Bioinformatics Seminar
Time: 11AM
Venue: Davis Auditorium and Online
15 April 2025
Machine-learning guided directed protein evolution
Daniel BrownWEHI Advanced Technology and Biology
The advent of deep learning has revolutionized approaches in protein design and optimization. In 2024, we explored de novo design of peptide binders. Building upon that work, we have since transitioned to protein sequence optimization, critical for progressing from hit sequences to lead biologics; a process traditionally reliant on slow, costly random mutagenesis or rational engineering. We are implementing active learning methods to accelerate this workflow, integrating design build test learn cycles with wet and dry lab feedback. I will present a general overview of active learning methodology and introduce some biological applications. Active learning has broad applications to hyperparameter optimisation, drug discovery, perturbation screens and cellular reprogramming.