computation PRS score from Pleio results

[sevi2018]

Hello, I ran the --blup option and I obtained a Beta for each trait that I used in the meta-analysis. Is there another option to compute also the SE? Also my understanding from some of the previous issues, you do not recommend combining the betas for each trait, so which method would you suggest to apply for PRS computation?

Thank you,

[cuelee]

Hello sevi2018,

Thanks for sharing this issue.

At present, PLEIO does not offer Standard Errors (SE) for BLUP estimates. Upon previous consideration, I concluded that adding this feature might not significantly enhance PLEIO’s functionality. This is mainly because Polygenic Risk Scores (PRS) require a focused understanding of target-specific disease signals, whereas PLEIO’s BLUP approach primarily captures the shared pleiotropic signal across traits.

I understand that PLEIO’s BLUP estimates might represent a diluted form of target-specific effects, influenced by a set of non-target traits. Therefore, an alternative approach worth considering is the adjustment of the target's observed effect, as obtained from raw GWAS data, using the BLUP estimates specific to that target. This approach could potentially be implemented without the need to estimate SE(BLUP).

This concept, though untested, aligns with methodologies used in other software tools like MTAG, which have demonstrated efficacy in enhancing PRS estimates. The idea of integrating pleiotropic signals, as proposed here, seems akin to the strategy used by MTAG.

I hope this clarification helps address your inquiries more comprehensively. I welcome any further questions you may have and am open to discussions or collaborations that could explore this methodology in greater depth.

Best regards, Cue

[sevi2018]

Thank you for your answer,

We recently published a paper using PLEIO, and since we shared the summary statistics people are starting to ask for information on beta and se. Along the review process, we have been asked to validate the loci identified with MTAG, so I guess at the moment we will share those estimates. However, it would be great to be included in a more detailed discussion and possible collaboration. Happy to talk. Best

On Wed, Jan 31, 2024 at 9:16 AM Cue Hyunkyu Lee @.***> wrote:

Hello sevi2018,

Thanks for sharing this issue.

At present, PLEIO does not offer Standard Errors (SE) for BLUP estimates, and upon careful consideration, I have concluded that adding this feature might not significantly enhance PLEIO’s functionality. This is mainly because Polygenic Risk Scores (PRS) require a focused understanding of target-specific disease signals, whereas PLEIO’s BLUP approach primarily captures the shared pleiotropic signal across traits.

I understand that PLEIO’s BLUP estimates might represent a diluted form of target-specific effects, influenced by a set of non-target traits. Therefore, an alternative approach worth considering is the adjustment of the target's observed effect, as obtained from raw GWAS data, using the BLUP estimates specific to that target. This approach could potentially be implemented without the need to estimate SE(BLUP).

This concept, though untested, aligns with methodologies used in other software tools like MTAG, which have demonstrated efficacy in enhancing PRS estimates. The idea of integrating pleiotropic signals, as proposed here, seems akin to the strategy used by MTAG.

I hope this clarification helps address your inquiries more comprehensively. I welcome any further questions you may have and am open to discussions or collaborations that could explore this methodology in greater depth.

Best regards, Cue

— Reply to this email directly, view it on GitHub https://github.com/cuelee/pleio/issues/25#issuecomment-1919191138, or unsubscribe https://github.com/notifications/unsubscribe-auth/AJHJD3NMFWPM26TZ4KLUQETYRJG2TAVCNFSM6AAAAABCP3J5GSVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSMJZGE4TCMJTHA . You are receiving this because you authored the thread.Message ID: @.***>