default_db.fasta — from the SortMeRNA 4.3.4 release (https://github.com/sortmerna/sortmerna); export $SORTMERNA_DB.

data/degs/ (the DESeq2 output of step 04) is the authoritative DEG base for all downstream steps.

Run order

# backbone (run the heavy CLI steps on Linux/HPC)
bash codes/01_download_sra.sh
THREADS=8  SORTMERNA_DB=~/sortmerna/database/smr_v4.3_default_db.fasta  bash codes/02_preprocess.sh
THREADS=12 bash codes/03_align_quantify.sh
Rscript codes/04_differential_expression.R
Rscript codes/05_orthology_mapping.R
Rscript codes/06_differential_interactome.R     # heavy: STRING co-expression over 8 studies
Rscript codes/07_rewired_analysis.R
# downstream
Rscript codes/08_enrichment.R
Rscript codes/09_network_modules.R
Rscript codes/10_figures_volcano.R
Rscript codes/11_figures_overlaps.R
Rscript codes/12_figures_marker.R
Rscript codes/13_supplementary_tables.R

Key parameters to check (all centralised in config.R)

Stage	Parameter	Value	Where
featureCounts	mode	-p -C -t exon -g gene_id	03_align_quantify.sh
DESeq2	low-count filter	total counts ≥ 10	04
DESeq2	reference level	Control (vs Sample)	04
DESeq2	LFC shrinkage	apeglm	04
DEG call	significance	padj < 0.05 and abs(log2FC) > 1	config.R (PADJ_THRESHOLD, LOG2FC_THRESHOLD)
Orthology	one-to-many tie-break	lowest FGSG numeric ID	05
Interactome	STRING confidence	combined score ≥ 400	06
Interactome	binarisation	+1/0/−1 at per-sample mean ± 1 SD	06
Interactome	Q cut-offs	activated Q ≥ 0.90, repressed Q ≤ 0.10	config.R (QCRIT1, QCRIT2)
Interactome	min frequency	max(N_ctrl, N_trt) ≥ 0.20	config.R (DELTA)
Rewired	differential-degree cut	top 25th percentile	config.R (DEGREE_PERCENTILE)
Rewired	exclusion	not a DEG	07
Enrichment	test / FDR	hypergeometric (phyper), Benjamini–Hochberg < 0.05	08
Modules	Louvain	seed = 42, default resolution; keep size ≥ 3 + sig. term	config.R, 09

Other things to watch: STAR needs ≈ 32 GB RAM for index/alignment; step 06 is the slowest (STRING co-expression across all studies); set THREADS per machine. Study 3 is dual RNA-seq — only F. graminearum-mapped reads are kept, which the F. graminearum-only STAR index enforces. The global seed (42) in config.R fixes Louvain clustering and layout. To re-run only the downstream analysis without touching the (authoritative) DEGs, start from step 05.

Methods summary (matches the manuscript)

• Trimming fastp (default). rRNA SortMeRNA default DB. Alignment STAR to the per-host RefSeq genome. Counts featureCounts (paired, exon, gene_id).
• DEGs DESeq2 with apeglm LFC shrinkage; ref = Control; genes with < 10 total counts removed; padj < 0.05 and |log2FC| > 1.
• Orthology FOXG / FPSE mapped to FGSG by shared KEGG Orthology; lowest FGSG ID on ties.
• Differential interactome STRING v12.0 (≥ 400). Per sample, expression discretised to +1/0/−1 at mean ± 1 SD; co-expression counted over the nine pairwise states; Q = N_trt / (N_ctrl + N_trt); activated Q ≥ 0.90, repressed Q ≤ 0.10, max(N_ctrl, N_trt) ≥ 0.20.
• Rewired proteins top 25th percentile of differential degree and not a DEG.
• Enrichment hypergeometric (phyper), BH FDR < 0.05; KEGG names via the KEGG REST API; GO from UniProt. Modules Louvain (igraph, seed = 42); modules with ≥ 3 members and a significant KEGG/GO term retained.

Tool / package versions

Validated environment (the versions this pipeline was run and checked with):

• CLI: sra-tools (bioconda current), FastQC 0.12.1, MultiQC 1.30, fastp 1.0.1, SortMeRNA 4.3.4, BBMap 38.96, STAR 2.7.11b, samtools 1.22.1, Subread/featureCounts 2.1.1.
• R 4.5.3: DESeq2 1.50.2, apeglm 1.32.0, dplyr 1.2.1, readr 2.2.0, readxl 1.4.5, openxlsx 4.2.8.1, tidyr 1.3.2, tibble 3.3.1, ggplot2 4.0.2, ggrepel 0.9.8, ggraph 2.2.2, igraph 2.2.3, pheatmap 1.0.13, VennDiagram 1.8.2, UpSetR 1.4.0, RColorBrewer 1.1.3, svglite 2.2.2, httr 1.4.8, tictoc 1.2.1, stringr 1.6.0, gridExtra 2.3.

The manuscript Methods cite the original analysis versions (e.g. R 4.5.2, DESeq2 1.44.0, igraph 2.0.3, httr 1.4.7); the results reproduce across these minor version differences.

Figure / table provenance

• Fig 4 (volcano) — 10_figures_volcano.R
• Fig 2 (DEG Venn), Fig 3 (DEG UpSet), Fig 5 (rewired UpSet), Fig 6 (rewired classification) — 11_figures_overlaps.R
• Fig 8 (marker heatmap) — 12_figures_marker.R
• Fig 1 (DEG KEGG) — 08_enrichment.R; Fig 7 (rewired PPI networks) — 09_network_modules.R
• Tables 2–10 — 05–09; Supplementary S1–S9 — 13_supplementary_tables.R