how to combine PRS from each chromosomes？

[hxckk]

Hello,

I am currently following a PRS (Polygenic Risk Score) calculation tutorial, and I have encountered a situation where:

a) I successfully generated the best PRS when considering the entire .bed file as a whole.

b) I divided the entire .bed file into 22 sub-files by chromosome and followed the same guidelines to calculate the best PRS for each chromosome.

However, when I attempted to aggregate all the best PRS results from (b), the outcome did not match the result from (a).

As such, I would like to inquire about the steps I should take to ensure that I obtain the same result as in (a) when calculating PRS for each individual chromosome. Your guidance on this matter would be greatly appreciated.

Thank you for your assistance.

[choishingwan]

Problem here is that the default is average score, which has a denominator of number of alleles included in the score (accounting for missingness). If you want to do it per chromosome, you can only use sum score (--score sum)

On Wed, Oct 25, 2023, 11:50 AM hxckk @.***> wrote:

Hello,

I am currently following a PRS (Polygenic Risk Score) calculation tutorial, and I have encountered a situation where:

a) I successfully generated the best PRS when considering the entire .bed file as a whole.

b) I divided the entire .bed file into 22 sub-files by chromosome and followed the same guidelines to calculate the best PRS for each chromosome.

However, when I attempted to aggregate all the best PRS results from (b), the outcome did not match the result from (a).

As such, I would like to inquire about the steps I should take to ensure that I obtain the same result as in (a) when calculating PRS for each individual chromosome. Your guidance on this matter would be greatly appreciated.

Thank you for your assistance.

— Reply to this email directly, view it on GitHub https://github.com/choishingwan/PRSice/issues/346, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYTTKXFV4TQMGZAX2E3YBEYK5AVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43ASLTON2WKOZRHE3DCNZUGEYDSNQ . You are receiving this because you are subscribed to this thread.Message ID: @.***>

[hxckk]

Hello,

Thank you for your prompt response! I appreciate your assistance.

I've tried using the --score sum command as you suggested, but unfortunately, I still encountered the same issue.

I've been closely following the PRS tutorial, and I successfully obtained the necessary data (i.e., EUR.QC.bed, EUR.QC.fam, and EUR.QC.bim). I proceeded with two separate analyses, as outlined below:

a) In the first analysis, I used the command recommended by the PRS tutorial to calculate the PRS, with the addition of --score sum as follows:

Rscript PRSice.R --prsice PRSice_linux --base Height.QC.gz --target EUR.QC --binary-target F --pheno EUR.height --cov EUR.cov --base-maf MAF:0.01 --base-info INFO:0.8 --stat OR --or --score sum --out EUR

b) In the second analysis, I divided the generated EUR.QC.bed file into 22 separate .bed files, each corresponding to one of the 22 chromosomes. I then calculated the PRS for each chromosome using the --score sum option, as shown in the script below:

for q in $(seq 1 22); do plink --bfile EUR.QC --chr $q --make-bed --out chr$q done

for chr in 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22; do Rscript PRSice.R --prsice PRSice_linux --base Height.QC.gz --target chr${chr} --binary-target F --pheno EUR.height --cov EUR.cov --base-maf MAF:0.01 --base-info INFO:0.8 --stat OR --or --score sum --out chr${chr} done

However, when I attempted to sum up the PRS results from each chromosome, the cumulative result did not match the outcome obtained in analysis a). (please refer to the attached document for details)

Your guidance on how to obtain consistent results when calculating PRS for each chromosome would be greatly appreciated. Thank you for your time and support.

-----原始邮件----- 发件人:"Shing Wan Choi" @.> 发送时间:2023-10-26 06:55:45 (星期四) 收件人: choishingwan/PRSice @.> 抄送: hxckk @.>, Author @.> 主题: Re: [choishingwan/PRSice] how to combine PRS from each chromosomes？ (Issue #346)

Problem here is that the default is average score, which has a denominator of number of alleles included in the score (accounting for missingness). If you want to do it per chromosome, you can only use sum score (--score sum)

On Wed, Oct 25, 2023, 11:50 AM hxckk @.***> wrote:

Hello,

I am currently following a PRS (Polygenic Risk Score) calculation tutorial, and I have encountered a situation where:

a) I successfully generated the best PRS when considering the entire .bed file as a whole.

b) I divided the entire .bed file into 22 sub-files by chromosome and followed the same guidelines to calculate the best PRS for each chromosome.

However, when I attempted to aggregate all the best PRS results from (b), the outcome did not match the result from (a).

As such, I would like to inquire about the steps I should take to ensure that I obtain the same result as in (a) when calculating PRS for each individual chromosome. Your guidance on this matter would be greatly appreciated.

Thank you for your assistance.

— Reply to this email directly, view it on GitHub https://github.com/choishingwan/PRSice/issues/346, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYTTKXFV4TQMGZAX2E3YBEYK5AVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43ASLTON2WKOZRHE3DCNZUGEYDSNQ . You are receiving this because you are subscribed to this thread.Message ID: @.***>

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

[hxckk]

Problem here is that the default is average score, which has a denominator of number of alleles included in the score (accounting for missingness). If you want to do it per chromosome, you can only use sum score (--score sum)

Hi Mr. Shing Wan Choi,

I attempted the task again, following the same code as previously, but I still cannot achieve results in (a) that match those in (b). What could be the issue? I am looking forward to your response. Thank you!

[choishingwan]

Try it with just one threshold e.g. --fastscore --bar-levels 1, to see if there's problem with your code regarding summing the scores Sam

On Fri, Oct 27, 2023 at 1:16 AM hxckk @.***> wrote:

Problem here is that the default is average score, which has a denominator of number of alleles included in the score (accounting for missingness). If you want to do it per chromosome, you can only use sum score (--score sum)

Hi Mr. Shing Wan Choi,

I attempted the task again, following the same code as previously, but I still cannot achieve results in (a) that match those in (b). What could be the issue? I am looking forward to your response. Thank you!

— Reply to this email directly, view it on GitHub https://github.com/choishingwan/PRSice/issues/346#issuecomment-1782307824, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ . You are receiving this because you commented.Message ID: @.***>

[hxckk]

Try it with just one threshold e.g. --fastscore --bar-levels 1, to see if there's problem with your code regarding summing the scores Sam … On Fri, Oct 27, 2023 at 1:16 AM hxckk @.> wrote: Problem here is that the default is average score, which has a denominator of number of alleles included in the score (accounting for missingness). If you want to do it per chromosome, you can only use sum score (--score sum) Hi Mr. Shing Wan Choi, I attempted the task again, following the same code as previously, but I still cannot achieve results in (a) that match those in (b). What could be the issue? I am looking forward to your response. Thank you! — Reply to this email directly, view it on GitHub <#346 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ . You are receiving this because you commented.Message ID: @.>

Hello Sam,

Thank you very much for your response. I have tried using "--fastscore --bar-levels 1" for both analyses a) and b), and this time the results matched! This indicates that the code I used was correct.

So, does this mean I should specify the same threshold for all chromosomes in all analyses?

[choishingwan]

Oh right now I see it. If you use per chromosome, you can't use the best score, you need to use --no-regress, and then do the optimization by hand, as each chromosome might select a different threshold, but we really want one threshold for the whole genome.

That's why I don't usually recommend doing c+t per chromosome, as it usually are too time consuming and io intensive for one to generate scores for 6000+ thresholds per chromosome

Sam

On Sat, Oct 28, 2023, 12:43 PM hxckk @.***> wrote:

Try it with just one threshold e.g. --fastscore --bar-levels 1, to see if there's problem with your code regarding summing the scores Sam … <#m5757006495489225859> On Fri, Oct 27, 2023 at 1:16 AM hxckk @.> wrote: Problem here is that the default is average score, which has a denominator of number of alleles included in the score (accounting for missingness). If you want to do it per chromosome, you can only use sum score (--score sum) Hi Mr. Shing Wan Choi, I attempted the task again, following the same code as previously, but I still cannot achieve results in (a) that match those in (b). What could be the issue? I am looking forward to your response. Thank you! — Reply to this email directly, view it on GitHub <#346 (comment) https://github.com/choishingwan/PRSice/issues/346#issuecomment-1782307824>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ . You are receiving this because you commented.Message ID: @.>

Hello Sam,

Thank you very much for your response. I have tried using "--fastscore --bar-levels 1" for both analyses a) and b), and this time the results matched! This indicates that the code I used was correct.

So, does this mean I should specify the same threshold for all chromosomes in all analyses?

— Reply to this email directly, view it on GitHub https://github.com/choishingwan/PRSice/issues/346#issuecomment-1783867029, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYWJHPWCVUVNIDI6Z6TYBUY4RAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTHA3DOMBSHE . You are receiving this because you commented.Message ID: @.***>

[hxckk]

Oh right now I see it. If you use per chromosome, you can't use the best score, you need to use --no-regress, and then do the optimization by hand, as each chromosome might select a different threshold, but we really want one threshold for the whole genome. That's why I don't usually recommend doing c+t per chromosome, as it usually are too time consuming and io intensive for one to generate scores for 6000+ thresholds per chromosome Sam … On Sat, Oct 28, 2023, 12:43 PM hxckk @.> wrote: Try it with just one threshold e.g. --fastscore --bar-levels 1, to see if there's problem with your code regarding summing the scores Sam … <#m5757006495489225859> On Fri, Oct 27, 2023 at 1:16 AM hxckk @.> wrote: Problem here is that the default is average score, which has a denominator of number of alleles included in the score (accounting for missingness). If you want to do it per chromosome, you can only use sum score (--score sum) Hi Mr. Shing Wan Choi, I attempted the task again, following the same code as previously, but I still cannot achieve results in (a) that match those in (b). What could be the issue? I am looking forward to your response. Thank you! — Reply to this email directly, view it on GitHub <#346 (comment) <#346 (comment)>>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ . You are receiving this because you commented.Message ID: @.> Hello Sam, Thank you very much for your response. I have tried using "--fastscore --bar-levels 1" for both analyses a) and b), and this time the results matched! This indicates that the code I used was correct. So, does this mean I should specify the same threshold for all chromosomes in all analyses? — Reply to this email directly, view it on GitHub <#346 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYWJHPWCVUVNIDI6Z6TYBUY4RAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTHA3DOMBSHE . You are receiving this because you commented.Message ID: @.>

Hello Sam,

You are absolutely right. I tried using the "--no-regress" command, and it generated a ".all_score file" for each chromosome. However, I'm uncertain about how to proceed with the next step of the analysis.

Additionally, I am planning to perform analyses using UKB imputation data, and due to storage limitations, I have to conduct PRS analysis on a per-chromosome basis. Since the C+T method is somewhat inconvenient for chromosome-specific analyses, do you have any recommendations for a more suitable method?

Thank you for your assistance!

[choishingwan]

You can use --target chr# for example which PRSice will automatically substitute # with 1-22. So as long as your file naming is consistent, this should work

On Sun, Oct 29, 2023, 11:30 AM hxckk @.***> wrote:

Oh right now I see it. If you use per chromosome, you can't use the best score, you need to use --no-regress, and then do the optimization by hand, as each chromosome might select a different threshold, but we really want one threshold for the whole genome. That's why I don't usually recommend doing c+t per chromosome, as it usually are too time consuming and io intensive for one to generate scores for 6000+ thresholds per chromosome Sam … <#m-184795729589143180> On Sat, Oct 28, 2023, 12:43 PM hxckk @.*> wrote: Try it with just one threshold e.g. --fastscore --bar-levels 1, to see if there's problem with your code regarding summing the scores Sam … <#m5757006495489225859> On Fri, Oct 27, 2023 at 1:16 AM hxckk @.> wrote: Problem here is that the default is average score, which has a denominator of number of alleles included in the score (accounting for missingness). If you want to do it per chromosome, you can only use sum score (--score sum) Hi Mr. Shing Wan Choi, I attempted the task again, following the same code as previously, but I still cannot achieve results in (a) that match those in (b). What could be the issue? I am looking forward to your response. Thank you! — Reply to this email directly, view it on GitHub <#346 https://github.com/choishingwan/PRSice/issues/346 (comment) <#346 (comment) https://github.com/choishingwan/PRSice/issues/346#issuecomment-1782307824>>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ . You are receiving this because you commented.Message ID: @.> Hello Sam, Thank you very much for your response. I have tried using "--fastscore --bar-levels 1" for both analyses a) and b), and this time the results matched! This indicates that the code I used was correct. So, does this mean I should specify the same threshold for all chromosomes in all analyses? — Reply to this email directly, view it on GitHub <#346 (comment) https://github.com/choishingwan/PRSice/issues/346#issuecomment-1783867029>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYWJHPWCVUVNIDI6Z6TYBUY4RAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTHA3DOMBSHE https://github.com/notifications/unsubscribe-auth/AAJTRYWJHPWCVUVNIDI6Z6TYBUY4RAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTHA3DOMBSHE . You are receiving this because you commented.Message ID: @.***>

Hello Sam,

You are absolutely right. I tried using the "--no-regress" command, and it generated a ".all_score file" for each chromosome. However, I'm uncertain about how to proceed with the next step of the analysis.

Additionally, I am planning to perform analyses using UKB imputation data, and due to storage limitations, I have to conduct PRS analysis on a per-chromosome basis. Since the C+T method is somewhat inconvenient for chromosome-specific analyses, do you have any recommendations for a more suitable method?

Thank you for your assistance!

— Reply to this email directly, view it on GitHub https://github.com/choishingwan/PRSice/issues/346#issuecomment-1784144502, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYR2KMKEN5NWTAOORX3YBZZARAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBUGE2DINJQGI . You are receiving this because you commented.Message ID: @.***>

[hxckk]

You can use --target chr# for example which PRSice will automatically substitute # with 1-22. So as long as your file naming is consistent, this should work … On Sun, Oct 29, 2023, 11:30 AM hxckk @.> wrote: Oh right now I see it. If you use per chromosome, you can't use the best score, you need to use --no-regress, and then do the optimization by hand, as each chromosome might select a different threshold, but we really want one threshold for the whole genome. That's why I don't usually recommend doing c+t per chromosome, as it usually are too time consuming and io intensive for one to generate scores for 6000+ thresholds per chromosome Sam … <#m-184795729589143180> On Sat, Oct 28, 2023, 12:43 PM hxckk @.> wrote: Try it with just one threshold e.g. --fastscore --bar-levels 1, to see if there's problem with your code regarding summing the scores Sam … <#m5757006495489225859> On Fri, Oct 27, 2023 at 1:16 AM hxckk @.> wrote: Problem here is that the default is average score, which has a denominator of number of alleles included in the score (accounting for missingness). If you want to do it per chromosome, you can only use sum score (--score sum) Hi Mr. Shing Wan Choi, I attempted the task again, following the same code as previously, but I still cannot achieve results in (a) that match those in (b). What could be the issue? I am looking forward to your response. Thank you! — Reply to this email directly, view it on GitHub <#346 <#346> (comment) <#346 (comment) <#346 (comment)>>>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ . You are receiving this because you commented.Message ID: @.> Hello Sam, Thank you very much for your response. I have tried using "--fastscore --bar-levels 1" for both analyses a) and b), and this time the results matched! This indicates that the code I used was correct. So, does this mean I should specify the same threshold for all chromosomes in all analyses? — Reply to this email directly, view it on GitHub <#346 (comment) <#346 (comment)>>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYWJHPWCVUVNIDI6Z6TYBUY4RAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTHA3DOMBSHE https://github.com/notifications/unsubscribe-auth/AAJTRYWJHPWCVUVNIDI6Z6TYBUY4RAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTHA3DOMBSHE . You are receiving this because you commented.Message ID: @.**> Hello Sam, You are absolutely right. I tried using the "--no-regress" command, and it generated a ".all_score file" for each chromosome. However, I'm uncertain about how to proceed with the next step of the analysis. Additionally, I am planning to perform analyses using UKB imputation data, and due to storage limitations, I have to conduct PRS analysis on a per-chromosome basis. Since the C+T method is somewhat inconvenient for chromosome-specific analyses, do you have any recommendations for a more suitable method? Thank you for your assistance! — Reply to this email directly, view it on GitHub <#346 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYR2KMKEN5NWTAOORX3YBZZARAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBUGE2DINJQGI . You are receiving this because you commented.Message ID: @.>

Hello Sam,

I wanted to express that using "target chr#" did indeed work. However, it does require placing all 22 chromosomes in the same folder.

My original intention was to analyze one chromosome at a time in a folder, primarily due to limited hard drive storage space. Following the acquisition of PRS information for a specific chromosome, my plan was to remove the other files and then proceed with the analysis of the next chromosome, repeating these steps until all 22 chromosomes had been analyzed. Afterward, I envisioned employing a specific statistical method to consolidate the PRS data from these 22 chromosomes to determine the optimal overall threshold and subsequently summing them up. Do you think this idea is feasible?

[choishingwan]

If storage is your concern, then this is completely unnecessary as the output of PRS software (specifically PRSice) scales with number of samples, not number of variant. And by using all chromosome at the same time, you can avoid generating the huge all-score file and only get the best score file, which should only contain 4 columns and should be much smaller in size than the all score file. It is also a good practice to keep all the genotype file in the same folder. But if for some reason that is not possible, then you can use the --target-list function instead, which allow a file containing the path to all genotype file as input. Hope this help

Sam

On Sun, Oct 29, 2023 at 10:24 PM hxckk @.***> wrote:

You can use --target chr# for example which PRSice will automatically substitute # with 1-22. So as long as your file naming is consistent, this should work … <#m-4242589899473809212> On Sun, Oct 29, 2023, 11:30 AM hxckk @.> wrote: Oh right now I see it. If you use per chromosome, you can't use the best score, you need to use --no-regress, and then do the optimization by hand, as each chromosome might select a different threshold, but we really want one threshold for the whole genome. That's why I don't usually recommend doing c+t per chromosome, as it usually are too time consuming and io intensive for one to generate scores for 6000+ thresholds per chromosome Sam … <#m-184795729589143180> On Sat, Oct 28, 2023, 12:43 PM hxckk @.> wrote: Try it with just one threshold e.g. --fastscore --bar-levels 1, to see if there's problem with your code regarding summing the scores Sam … <#m5757006495489225859> On Fri, Oct 27, 2023 at 1:16 AM hxckk @.> wrote: Problem here is that the default is average score, which has a denominator of number of alleles included in the score (accounting for missingness). If you want to do it per chromosome, you can only use sum score (--score sum) Hi Mr. Shing Wan Choi, I attempted the task again, following the same code as previously, but I still cannot achieve results in (a) that match those in (b). What could be the issue? I am looking forward to your response. Thank you! — Reply to this email directly, view it on GitHub <#346 https://github.com/choishingwan/PRSice/issues/346 <#346 https://github.com/choishingwan/PRSice/issues/346> (comment) <#346 https://github.com/choishingwan/PRSice/issues/346 (comment) <#346 (comment) https://github.com/choishingwan/PRSice/issues/346#issuecomment-1782307824>>>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ . You are receiving this because you commented.Message ID: @.> Hello Sam, Thank you very much for your response. I have tried using "--fastscore --bar-levels 1" for both analyses a) and b), and this time the results matched! This indicates that the code I used was correct. So, does this mean I should specify the same threshold for all chromosomes in all analyses? — Reply to this email directly, view it on GitHub <#346 https://github.com/choishingwan/PRSice/issues/346 (comment) <#346 (comment) https://github.com/choishingwan/PRSice/issues/346#issuecomment-1783867029>>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYWJHPWCVUVNIDI6Z6TYBUY4RAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTHA3DOMBSHE https://github.com/notifications/unsubscribe-auth/AAJTRYWJHPWCVUVNIDI6Z6TYBUY4RAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTHA3DOMBSHE https://github.com/notifications/unsubscribe-auth/AAJTRYWJHPWCVUVNIDI6Z6TYBUY4RAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTHA3DOMBSHE https://github.com/notifications/unsubscribe-auth/AAJTRYWJHPWCVUVNIDI6Z6TYBUY4RAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTHA3DOMBSHE . You are receiving this because you commented.Message ID: @.> Hello Sam, You are absolutely right. I tried using the "--no-regress" command, and it generated a ".all_score file" for each chromosome. However, I'm uncertain about how to proceed with the next step of the analysis. Additionally, I am planning to perform analyses using UKB imputation data, and due to storage limitations, I have to conduct PRS analysis on a per-chromosome basis. Since the C+T method is somewhat inconvenient for chromosome-specific analyses, do you have any recommendations for a more suitable method? Thank you for your assistance! — Reply to this email directly, view it on GitHub <#346 (comment) https://github.com/choishingwan/PRSice/issues/346#issuecomment-1784144502>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYR2KMKEN5NWTAOORX3YBZZARAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBUGE2DINJQGI https://github.com/notifications/unsubscribe-auth/AAJTRYR2KMKEN5NWTAOORX3YBZZARAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBUGE2DINJQGI . You are receiving this because you commented.Message ID: @.>

Hello Sam,

I wanted to express that using "target chr#" did indeed work. However, it does require placing all 22 chromosomes in the same folder.

My original intention was to analyze one chromosome at a time in a folder, primarily due to limited hard drive storage space. Following the acquisition of PRS information for a specific chromosome, my plan was to remove the other files and then proceed with the analysis of the next chromosome, repeating these steps until all 22 chromosomes had been analyzed. Afterward, I envisioned employing a specific statistical method to consolidate the PRS data from these 22 chromosomes to determine the optimal overall threshold and subsequently summing them up. Do you think this idea is feasible?

— Reply to this email directly, view it on GitHub https://github.com/choishingwan/PRSice/issues/346#issuecomment-1784376365, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYSI4NQWOYL5MGLMFADYB4FVNAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBUGM3TMMZWGU . You are receiving this because you commented.Message ID: @.***>

[hxckk]

Certainly! Thank you for your assistance, Sam. Your help is greatly appreciated!

[hxckk]

Dear Sam,

I trust this message finds you well. Following your insightful guidance, I employed PRSice and LASSO methodologies to compute Polygenic Risk Scores (PRS). However, upon reviewing the results, a discrepancy has emerged, as illustrated in the attached image. I would appreciate your expertise in elucidating the potential reasons for these divergent outcomes.

Your assistance in interpreting these results would be highly valued.

-----原始邮件----- 发件人:"Shing Wan Choi" @.> 发送时间:2023-10-30 11:00:23 (星期一) 收件人: choishingwan/PRSice @.> 抄送: hxckk @.>, Author @.> 主题: Re: [choishingwan/PRSice] how to combine PRS from each chromosomes？ (Issue #346)

If storage is your concern, then this is completely unnecessary as the output of PRS software (specifically PRSice) scales with number of samples, not number of variant. And by using all chromosome at the same time, you can avoid generating the huge all-score file and only get the best score file, which should only contain 4 columns and should be much smaller in size than the all score file. It is also a good practice to keep all the genotype file in the same folder. But if for some reason that is not possible, then you can use the --target-list function instead, which allow a file containing the path to all genotype file as input. Hope this help

Sam

On Sun, Oct 29, 2023 at 10:24 PM hxckk @.***> wrote:

You can use --target chr# for example which PRSice will automatically substitute # with 1-22. So as long as your file naming is consistent, this should work … <#m-4242589899473809212> On Sun, Oct 29, 2023, 11:30 AM hxckk @.> wrote: Oh right now I see it. If you use per chromosome, you can't use the best score, you need to use --no-regress, and then do the optimization by hand, as each chromosome might select a different threshold, but we really want one threshold for the whole genome. That's why I don't usually recommend doing c+t per chromosome, as it usually are too time consuming and io intensive for one to generate scores for 6000+ thresholds per chromosome Sam … <#m-184795729589143180> On Sat, Oct 28, 2023, 12:43 PM hxckk @.> wrote: Try it with just one threshold e.g. --fastscore --bar-levels 1, to see if there's problem with your code regarding summing the scores Sam … <#m5757006495489225859> On Fri, Oct 27, 2023 at 1:16 AM hxckk @.> wrote: Problem here is that the default is average score, which has a denominator of number of alleles included in the score (accounting for missingness). If you want to do it per chromosome, you can only use sum score (--score sum) Hi Mr. Shing Wan Choi, I attempted the task again, following the same code as previously, but I still cannot achieve results in (a) that match those in (b). What could be the issue? I am looking forward to your response. Thank you! — Reply to this email directly, view it on GitHub <#346 https://github.com/choishingwan/PRSice/issues/346 <#346 https://github.com/choishingwan/PRSice/issues/346> (comment) <#346 https://github.com/choishingwan/PRSice/issues/346 (comment) <#346 (comment) https://github.com/choishingwan/PRSice/issues/346#issuecomment-1782307824>>>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ . You are receiving this because you commented.Message ID: @.> Hello Sam, Thank you very much for your response. I have tried using "--fastscore --bar-levels 1" for both analyses a) and b), and this time the results matched! This indicates that the code I used was correct. So, does this mean I should specify the same threshold for all chromosomes in all analyses? — Reply to this email directly, view it on GitHub <#346 https://github.com/choishingwan/PRSice/issues/346 (comment) <#346 (comment) https://github.com/choishingwan/PRSice/issues/346#issuecomment-1783867029>>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYWJHPWCVUVNIDI6Z6TYBUY4RAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTHA3DOMBSHE https://github.com/notifications/unsubscribe-auth/AAJTRYWJHPWCVUVNIDI6Z6TYBUY4RAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTHA3DOMBSHE https://github.com/notifications/unsubscribe-auth/AAJTRYWJHPWCVUVNIDI6Z6TYBUY4RAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTHA3DOMBSHE https://github.com/notifications/unsubscribe-auth/AAJTRYWJHPWCVUVNIDI6Z6TYBUY4RAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTHA3DOMBSHE . You are receiving this because you commented.Message ID: @.> Hello Sam, You are absolutely right. I tried using the "--no-regress" command, and it generated a ".all_score file" for each chromosome. However, I'm uncertain about how to proceed with the next step of the analysis. Additionally, I am planning to perform analyses using UKB imputation data, and due to storage limitations, I have to conduct PRS analysis on a per-chromosome basis. Since the C+T method is somewhat inconvenient for chromosome-specific analyses, do you have any recommendations for a more suitable method? Thank you for your assistance! — Reply to this email directly, view it on GitHub <#346 (comment) https://github.com/choishingwan/PRSice/issues/346#issuecomment-1784144502>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYR2KMKEN5NWTAOORX3YBZZARAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBUGE2DINJQGI https://github.com/notifications/unsubscribe-auth/AAJTRYR2KMKEN5NWTAOORX3YBZZARAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBUGE2DINJQGI . You are receiving this because you commented.Message ID: @.>

Hello Sam,

I wanted to express that using "target chr#" did indeed work. However, it does require placing all 22 chromosomes in the same folder.

My original intention was to analyze one chromosome at a time in a folder, primarily due to limited hard drive storage space. Following the acquisition of PRS information for a specific chromosome, my plan was to remove the other files and then proceed with the analysis of the next chromosome, repeating these steps until all 22 chromosomes had been analyzed. Afterward, I envisioned employing a specific statistical method to consolidate the PRS data from these 22 chromosomes to determine the optimal overall threshold and subsequently summing them up. Do you think this idea is feasible?

— Reply to this email directly, view it on GitHub https://github.com/choishingwan/PRSice/issues/346#issuecomment-1784376365, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYSI4NQWOYL5MGLMFADYB4FVNAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBUGM3TMMZWGU . You are receiving this because you commented.Message ID: @.***>

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

[choishingwan]

Don't see any image.

What discrepancy are you seeing? We do expect different methods to give sulfurous different results

On Sat, Jan 27, 2024, 10:41 AM hxckk @.***> wrote:

Dear Sam,

I trust this message finds you well. Following your insightful guidance, I employed PRSice and LASSO methodologies to compute Polygenic Risk Scores (PRS). However, upon reviewing the results, a discrepancy has emerged, as illustrated in the attached image. I would appreciate your expertise in elucidating the potential reasons for these divergent outcomes.

Your assistance in interpreting these results would be highly valued.

-----原始邮件----- 发件人:"Shing Wan Choi" @.> 发送时间:2023-10-30 11:00:23 (星期一) 收件人: choishingwan/PRSice @.> 抄送: hxckk @.>, Author @.> 主题: Re: [choishingwan/PRSice] how to combine PRS from each chromosomes？ (Issue #346)

If storage is your concern, then this is completely unnecessary as the output of PRS software (specifically PRSice) scales with number of samples, not number of variant. And by using all chromosome at the same time, you can avoid generating the huge all-score file and only get the best score file, which should only contain 4 columns and should be much smaller in size than the all score file. It is also a good practice to keep all the genotype file in the same folder. But if for some reason that is not possible, then you can use the --target-list function instead, which allow a file containing the path to all genotype file as input. Hope this help

Sam

On Sun, Oct 29, 2023 at 10:24 PM hxckk @.***> wrote:

You can use --target chr# for example which PRSice will automatically substitute # with 1-22. So as long as your file naming is consistent, this should work … <#m-4242589899473809212> On Sun, Oct 29, 2023, 11:30 AM hxckk @.*> wrote: Oh right now I see it. If you use per chromosome, you can't use the best score, you need to use --no-regress, and then do the optimization by hand, as each chromosome might select a different threshold, but we really want one threshold for the whole genome. That's why I don't usually recommend doing c+t per chromosome, as it usually are too time consuming and io intensive for one to generate scores for 6000+ thresholds per chromosome Sam … <#m-184795729589143180> On Sat, Oct 28, 2023, 12:43 PM hxckk @.> wrote: Try it with just one threshold e.g. --fastscore --bar-levels 1, to see if there's problem with your code regarding summing the scores Sam … <#m5757006495489225859> On Fri, Oct 27, 2023 at 1:16 AM hxckk @.> wrote: Problem here is that the default is average score, which has a denominator of number of alleles included in the score (accounting for missingness). If you want to do it per chromosome, you can only use sum score (--score sum) Hi Mr. Shing Wan Choi, I attempted the task again, following the same code as previously, but I still cannot achieve results in (a) that match those in (b). What could be the issue? I am looking forward to your response. Thank you! — Reply to this email directly, view it on GitHub <#346 https://github.com/choishingwan/PRSice/issues/346 <#346 https://github.com/choishingwan/PRSice/issues/346> (comment) <#346 https://github.com/choishingwan/PRSice/issues/346 (comment) <#346 (comment) < https://github.com/choishingwan/PRSice/issues/346#issuecomment-1782307824

, or unsubscribe

https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ < https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ

https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ < https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ

https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ < https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ

https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ < https://github.com/notifications/unsubscribe-auth/AAJTRYVNKJH2KS5CA42ZKSDYBM7SLAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBSGMYDOOBSGQ

. You are receiving this because you commented.Message ID: @.> Hello Sam, Thank you very much for your response. I have tried using "--fastscore --bar-levels 1" for both analyses a) and b), and this time the results matched! This indicates that the code I used was correct. So, does this mean I should specify the same threshold for all chromosomes in all analyses? — Reply to this email directly, view it on GitHub <#346 https://github.com/choishingwan/PRSice/issues/346 (comment) <#346 (comment) < https://github.com/choishingwan/PRSice/issues/346#issuecomment-1783867029

, or unsubscribe

https://github.com/notifications/unsubscribe-auth/AAJTRYWJHPWCVUVNIDI6Z6TYBUY4RAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTHA3DOMBSHE < https://github.com/notifications/unsubscribe-auth/AAJTRYWJHPWCVUVNIDI6Z6TYBUY4RAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTHA3DOMBSHE

https://github.com/notifications/unsubscribe-auth/AAJTRYWJHPWCVUVNIDI6Z6TYBUY4RAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTHA3DOMBSHE < https://github.com/notifications/unsubscribe-auth/AAJTRYWJHPWCVUVNIDI6Z6TYBUY4RAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTHA3DOMBSHE

. You are receiving this because you commented.Message ID: @.**> Hello Sam, You are absolutely right. I tried using the "--no-regress" command, and it generated a ".all_score file" for each chromosome. However, I'm uncertain about how to proceed with the next step of the analysis. Additionally, I am planning to perform analyses using UKB imputation data, and due to storage limitations, I have to conduct PRS analysis on a per-chromosome basis. Since the C+T method is somewhat inconvenient for chromosome-specific analyses, do you have any recommendations for a more suitable method? Thank you for your assistance! — Reply to this email directly, view it on GitHub <#346 (comment) < https://github.com/choishingwan/PRSice/issues/346#issuecomment-1784144502

, or unsubscribe

https://github.com/notifications/unsubscribe-auth/AAJTRYR2KMKEN5NWTAOORX3YBZZARAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBUGE2DINJQGI < https://github.com/notifications/unsubscribe-auth/AAJTRYR2KMKEN5NWTAOORX3YBZZARAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBUGE2DINJQGI

. You are receiving this because you commented.Message ID: @.***>

Hello Sam,

I wanted to express that using "target chr#" did indeed work. However, it does require placing all 22 chromosomes in the same folder.

My original intention was to analyze one chromosome at a time in a folder, primarily due to limited hard drive storage space. Following the acquisition of PRS information for a specific chromosome, my plan was to remove the other files and then proceed with the analysis of the next chromosome, repeating these steps until all 22 chromosomes had been analyzed. Afterward, I envisioned employing a specific statistical method to consolidate the PRS data from these 22 chromosomes to determine the optimal overall threshold and subsequently summing them up. Do you think this idea is feasible?

— Reply to this email directly, view it on GitHub < https://github.com/choishingwan/PRSice/issues/346#issuecomment-1784376365 , or unsubscribe < https://github.com/notifications/unsubscribe-auth/AAJTRYSI4NQWOYL5MGLMFADYB4FVNAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBUGM3TMMZWGU

. You are receiving this because you commented.Message ID: @.***>

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

— Reply to this email directly, view it on GitHub https://github.com/choishingwan/PRSice/issues/346#issuecomment-1913189600, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYSIUY5MUYMHUUJI7ODYQUNZRAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSMJTGE4DSNRQGA . You are receiving this because you commented.Message ID: @.***>

[kk-hxc]

sorry, i've unloaded the picture again.

[choishingwan]

As mentioned, it is only normal for different methods to result in different PRS. That's the fundamental of it.

On Sat, Jan 27, 2024 at 9:02 PM kk-hxc @.***> wrote:

sorry, i've unloaded the picture again.

db2adc2bcce3f6281004bb99fc20e5f.png (view on web) https://github.com/choishingwan/PRSice/assets/143304884/1d6a9384-6888-4b7c-8736-838720344aa5

— Reply to this email directly, view it on GitHub https://github.com/choishingwan/PRSice/issues/346#issuecomment-1913411359, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJTRYXEU7EWSSB2SMEUSR3YQWWVBAVCNFSM6AAAAAA6PTVG72VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSMJTGQYTCMZVHE . You are receiving this because you commented.Message ID: @.***>