liveBook · Manning

Ensemble consensus

As explained in Build AI Drug Discovery Pipelines

Ensemble consensus is the degree to which the individual models in an ensemble produce mutually agreeing predictions for an input.

Ensemble consensus measures how closely the individual models in an ensemble agree on a prediction. In a Random Forest, stronger internal agreement provides a cheap uncertainty signal for deciding which drug-discovery predictions to trust. section 5.3.2

Use consensus as a practical confidence check:

  • Train a diverse collection of decision trees with bagging and feature subsampling. section 5.3.2

CYP3A4 inhibition

A Random Forest trained on molecular fingerprints predicts cytochrome P450 inhibition. Applying a confidence threshold of 0.8 separates predictions considered inside the applicability domain from those outside it.

How Build AI Drug Discovery Pipelines covers it

How deep it goes
Connects consensus to Random Forest diversity and applicability-domain filtering, but does not detail a standalone agreement formula.
Examples
Demonstrates consensus with CYP450 inhibition classification, molecular fingerprints, confidence masks, and in-domain versus out-of-domain PR AUC.
What you can do after
Diagnose prediction reliability by separating confident applicability-domain cases from less trustworthy cases.
How it connects
Ties ensemble consensus to Random Forests, decision trees, uncertainty estimation, and applicability domains.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Build AI Drug Discovery Pipelines. The section links point to the parts of the book it draws on.