PRSice threshold result is not as expected

Question

[landscape95]

Hi everyone,

I am doing the PRS with PRSice, I set as follow

...
--lower 0 \
--interval 0.01 \
--upper 0.85 \
...

the result in the .prsice is as follow

Set Threshold   R2  P   Coefficient Standard.Error  Num_SNP
Base    0.11    0.0573236   2.14869e-05 1.41723 0.323538    1
Base    0.16    0.0820505   2.23083e-07 2.28802 0.42246 2
Base    0.23    0.124923    3.67078e-11 3.3304  0.468058    3
Base    0.44    0.212734    5.69775e-21 6.80455 0.624526    10
Base    0.46    0.225492    1.00129e-22 7.28094 0.631047    12
Base    0.57    0.223171    2.12793e-22 7.39902 0.647995    13
Base    0.6 0.238986    1.0476e-24  8.42251 0.68701 21
Base    0.62    0.244352    1.56133e-25 9.09485 0.724074    27
Base    0.7 0.251666    1.06589e-26 9.4993  0.731542    31
Base    0.76    0.2574  1.20287e-27 10.3398 0.775632    44
Base    0.77    0.265801    4.31103e-29 12.4527 0.898506    116
Base    0.78    0.265607    4.6651e-29  12.4531 0.899346    117
Base    0.79    0.26775 1.9452e-29  12.616  0.902045    122
Base    0.81    0.268386    1.49709e-29 12.6483 0.901663    123
Base    0.82    0.268006    1.75044e-29 12.659  0.904032    124
Base    0.83    0.266628    3.07871e-29 12.6872 0.911902    131
Base    0.84    0.269101    1.11373e-29 12.9602 0.920808    146
Base    0.85    0.274744    1.0325e-30  13.4441 0.9301  171
Base    1   0.292076    3.96269e-34 118.58  7.54246 110070

I wonder why even though I have set the parameters as above the threshold jump is not as expected 0 - 0.01 - 0.02 - ... - 0.85. Does any know what happens with it?

Your help is really appreciated!

Answer 1

[Sam]

That is because we tried to optimize the speed of PRSice by ignoring thresholds that did not add any additional SNPs from previous thresholds.

For example, if my SNPs are SNP P-value SNP1 0.04 SNP2 0.045 SNP3 0.1 SNP4 0.2

Then with your threshold selection (0 - 0.85, step size of 0.01), we will only generate results for threshold

0.04, 0.05, 0.1 and 0.2, ignoring all thresholds in-between. You can savely duplicate the results for those other threshold if you want (e.g. results for threshold 0.05 to 0.09 should be identical as no new SNPs are added)