about cutoff for HVP selection?

[xflicsu]

Dear MAnorm2 developer,
We get result of HVPs by HyperChIP of MAnorm2 as follow.
We wonder what cutoff is better for HVP selection?
Two questions:

1. padj<0.05 is one basic cutoff?
2. What is the role of "observed.var" and "fold.change" and how can we select them?

#> observed.mean observed.var prior.var fold.change pval padj
#> 1 4.123172 0.62767325 3.3719506 0.18614545 0.108561188 0.3391233
#> 2 8.835486 0.01362584 0.2844984 0.04789425 0.008610023 0.1013005
#> 3 5.794082 0.16260797 1.4398955 0.11293040 0.043945034 0.2164067
#> 4 4.580327 4.82749982 2.6687559 1.80889526 0.247900708 0.5171521
#> 5 8.302970 0.07726829 0.3872625 0.19952434 0.122594081 0.3598123
#> 6 4.440024 1.65003664 2.8669925 0.57552876 0.639231994 0.8224108

[tushiqi]

@xflicsu
The observed.var variable refers to the sample variance calculated from the normalized signal intensities associated with each peak region, and the fold.change variable refers to the ratio of observed.var to prior.var, which is read from the fitted mean-variance curve.

Yes, you are right. To select HVPs you first need to set a cutoff for padj (both 0.1 and 0.05 are frequently used). Then, you should further impose the restriction fold.change>1 to filter out the LVPs (lowly variable peaks).

Please feel free to contact me if you have further questions.