Closed kimsjune closed 1 year ago
Dear kimsjune, the detection p-values on Infinium arrays are in fact "p-values", i.e. they are not p-values in a strict sense, but represent one minus quantile of the negative probe signal distribution into which the actual probe intensity falls (actually a summary of two such quantiles, for methylated and unmethylated probes, respectively). Since there are only around 600 negative control probes, it is easily possible that they are exactly zero. Furthermore, this means that the smallest non-negative detection p-value would be between 0.001 and 0.002. Hope this helps.
lutsik,
Thank you for your swift response. This clarifies a lot.
Best, SK
As demonstrated on page 12 of the manual, dpval() just returns zeros. Does this mean that the p-values are too small to be calculated? summary(dpval(rnb.set.example)) shows that nearly all probes across nearly all samples are just zero. Do these zeroes represent 0.01, 0.0001 or 0.0000000000001? It's difficult to tell. My guess is that floating points just get lost somewhere or not reported fully. P-values should be represented in scientific notation.