BIGA

[Feng Yu]

Thanks a lot for the BIGA platform, I would like to ask about how long it takes to run a Popcorn massive analysis? And will LAVA be able to run databases other than PGC in the future?

[yujue23]

Hello,

Thank you for using BIGA! Currently, processing one trait in Popcorn massive analysis takes approximately 25 minutes.

For those utilizing the CHIMGEN dataset,  it will require 18 hours in total for the Popcorn analysis plus 2-4 hours data harmonization. For those using the Biobank Japan dataset, it will require about 70 hours plus 2-4 hours of data harmonization.

Currently, all LDSC and SumHer massive analyses are finished within 24 hours, and there is an upcoming upgrade to our job processing system very soon, which will ensure that all Popcorn and LAVA jobs are completed within a 24-hour period.

For your question in terms of LAVA, we have prepared all curated datasets but we haven't added it to our platform due to the extensive processing time. After this upgrade, we will add more datasets to LAVA and Popcorn analysis.

We also would like to hear your preferences regarding datasets with LAVA and Popcorn analysis. What would you like us to include next?

Best,

Yujue

[Feng Yu]

Thank you very much for your reply, I tried to run a POPCORN analysis and I found that there are many traits that have a corr_p of 0. Should we consider the value as 1 E-300 or NA?For the LAVA analysis, I'm interested in all datasets except PGC hahaha.

Looking forward to your reply.
Yu

[yujue23]

Hi Yu,

The value in corr_p_1 is provided by Popcorn software itself and represents the p-value for the null hypothesis that the correlation is 1. I believe it is not 'NA' but rather a number very close to 0 (Popcorn gives 0 directly in their outputs).

From my experience, when sorting the corr_p_1 values in descending order, the first few entries are 0, followed by values such as 3.3307e-16. This indicates that the p-value is less than 3.3307e-16.

The corr_p_0 is provided by BIGA and represents the p-value for the null hypothesis that the correlation is 0. This allows for p-values with higher precision, often less than 1E-300.

If you have any questions, please feel free to ask. Additionally, we will be adding more datasets into LAVA soon!

Best,

Yujue

[Feng Yu]

Thank you in particular for your response, so when I write the methods section of the paper, can I go ahead and write it like this, "For the results of the Popcorn analysis, we use the p-value of the null hypothesis with a correlation of 0 because this allows the p-value to have a higher precision."

Looking forward to your reply.
Yu

[yujue23]

Hi Yu,

Both p values are calculated by z score and standard error with different hypothesis of correlation (cor=0 or 1),  I think which p-value you choose to use depends on what your hypothesis is.

If you want to test whether the correlation is 1, you should still use the p-value of corr_p_1, or you can calculate it by yourself based on z-score and standard error provided by Popcorn. BIGA use `scipy.stats.chi2` function to get a higher precision of p value.

Best,

Yujue