Query about difference in results

[Grace Mathew]

Dear WebGestalt team,

I am writing to ask for clarification about a difference I observe between results generated with the WebGestalt web server and those obtained with the WebGestaltR package for the same dataset.

For ORA on GO Biological Process (and similarly for CC and MF), I used:

• Organism: Homo sapiens

• Method: ORA

• Gene ID type: Gene symbol

• Interest gene list: [same list used in both]

• Reference gene list (background): [same list used in both]

• Database: geneontology_Biological_Process

In R, I used:

r

result_BP <- WebGestaltR( enrichMethod = "ORA", organism = "hsapiens", interestGene = gene_list, referenceGene = background_genes, interestGeneType = "genesymbol", referenceGeneType = "genesymbol", enrichDatabase = "geneontology_Biological_Process", minNum = 5, maxNum = 2000, fdrMethod = "BH", sigMethod = "fdr", fdrThr = 0.05, setCoverNum = 10, reportNum = 40, dagColor = "continuous", isOutput = TRUE, outputDirectory = "ORA_results", projectName = "GO_BP" )

On the web server, I set the “Advanced parameters” to:

• Minimum number of analytes for a category: 5

• Maximum number of analytes for a category: 2000

• Multiple test adjustment: BH

• Significance level: FDR 0.05

• Redundancy removal: Weighted set cover (fast)

• Number of categories expected from set cover: 10

• Number of clusters (k) for k‑Medoid: 10 (not selected)

• Number of categories visualized in the report: 40

• Color in DAG: Continuous

Despite using the same gene list, background, organism, GO category, and what I believe are matching advanced parameters, the results are not identical between the web server and the R package. For example:

• Total number of enriched GO BP terms differs: web tool = 42, R Code= 126]

• Mapped input: web tool = 10509, R Code= 10545

• Gene Set: web tool = 798, R Code = 780

• Overlap: web tool = 610, R code = 619

• Some GO terms appear only in one output but not the other.

Could you please help me understand:

1. Whether the web server and the current CRAN version of WebGestaltR use exactly the same backend, databases, and default settings for ORA.

2. If there are any additional parameters (e.g., related to redundancy removal, filtering, or version of GO / annotation) that I should set in WebGestaltR to reproduce the web server results exactly.

3. Whether the observed differences might be due to database version or other updates on the server side compared with the R package.

I would be happy to provide example input files (gene list and background), output tables if that would help in diagnosing the issue.

Thank you very much for your time and for providing such a useful tool.

Best regards.

[John Elizarraras]

Your first question addresses the cause. The WebGesalt server uses a newer version of both the R package and the underlying data. Because of some difficulties with using Rust in packages published on CRAN, you have to install the new package from our GitHub repository. Instructions are here: https://bzhanglab.github.io/WebGestaltR/. The CRAN package only has access to the 2019 version of the data, while the new package uses GMT files from 2024. 

If you are good with the R package results that you have, we still maintain a 2019 version of the web server: https://2019.webgestalt.org/. However, the new package and website is much faster if you are using GSEA.