AlexTISYoung
commented
1 year ago Hi Sam,
Thanks for the email. I haven't noticed this issue before. My guess is that there might be some issue in the way snipar is reading your pedigree file/how your pedigree file is specified.
Your results look to me like the maternal PGI is not matched properly to the proband and paternal PGI since the coefficient on the maternal PGI is not statistically significant. I would suggest checking the correlation between proband, paternal, and maternal PGIs as output in the pgs.txt file. The correlation of the proband and maternal PGI should be at least 0.5, and if it isn't, then something has gone wrong in matching the mothers' genotypes to the right family.
Alex.
On Thu, 31 Aug 2023 at 14:28, Sam Ghatan @.***> wrote:
Hi there,
First of all, I'd like to express my appreciation for creating the snipar tool – it's been incredibly valuable in my research.
However, I've encountered some discrepancies when using snipar to calculate PGIs using the pgs.py script. Specifically, I've noticed discrepancies in the calculated scores and direct/indirect effects when working with non-imputed complete trios compared to results from PLINK and PRSice2 using the same input files.
When using snipar's pgs.py script to compute scores and effects for fathers and mothers within complete trios, I've observed unusual results, particularly with fathers having significantly larger indirect effects compared to mothers. In contrast, when constructing PGIs using PLINK and PRSice2 with identical input files, the resulting scores and direct/indirect effects seem to align more sensibly.
I am currently looking into solutions but was hoping you could point me in the right direction. Is it expected to observe such variations between snipar and PLINK/PRSice2?
I appreciate the help.
Below is an excerpt of the code I've been using as well as the results.
/opt/software/versions/Python-3.8.3/bin/python3.8 ~/bin/pgs.py ./PGS/nonImp_direct --bed /home/r055442/GenR_trio_duos/European/Unphased_plink/EU_trio_duos_qc --weights TB_snps_GWS.txt --pedigree king_pedigree.txt --SNP SNP --beta_col b --A1 A1 --A2 A2 --no_am_adj
snipar results:
intercept -0.3184900915503549 0.4213717452056926 AGECHILD13 0.6187514297850482 0.065049096367718 GENDER 0.5119690037751246 0.0456294872363309 PC1 58.295617419578875 89.89010678220191 PC2 14.199614550388121 16.506109847826696 PC3 -11.658331730011618 16.342479733621758 PC4 -3.4131219226976306 10.229518152133148 PC5 6.125962910592555 9.75580559131821 PC6 2.0867224608324513 14.909449205213575 PC7 -1.9778547881605764 4.020370424302356 PC8 1.0566154135484203 3.0492383101905407 PC9 -0.10356116267358671 2.0601000784087495 PC10 4.670396739910904 7.887381787071474 proband 0.06989054168799796 0.03345105140816477 paternal 0.08266941781371662 0.02959574046976095 maternal -0.019148471231257614 0.029618742638294098
PRSice/Plink scores results:
Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) -11.78773 13.95345 -0.845 0.3984 proband 0.04540 0.03235 1.403 0.1607 father 0.06621 0.02881 2.298 0.0217 mother 0.05871 0.02868 2.047 0.0408 I(AGECHILD13^2) 0.00832 0.07041 0.118 0.9059 AGECHILD13 0.66761 1.96951 0.339 0.7347 GENDER 3.01988 1.94918 1.549 0.1215 PC1 88.17446 94.16606 0.936 0.3492 PC2 31.25962 18.75293 1.667 0.0957 . PC3 -13.27491 17.16911 -0.773 0.4395 PC4 -2.44231 12.83195 -0.190 0.8491 PC5 2.60604 11.24240 0.232 0.8167 PC6 -2.90147 15.41586 -0.188 0.8507 PC7 4.32752 5.38331 0.804 0.4216 PC8 4.18743 3.97015 1.055 0.2917 PC9 -1.04109 2.17161 -0.479 0.6317 PC10 2.63566 8.00310 0.329 0.7419 AGECHILD13:GENDER -0.18463 0.14344 -1.287 0.1982
— Reply to this email directly, view it on GitHub https://github.com/AlexTISYoung/snipar/issues/31, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABQQS6OV54YF7R7FBRAC7ZLXYB7PLANCNFSM6AAAAAA4F72X3M . You are receiving this because you are subscribed to this thread.Message ID: @.***>
PandaPowell
Hi there,
First of all, I'd like to express my appreciation for creating the snipar tool – it's been incredibly valuable in my research.
However, I've encountered some discrepancies when using snipar to calculate PGIs using the pgs.py script. Specifically, I've noticed discrepancies in the calculated scores and direct/indirect effects when working with non-imputed complete trios compared to results from PLINK and PRSice2 using the same input files.
When using snipar's pgs.py script to compute scores and effects for fathers and mothers within complete trios, I've observed unusual results, particularly with fathers having significantly larger indirect effects compared to mothers. In contrast, when constructing PGIs using PLINK and PRSice2 with identical input files, the resulting scores and direct/indirect effects seem to align more sensibly.
I am currently looking into solutions but was hoping you could point me in the right direction. Is it expected to observe such variations between snipar and PLINK/PRSice2?
I appreciate the help.
Below is an excerpt of the code I've been using as well as the results.
/opt/software/versions/Python-3.8.3/bin/python3.8 ~/bin/pgs.py ./PGS/nonImp_direct \ --bed /home/r055442/GenR_trio_duos/European/Unphased_plink/EU_trio_duos_qc \ --weights TB_snps_GWS.txt \ --pedigree king_pedigree.txt \ --SNP SNP \ --beta_col b \ --A1 A1 \ --A2 A2 \ --no_am_adj
snipar results:
intercept -0.3184900915503549 0.4213717452056926 AGECHILD13 0.6187514297850482 0.065049096367718 GENDER 0.5119690037751246 0.0456294872363309 PC1 58.295617419578875 89.89010678220191 PC2 14.199614550388121 16.506109847826696 PC3 -11.658331730011618 16.342479733621758 PC4 -3.4131219226976306 10.229518152133148 PC5 6.125962910592555 9.75580559131821 PC6 2.0867224608324513 14.909449205213575 PC7 -1.9778547881605764 4.020370424302356 PC8 1.0566154135484203 3.0492383101905407 PC9 -0.10356116267358671 2.0601000784087495 PC10 4.670396739910904 7.887381787071474 proband 0.06989054168799796 0.03345105140816477 paternal 0.08266941781371662 0.02959574046976095 maternal -0.019148471231257614 0.029618742638294098
PRSice/Plink scores results:
Coefficients: Estimate Std. Error t value Pr(>|t|)
(Intercept) -11.78773 13.95345 -0.845 0.3984
proband 0.04540 0.03235 1.403 0.1607
father 0.06621 0.02881 2.298 0.0217 mother 0.05871 0.02868 2.047 0.0408 I(AGECHILD13^2) 0.00832 0.07041 0.118 0.9059
AGECHILD13 0.66761 1.96951 0.339 0.7347
GENDER 3.01988 1.94918 1.549 0.1215
PC1 88.17446 94.16606 0.936 0.3492
PC2 31.25962 18.75293 1.667 0.0957 . PC3 -13.27491 17.16911 -0.773 0.4395
PC4 -2.44231 12.83195 -0.190 0.8491
PC5 2.60604 11.24240 0.232 0.8167
PC6 -2.90147 15.41586 -0.188 0.8507
PC7 4.32752 5.38331 0.804 0.4216
PC8 4.18743 3.97015 1.055 0.2917
PC9 -1.04109 2.17161 -0.479 0.6317
PC10 2.63566 8.00310 0.329 0.7419
AGECHILD13:GENDER -0.18463 0.14344 -1.287 0.1982