Cross-Disorder GRN Thesis Pipeline

This repository analyzes cross-disorder differential expression in the context of brain gene regulatory networks (GRNs).
It combines:

1. Tissue-specific GRN matrices (PANDA/GRAND-derived; bipartite TF -> target edges)
2. Cross-study table showing psychiatric disorders
3. Python graph filtering/scoring/visualization
4. R-based enrichment, overlap, and meta-disorder analyses

Repository Layout

• src/run.py: main Python pipeline for GRN + DEG integration.
• src/graph_utils.py: data loading, Ensembl mapping, GRN filtering/merging helpers.
• src/graph_algos.py: regulatory scoring, summary stats, overlap metrics, CSV writers.
• src/graph_viz.py: PyVis/Matplotlib graph visualizations.
• src/degensembl.py: helper script to fill missing Ensembl IDs in DEG tables.
• src/thesisp1.R: primary R analysis pipeline (enrichment, GSEA, overlap tests, networks).
• src/p1thesis.R: older/alternate thesis pipeline version.
• src/p1permutationtest.R: standalone permutation overlap test script.
• src/p1deganalysis.R, src/p1geneontology.R, src/p1venndiagrams.R, src/wingodatasetanalysis.R: past exploratory/auxiliary analysis scripts.
• src/data/: input data files.
• src/results/: generated figures, tables, and HTML network visualizations.
• src/lib/: vendored JS/CSS assets used by HTML network outputs.
• src/run_merge.sbatch: SLURM wrapper to run random walks

Data Inputs

GRN Matrices (CSV)

Used by run.py with edge thresholding:

• src/data/Brain_Other.csv
• src/data/Brain_Basal_Ganglia.csv
• src/data/Brain_Cerebellum.csv

Format: first column = TF, remaining columns = target genes, cells = edge weights.

DEG Tables (CSV)

Main Python default:

• src/data/DEGDataStrictLFC.csv

Other variants included:

• DEGData.csv, DEGDataStrictestLFC.csv, DEGDataSample.csv

Expected columns:

• DISORDER, STUDY, YEAR, TISSUE, GENEID, LOG2FC, PVAL (some files also include -LOGPVAL)

Python Pipeline (src/run.py)

Run from inside src/:

cd src
python run.py

What it does:

1. Reads brain GRN matrices and filters edges by weight (threshold = 1.7).
2. Merges tissue GRNs with averaged duplicate TF->gene edge weights.
3. Converts IDs to Ensembl where needed (Ensembl REST API).
4. Loads disorder DEG lists for:
   • AD, ADHD, ASD, BD, MDD, OCD, SZ
5. Builds per-disorder network variants:
   • DEG TF only (tf_grn)
   • DEG targets only (deg_grn)
   • DEG TF + DEG targets (detf_deggrn, when distinct)
6. Computes regulator scores, edge/log2FC summaries, and pairwise Jaccard overlaps.
7. Writes per-disorder and cross-disorder outputs to src/results/.

Key Python outputs:

• src/results/deggrn_disorder_summary.csv
• src/results/deggrn_overlap_summary.csv
• src/results/deggrn_jaccard_*.csv
• src/results/overlay_tf_grn_disorders.html
• src/results/pyviz_*_deggrn.html and src/results/deggrn_*.html

R Pipeline (src/thesisp1.R)

Run from inside src/:

cd src
Rscript thesisp1.R

High-level steps:

1. Load and clean DEG table (DEGData.csv)
2. Collapse to per-gene/per-disorder meta effects
3. Annotate with HGNC/Entrez (biomaRt)
4. GO/KEGG enrichment and GSEA
5. Pairwise overlap statistics (Fisher/permutation variants)
6. Cross-disorder visualizations and network hub analyses
7. Export tables/figures under src/results/tables, src/results/figures, src/results/networks, src/results/enrichment

Environment Requirements

Python

Minimum packages used by core pipeline:

• requests
• numpy
• matplotlib
• networkx
• pyvis
• urllib3

R

Scripts use many packages, including:

• CRAN: tidyverse, dplyr, data.table, ggplot2, igraph, ggraph, pheatmap, circlize, RColorBrewer, ggupset
• Bioconductor: biomaRt, clusterProfiler, org.Hs.eg.db, enrichplot, ComplexHeatmap, DOSE, GEOquery, limma

Notes

• run.py assumes execution from src/ (relative paths like data/... and results/...).
• Ensembl ID conversion in Python requires internet access to Ensembl REST.
• R annotation/enrichment steps require internet access for some biomaRt/Bioc calls.
• src/p1thesis.R and other p1*.R scripts are older/experimental variants; thesisp1.R is the most complete end-to-end R workflow in this repo.