bed_autoSVD PC number calculation error?

[gradsmjin814]

obj.bed=bed(YOBP_OE_cox_bed)
|--------------------------------------------------|
|==================================================|
obj.svd <- runonce::save_run(

• bed_autoSVD(obj.bed, k=20, ncores=nb_cores(),min.maf=0.01, min.mac=10),
• file=paste0(SWE_OE, "_YOBP_OE_cox_tmp_PCA.rds")
• )
  user system elapsed
  1931.089 19.170 2921.727

PC<-predict(obj.svd); head(PC)
[,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8]
[1,] -15.339625 -0.6784555 4.219289 -4.806412 17.0022308 -3.276986 -9.4454624 1.886276
[2,] -18.129878 -5.1819836 1.729590 -10.008311 14.5715798 2.004858 -1.8473338 -3.185667
[3,] -8.836365 -0.3806917 8.346467 -5.900797 -4.0331445 -9.286730 2.1223432 7.873380
[4,] -8.545034 1.8428507 2.147583 -3.161611 -7.6413284 -9.879551 7.4082319 4.351154
[5,] -13.234423 -0.5214976 6.259730 -10.428578 5.5295740 -10.062408 -11.1693163 -6.209751
[6,] -13.033614 -4.3277959 8.582236 -9.232543 -0.7709548 7.052743 0.6112591 -5.495643
[,9] [,10]
[1,] -14.0802928 20.2251012
[2,] -11.4811502 2.1643478
[3,] -13.9776866 0.5716711
[4,] 0.7564992 -3.1917674
[5,] -4.8962562 -0.2635853
[6,] -12.6963200 -1.6274658

Not sure why it only gives me PC scores up to 10 when I ordered the SVD for 20?

[privefl]

Are you sure you didn't previously run this with k=10, and save_run() stored and is reusing that older result?

[gradsmjin814]

I did actually save it k=10 before, but should it not have k=20 when I rerun and save the rds file again? I mean I expect the result to overwrite the previously calculated singular value decomposition

…

On Thu, 26 Dec 2024, 07:43 Florian Privé, ***@***.***> wrote: Are you sure you didn't previously run this with k=10, and save_run() stored and is reusing that older result? — Reply to this email directly, view it on GitHub <#530 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A6VDUPVOMRBZDFE7F75H2X32HOQSRAVCNFSM6AAAAABUGKLKK2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDKNRSGIYTQMBZGE> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[privefl]

If the rds file already exists, it just uses it.
You'll have to remove the file yourself if you want to run some new code.

I can probably add a TIME attribute in next versions to give information on when this was computed and stored, if that helps.
And also maybe a warning if code has changed.