FUMA snp2gene job timeout error (Job ID: 746341) – request for assistance

[Zhi'ang Cheng]

Dear FUMA support team,

I am writing to report an issue encountered during the execution of a SNP-to-gene mapping job on the FUMA platform.

Job details:

• Job ID: 746341

• Job title: AD_Metaanalysis_Freeze3_GxS_fulldc_allSNP_positional10kb_MHCincl

• Step: snp2gene

• Error message: ERROR: timeout

• Description: The job exceeded the 8-hour processing time limit and was terminated before completion.

This analysis includes a full GWAS meta-analysis dataset with positional mapping (10kb) and the MHC region included. I understand that this type of analysis may be computationally intensive, and I would like to kindly ask for your advice on how to resolve or prevent this issue.

Could you please help clarify:

1. Whether this timeout is due to server load or dataset complexity

2. Recommended strategies to reduce computation time (e.g., parameter settings, filtering thresholds, or excluding MHC region)

3. Whether it is possible to rerun or resume the same job without restarting from scratch

4. Any best practices for handling large-scale GWAS summary statistics in FUMA

I would greatly appreciate your guidance on how to successfully complete this analysis.

Thank you very much for your time and support.

Best regards,

Charlie Cheng

[Tanya Phung]

Hi Charlie, 

It appears that the pipeline was stuck at the step to define genomic risk loci. Could you send the input GWAS sumstat that you use to submit to FUMA? I will take a look next week. 

Best,
Tanya

[Zhi'ang Cheng]

The input GWAS sumstat was as followed: https://drive.google.com/file/d/1pM1T076sUVrYdo5RKJ2vuhed9Zut7Eg_/view?usp=sharing

[Zhi'ang Cheng]

[Tanya Phung]

Hi Charlie, 

Thank you for sending the input gwas sumstat. 

The file is relatively large and that's part of the reason why the job is exceeding 8 hours. You can try to run per chromosome or combine a few chromosomes. Perhaps try first with just chromosome 22 to check how long it takes to finish. 

When testing with a subset of chromosome 1 only, one other that I noticed is that the lead snps are very close together and it causes the problem of one large genomic risk loci. The logic of generating the genomic risk loci in FUMA has not been changed since the original version and it generally works well so I think that it could be some property of your data. One thing that you can do is to lower the merge distance: 

Hope that helps,
Tanya