19 months ago by
EMBL Heidelberg, Germany
It is common for experiments to be inconclusive but this being said, there are still a few things to check before giving up on the data. First have a look at the histogram of the original p-values and read this blog on interpreting p-value histograms. If you're in a pathological case, you may want to seek help in figuring out if there's anything that can be done to still use the data. Assuming you're not in a pathological case, then you can apply the FDR correction. If all the corrected p-values are close to 1, it means that for any threshold on the false positive rate you can't reject the null hypothesis. However, statistical significance is not related to biological relevance. You don't say what you're testing but you should also have a look at the magnitude of the effect you're measuring for example by ranking the genes by strength of the effect or plotting a histogram of it. If all you need are a few genes for follow up experiments, that may give you indications on what to do next.
Another common thing to do is to filter the data based on some criterion independent of what's tested (e.g. removing genes associated with technical artefacts or genes not expressed in the system under consideration).