Zachary Wu
Overview
Zachary Wu is a research scientist at DeepMind, working on protein engineering, biomolecular modelling, and machine learning. He has been affiliated with the California Institute of Technology and has co-authored several papers on protein engineering and machine learning.
Notable Works
- Protein sequence design with deep generative models: This work focuses on protein engineering, aiming to identify protein sequences with optimised properties. By combining machine learning with experimental efforts, the process of protein sequence generation can be improved.
- Advances in machine learning for directed evolution: Wu and his colleagues discuss how machine learning can expedite directed evolution by allowing researchers to perform expensive experimental screens in silico. They also address the challenge of gathering sequence-function data for training ML models and suggest using widely available raw protein sequence data instead.
- Machine-learning-guided directed evolution for protein engineering: Wu and his co-authors introduce the steps required to build machine-learning sequence-function models and use them to guide engineering. They illustrate the underlying principles through case studies.
- Machine learning-assisted directed protein evolution with combinatorial libraries: The paper proposes incorporating machine learning with directed evolution to reduce the experimental effort and improve outcomes. It suggests that machine learning-guided directed evolution can find variants with higher fitness than other directed evolution approaches.
- Learned protein embeddings for machine learning: The authors compare the predictive power of Gaussian process models trained using embeddings with those trained on existing representations. They find that embeddings enable accurate predictions despite having significantly fewer dimensions.
- Stereoselective Enzymatic Synthesis of Heteroatom-Substituted Cyclopropanes: This work focuses on engineering variants of Cytochrome P450BM3 to catalyse the synthesis of nitrogen-, oxygen-, and sulfur-substituted cyclopropanes, expanding the catalytic functions of iron heme proteins.
Zachary Wu
Biography
Zachary Wu is a research scientist at DeepMind, working in protein engineering, biomolecular modelling, and machine learning. He was previously affiliated with the California Institute of Technology.
Research
Wu's research focuses on protein engineering and machine learning. He has published work on using machine learning to generate protein sequences, improve directed evolution, and optimise protein functions. He has also worked on signal peptides generated by attention-based neural networks and machine learning-assisted directed protein evolution with combinatorial libraries.
Publications
- Protein sequence design with deep generative models
- Advances in machine learning for directed evolution
- Signal Peptides Generated by Attention-Based Neural Networks
- Machine-learning-guided directed evolution for protein engineering
- Machine learning-assisted directed protein evolution with combinatorial libraries
- Machine Learning in Protein Engineering
- Learned protein embeddings for machine learning
- Stereoselective Enzymatic Synthesis of Heteroatom-Substituted Cyclopropanes
Zachary Wu
Biography
Zachary Wu is a research scientist at DeepMind, working in protein engineering, biomolecular modelling, and machine learning. He was previously affiliated with the California Institute of Technology.
Research
Wu's research focuses on protein engineering and machine learning. He has published work on using machine learning to generate protein sequences, and on machine learning-guided directed evolution for protein engineering. He has also published on advances in machine learning for directed evolution, and on using attention-based neural networks to generate signal peptides.
Publications
- Protein sequence design with deep generative models
- Advances in machine learning for directed evolution
- Signal Peptides Generated by Attention-Based Neural Networks
- Machine-learning-guided directed evolution for protein engineering
- Machine learning-assisted directed protein evolution with combinatorial libraries
- Learned protein embeddings for machine learning
- Stereoselective Enzymatic Synthesis of Heteroatom-Substituted Cyclopropanes