liveBook · Manning

Expected improvement

As explained in Build AI Drug Discovery Pipelines

Expected improvement is an acquisition criterion that estimates the expected amount by which a candidate will improve on the best observed result.

Expected improvement (EI) is an acquisition function that estimates how much a candidate could improve on the best observed result. section 9.2.5

In virtual screening, EI helps balance exploration of uncertain chemical space with exploitation of molecules predicted to perform well. It works within an active-learning loop that uses a surrogate model to prioritize compounds for expensive evaluation. section 9.2.5 section 9.4

Prioritizing docking experiments

A neural network surrogate predicts binding affinities from molecular fingerprints while EI helps select compounds for full docking. In a VEGFR-2 kinase experiment using target PDB ID 4AG8, active learning recovered 73 of the top 100 molecules after 10 iterations. section 9.4

How Build AI Drug Discovery Pipelines covers it

How deep it goes
Defines EI as an acquisition function but defers its implementation to a Python notebook rather than showing its code here.
Examples
Uses VEGFR-2 kinase screening, PDB ID 4AG8, molecular fingerprints, docking, and recovery of 73 top-100 molecules after 10 iterations.
What you can do after
Apply it via an active-learning acquisition strategy to prioritize promising compounds for expensive docking or related calculations.
How it connects
Connects expected improvement with active learning, surrogate models, molecular fingerprints, virtual screening, and exploration-exploitation trade-offs.

Also explained in

Compare how 3 books explain expected improvement →

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.