Skip to content

README.md

Latest commit

 

History

History
86 lines (63 loc) · 5.55 KB

README.md

File metadata and controls

86 lines (63 loc) · 5.55 KB
Raw

MESSI manuscript: analysis and figure code

Code that turns the output of the MESSI Nextflow pipeline into the derived tables and figures of the MESSI manuscript. Manuscript text is not in this repository.

Requirements

  • R 4.3.3. Packages used: ggplot2 4.0.0, patchwork 1.3.0, cowplot 1.1.3, dplyr 1.1.4, tidyr 1.3.1, readr 2.1.5, stringr, tibble, purrr, here 1.0.1, ggrepel 0.9.5, survival 3.8-3, data.table 1.15.4, svglite 2.1.3, png 0.1-8, msigdbr 10.0.1, msigdbdf, AnnotationDbi 1.64.1, org.Hs.eg.db 3.18.0, illuminaHumanv4.db 1.26.0.
  • GNU make and bash. The chain was run on Windows with Git Bash.

Data

Input data are on Zenodo: 10.5281/zenodo.23965904 (version used for the manuscript). The concept DOI 10.5281/zenodo.18842127 resolves to the latest version. Nothing under data/raw/ or data/processed/ is tracked.

Unpack the archives so that data/raw/ has the layout below:

unzip final_pipeline_classification_results.zip -d data/raw
unzip final_pipeline_survival_results.zip -d data/raw
unzip known_gene_sets.zip      # run from the repository root
data/raw/
  final_classification_results/{bulk,multimodal,covid_only,htx_only,pca50-all-data,simulated}/copied_results_*/
  final_survival_results/copied_results_*/
  known_gene_sets/

Two gene lists are not redistributed and must be added to data/raw/known_gene_sets/ under these exact names. make process and make figures stop at recall_at_k.R without them.

File Source Used here
NCG_cancerdrivers_annotation_supporting_evidence.tsv Network of Cancer Genes, http://network-cancer-genes.org/ NCG 7.1
sfari_autism.csv SFARI Gene, https://gene.sfari.org/database/human-gene/ Downloaded October 2026

A different release changes the known-gene tables. With the files used for the manuscript, recall_at_k.R prints [gene sets] cancer=556, AD=703, DKD=227, autism=1289. The Hallmark allograft-rejection set comes from msigdbr.

Run

From the repository root:

make process   # data/raw -> derived tables (data/processed, results): about 2 minutes
make figures   # derived tables -> figures: about 3 minutes

make figures ends by sorting the figures:

  • results/manuscript_figures/: the figures in the manuscript, named by their number in the manuscript (table below).
  • results/other_figures/: everything else the scripts write, such as single panels and figures not used in the manuscript.

Times are from a Windows laptop. Logs go to logs/. Each step stops at the first failing script. Figures and derived tables are not tracked; they are written when you run the chain.

src/biology_validation/run_fig4_null.R is not part of make figures. It computes the permutation null behind Fig 4C and Supplementary Fig. S5 (about 6 minutes on 16 cores) and writes the two tables tracked under results/biology_validation/tables/, which figure4_biology.R reads. Set MESSI_NULL_WORK to keep its cache and MESSI_NULL_OUT to write elsewhere.

Figures and scripts

Fig 1, Fig 2 and Fig 5A are BioRender artwork and have no script. Fig 5A is results/sc_new/sc_design_concept.png.

Manuscript figure File in results/manuscript_figures/ Script
Fig 1 Fig1_overview Artwork: data_design_figs/Fig1_overview.png
Fig 2 Fig2_architecture Artwork: data_design_figs/Fig2_architecture.png
Fig 3 Fig3_bulk_benchmark src/benchmark_stats/figure3_bulk.R
Fig 4 Fig4_biology_validity src/biology_validation/figure4_biology.R
Fig 5 Fig5_single_cell src/benchmark_stats/figure_sc_main.R, figure5_single_cell.R
Fig 6 Fig6_simulation src/simulated_brief/figure6_simulation.R
Fig 7 Fig7_survival src/survival/figure7_survival.R
Supplementary Fig. S1 FigS1_datasets_overview src/benchmark_stats/plot_benchmark_overview.R
Supplementary Fig. S2 FigS2_CD_diagrams src/benchmark_stats/plot_benchmark_stats.R
Supplementary Fig. S3A FigS3A_PCA50_vs_raw src/benchmark_stats/pca50_baseline_comparison.R
Supplementary Fig. S3B, C FigS3BC_single_modality src/benchmark_stats/single_modality_baseline.R
Supplementary Fig. S4 FigS4_bulk_cost src/cost/plot_cost.R
Supplementary Fig. S5 FigS5_known_gene_per_dataset src/biology_validation/figure4_biology.R
Supplementary Fig. S6 FigS6_single_cell_variants src/benchmark_stats/figure_sc_main.R, figure5_single_cell.R
Supplementary Fig. S7A FigS7A_sim_prediction src/simulated_brief/figure6_simulation.R
Supplementary Fig. S7B FigS7B_sim_recovery src/simulated_brief/figure6_simulation.R
Supplementary Fig. S8 FigS8_survival_per_dataset src/survival/figure7_survival.R
Supplementary Fig. S9 FigS9_calibration src/benchmark_stats/plot_benchmark_stats.R

Other folders

  • scripts/: the build scripts called by the Makefile.
  • src/prep/make_short_paths.sh: copies raw files with long paths to data/processed/*_src/, because R on Windows cannot open paths over 260 characters.
  • simulated_data_gen_code/: generator of the 45 simulated datasets (create_data.R, simulation.R; seed 101) and the MAE and MuData converter (converter_mae_mudata/).
  • src/cleaning_scripts/, src/get_dataset_metadata.R, src/table1_dataset_description/, src/table2_method_description/: earlier data preparation and table scripts. They are not part of the chain, and some expect input paths that no longer exist.