Stan for multi-omics data integration

I am working on multi-omics data integration and using Stan for Bayesian inference. Do we have a multi-omics specialized field guide (something similar to Stan for Ecology, Stan for Cognitive Science)?

3 Likes

I’m not aware of any existing group but I’d be keen to help set one up as I also work with Stan in this area.

2 Likes

Thank you for replying. I’m not sure how much work involved in setting up a group. I’ll be happy to contribute to it.

@martinmodrak @stemangiola

I would also be in! Although I’m not experienced in setting up groups

1 Like

Do you have anything in mind, like what kind of resource you’d like to see in a field guide for omics? I rewrote some of the typical RNA-seq analysis method in Stan (which are basically a GLM, which I mae hierarchical, since the most popular packages don’t support random/mixed effects), and a few other models for that kind of data. But omics is pretty vast, so I’m not sure what a “basic toolbox” and resource guide would look like.

1 Like

there are a bunch of field guides linked to from here: https://mc-stan.org/users/documentation/
scroll down and you’ll find “Specialized Field Guides” -
e.g.:
https://cognitive-science-stan.github.io
https://github.com/cognitive-science-stan/cognitive-science-stan.github.io

if you set something up, happy to list it

2 Likes

I would imagine having an omics field guide for different omics types - microbiome-omics, bulk RNA seq, sc RNA-seq, sc proteomics, spatial sc omics, spatial omics. There are published papers that proposed Bayesian methods for microbiome-omics and sc data, some papers used Stan, but not many do it.
It seems that there are a lot of methods overlapping for the above omics data. I thought of having a specialized field guide, so that would be helpful to reduce redundancy. This guide can also point to Stan Bayesian inference for multi-table, multi-view, multi-domain data.

1 Like

When you say field guide I imagine something like Broad Institute’s Genome Analysis Toolkit (GATK) best practices, although that is essentially bioinformatics pipelines with very little if any bayesian statistics (that I know of, or is exposed in any way).

The actual analyses probably have Bayesian versions, like you mentioned, but it would require that someone compiles them and looks into the methods and maybe reimplements them (or snippets) in Stan. It’s probably quite a bit of work, but there are probably people out there (some already above) willing to contribute if it’s lined up.

I’m mostly just passing by the omics field, but I’d be happy to share my Bayesian version of standard RNA-seq analysis methods (and maybe nonstandard ones as well), which I’m probably doing anyway upon publication.