Active substances calculationΒΆ

Depending on the model settings, the set of active substances for a specified effect can be computed in several ways:

  1. From the list of substances with available points of departure (POD) data for the specified effect. If there is a POD, then the substance is considered an active substance, with membership 1. If not, the membership is 0, and the substance is excluded from the list of active substances.

  2. From one or more in-silico (QSAR and/or molecular docking) models. The results of the in-silico models should be provided as QSAR membership models data and/or molecular docking models data. Binding energies from molecular docking models are first translated to crips memberships using a threshold value. The results from multiple in-silico models can be combined in any of four membership calculation methods:

    1. (crisp, any) the substance is considered an active substance if any in-silico model indicates activity;

    2. (crisp, majority) the substance is considered an active substance if the majority of in-silico models indicates activity;

    3. (probabilistic, ratio) the membership probability is the fraction of in-silico models that indicate activity;

    4. (probabilistic, Bayesian) the membership probability is calculated using a Bayesian model according to Kennedy et al. [Kennedy et al., 2020] and a specified prior probability (which is by default 0.5).

For substances within the scope of the assessment but without in-silico data, the default is to omit them in the AG. There is an option however to include such substances with a default membership probability.

  1. From a combination of 1 and 2, using either the union (OR) method or the intersection (AND method) of results.