Hi Sam,

sounds useful. I also suggest you take a look at what database you are using. They are not all created equal. These papers should help.

Miryala, Sravan Kumar, Anand Anbarasu, and Sudha Ramaiah. “Discerning Molecular Interactions: A Comprehensive Review on Biomolecular Interaction Databases and Network Analysis Tools.” Gene 642 (February 5, 2018): 84–94. https://doi.org/10.1016/j.gene.2017.11.028.

Mubeen, Sarah, Charles Tapley Hoyt, André Gemünd, Martin Hofmann-Apitius, Holger Fröhlich, and Daniel Domingo-Fernández. “The Impact of Pathway Database Choice on Statistical Enrichment Analysis and Predictive Modelling.” Frontiers in Genetics 10 (2019). https://www.frontiersin.org/articles/10.3389/fgene.2019.01203.

Thank you Peter,

Those look like useful articles.

I'm also conscious that some of these tools were developed for RNA data rather than proteins.

Sam

I personally think one of the biggest issues is going from protein code to gene code prior to pathway analysis etc. It's really a "non-trivial" exercise and there are numerous cases where one protein does not equal one gene and vice versa. How is this handled? There are potentially significant effects on the data analysis and interpretation. We should at least flag proteins/genes where there is ambiguity.

I'm currently using the protein IDs as produced by the Max Quant of Proteome Discoverer output.

I've used `clusterProfiler` for GO term enrichment initially, which accepts these IDs, however there is a chance that this package maps the protein IDs to gene IDs prior to network analysis.

Hence I'm keen to find a package designed with protein analysis in mind.

Hi, I don't know the extent of your shiny application, but you can also use EnrichR (in R), which is also good, though, it upload the data to server during calculation.

Plus, I am curious how you will handle the missing IDs during ID mapping? ClusterProfiler can map IDs, but there are always some % of ID that are missing.