Are there any examples of calculating the genetic risk score from different SNPs for a specific disease (e.g., diabetes) on the Research Analysis Platform (RAP)?
That said, here is an example of a run down of PRS techniques. https://pubmed.ncbi.nlm.nih.gov/34304866/. This and other techniques (PRSice for example) can quickly be implemented on the UKBRAP using Jupyter notebooks, cloud workstations, or building them into Apps.
Most tools are on swiss army knife and there are a few that can be uploaded using Jupyter lab. I haven't started working on it yet but it would be great if we could help each other out in the near future.
I have considered it, and contacted the author of PRSICE-2. He stated that there are some memory issues with using it on imputed data with bgen format. The app on DNAnexus does not allow for the bgen format to be used, only bed, bim, fam, which is the format used for the genotyped data. If you figure out a way around this, please let me know. For now, I plan to use the pipeline instead. Make sure that the polygenic score used comes from an external cohort or a meta-analysis of GWAS external to UK biobank.
I'm rather new to the RAP, and I'm also looking into PRS calculations. Thanks for the pipeline tip. I noticed that UK Biobank has released som calculated PRS data fields. According to the release timeline these were made available in July. However I can not retrieve these fields using either spark or cohort browser. Has anyone else had any luck?
Ah, I just noticed that the latest UK Biobank release hasn't arrived to the RAP yet according to the Data Release page. That explains why I couldn't find the data fields.
Comments
9 comments
Do you mean Polygenic Risk Scores (PRS), specifically?
Yes. Polygenic Risk Scores (PRS).
Hi Fu-Rong!
I just checked PubMed and there are a few articles: https://pubmed.ncbi.nlm.nih.gov/?term=polygenic+risk+score+ukb, https://pubmed.ncbi.nlm.nih.gov/?term=PRS+AND+UKBiobank
That said, there are a fair amount of UKBioBank projects that address PRS -- searchable here. https://www.ukbiobank.ac.uk/enable-your-research/approved-research?query=polygenic+risk+score#articles
That said, here is an example of a run down of PRS techniques. https://pubmed.ncbi.nlm.nih.gov/34304866/. This and other techniques (PRSice for example) can quickly be implemented on the UKBRAP using Jupyter notebooks, cloud workstations, or building them into Apps.
I've previously built PRSice2 into a Jupyter notebook in a few minutes. Check this out and then it should be pretty simple to build the logic into an R notebook: https://www.prsice.info. Heres a quick spin-up on UKBRAP jupyter notebooks in case you need it: https://dnanexus.gitbook.io/uk-biobank-rap/working-on-the-research-analysis-platform/using-jupyterlab-on-the-research-analysis-platform.
If you wanted to use this in a less jupytery way, you could always build an app out of it. If you wanted to get really fancy, you could probably use papermill, but the most straightforward way IMO is to use the dx-app-wizard. Check this out if you decide to go there: https://dnanexus.gitbook.io/uk-biobank-rap/working-on-the-research-analysis-platform/bringing-your-app-or-workflow-to-the-research-analysis-platform.
Lemme know if youre looking for something more specific -- always happy to chat.
Ben
There is a new manuscript that just got published. Here is the link for the pipeline:
https://2cjenn.github.io/PRS_Pipeline/#SNP_Extraction
Most tools are on swiss army knife and there are a few that can be uploaded using Jupyter lab. I haven't started working on it yet but it would be great if we could help each other out in the near future.
Thanks!
May
Hi May @May Baydoun? ,
Have you used PRSice-2 for PRS calculation? Do you know any resources particularly helpful for this analysis?
Sincerely,
Anand
Hi Anand,
I have considered it, and contacted the author of PRSICE-2. He stated that there are some memory issues with using it on imputed data with bgen format. The app on DNAnexus does not allow for the bgen format to be used, only bed, bim, fam, which is the format used for the genotyped data. If you figure out a way around this, please let me know. For now, I plan to use the pipeline instead. Make sure that the polygenic score used comes from an external cohort or a meta-analysis of GWAS external to UK biobank.
May
Hi May,
Thanks for the details.
I am just starting with the analysis so I think I will follow the pipeline you mentioned. I will update you once I go through the PRS Pipeline paper.
Sincerely,
Anand
Hi!
I'm rather new to the RAP, and I'm also looking into PRS calculations. Thanks for the pipeline tip. I noticed that UK Biobank has released som calculated PRS data fields. According to the release timeline these were made available in July. However I can not retrieve these fields using either spark or cohort browser. Has anyone else had any luck?
Johan
Ah, I just noticed that the latest UK Biobank release hasn't arrived to the RAP yet according to the Data Release page. That explains why I couldn't find the data fields.
Please sign in to leave a comment.