checking GO enrichment of the top relationships for C18orf21 and C1orf131 in GWPS in the notebook map_evaluation_comparison_clean.ipynb is outputing empty dataframes

[brunicardoso]

Hello,

Thanks for the great framework!

Code for checking GO enrichment of the top relationships for C18orf21 and C1orf131 in GWPS in the notebook map_evaluation_comparison_clean.ipynb is outputing empty dataframes.

I was not able to find a GO file in the benchmark annotation folder. Not sure if that is the issue source.

I am running the notebook in Google Colab and have cloned repo.

Thank you!

Alex

[brunicardoso]

I noticed the file c5.go.v2023.2.Hs.symbols.gmt that comes with the installation (pip install) was probably corrupted since it only had 1kb. I replaced it with a file I downloaded directly from https://www.gsea-msigdb.org/and could reproduce the output of the original notebook.
Also, pointing to the folder path directly cloned from the repo (in which the file is fine) in the enrichment function parameters worked as well. See code below.

benchmark_data_dir` = 'EFAAR_benchmarking/efaar_benchmarking/benchmark_annotations/'
dataset = 'GWPS'
src = "GO"
topx = 25
enr_p_lim = .01
for gene in ['C18orf21', 'C1orf131']:
    print(gene)
    res = compute_top_similars(maps[dataset], pert_label_cols[dataset], gene, topx=topx)
    print(', '.join(sorted(res.pert)))
    print(enrichment(res.pert, list(maps[dataset].metadata[pert_label_cols[dataset]]), source=src, benchmark_data_dir = benchmark_data_dir, pval_thr=enr_p_lim)[['cluster', 'pval']])
    print('\n')

[safiyecelik]

Hi Alex,

Thanks for reaching out, and my apologies for the delayed response - the original GitHub notification got buried in my inbox.

This repo relies on Git Large File Storage (LFS). Instead of storing large data files (like the GO annotation file you mentioned) directly in the repo, we use small pointer files. The actual data lives on a separate LFS server. This approach keeps the main repo lightweight and efficient.

If Git LFS isn't installed and configured on your system when you clone the repo, Git will only download these small pointer files. The actual large files needed by map_evaluation_comparison_clean.ipynb will be missing from your local copy. This is almost certainly why you're seeing empty dataframes and couldn't find the GO file. You can confirm which files are being tracked by LFS with the command: !git lfs ls-files.

While downloading the annotations from MSigDB got the notebook running for you, properly utilizing Git LFS would guarantee you're working with the exact version of the GO annotations that the notebook was originally tested with. I recommend you install and initialize Git LFS in your Google Colab environment (using below commands) and then re-clone the repository. This will ensure all the necessary large files are downloaded correctly.

!sudo apt-get update
!sudo apt-get install git-lfs
!git lfs install

Let me know if you have any further questions.

Best,
Safiye