19 Disclosure filter configurations

DataSHIELD and OmicSHIELD have different filters that can be configured by each study server independently. This filters are automatically setted when installing the server packages (dsOmics and dsBase) and take the default values, which are designed to be sensible. This values can be configured individually through the web interface of Opal so that each cohort can establish their disclosure risk. First, we need to understand a little bit what is the function of each filter:

• For the dsBase filters, take a look at the official documentation
• For the OmicSHIELD filters:
  • default.nfilter.diffP.epsilon: \(\epsilon\)-privacy level for the differential-privacy (defaults to 3)
  • default.nfilter.diffP.resampleN: Number of resamples for the assessment of the l1-sensitivity of the differential-privacy. Largest is more safe but more slow (defaults to 3)
  • default.nfilter.MAF: Filter for the calculation of the minor allele frequencies (MAF), MAF values lower than this threshold will not be returned (defaults to 0.05)

To change this values, we have to login to the Opal web interface of our server, we can visit the Opal demo website as an example. We have to login as an administrator to change the filter settings, for the Opal demo: administrator/password are the credentials.

Once logged in, we navigate to Administration > DataSHIELD.

On this configuration page we have to locate the “Profiles” section (information about them here), there we will find the different analysis profiles of our server, we can set up different filter configurations for each of the profiles. Following the Opal demo example we will be modifying the filters for the “omics” profile, which has the dsOmics package installed. To do so, we select the profile and go to the “Options” tab (illustrated on the figure 19.1). There we can manually edit the filter values at our own risk.

Figure 19.1: DataSHIELD filters on the Opal web interface