# Protein Sequence Design with Batch Bayesian Optimisation

arXiv: https://arxiv.org/abs/2303.10429 · HTML: https://mlsanity.com/paper/2303.10429/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 4 of its 36 references that are in the map.
- First posted on arXiv 18 Mar 2023.

## Builds on (most cited first, 4 of 4)

- [Accelerating Bayesian Optimization for Biological Sequence Design with Denoising Autoencoders](https://mlsanity.com/paper/2203.12742/) (2022)
- [The Parallel Knowledge Gradient Method for Batch Bayesian Optimization](https://mlsanity.com/paper/1606.04414/) (2016)
- [AntBO: Towards Real-World Automated Antibody Design with Combinatorial Bayesian Optimisation](https://mlsanity.com/paper/2201.12570/) (2022)
- [Adaptive machine learning for protein engineering](https://mlsanity.com/paper/2106.05466/) (2021)

## Built on by (most cited first, 1 of 1)

- [Deep learning-guided evolutionary optimization for protein design](https://mlsanity.com/paper/2603.02753/) (2026)

## Abstract (arXiv)

Protein sequence design is a challenging problem in protein engineering, which aims to discover novel proteins with useful biological functions. Directed evolution is a widely-used approach for protein sequence design, which mimics the evolution cycle in a laboratory environment and conducts an iterative protocol. However, the burden of laboratory experiments can be reduced by using machine learning approaches to build a surrogate model of the protein landscape and conducting in-silico population selection through model-based fitness prediction. In this paper, we propose a new method based on Batch Bayesian Optimization (Batch BO), a well-established optimization method, for protein sequence design. By incorporating Batch BO into the directed evolution process, our method is able to make more informed decisions about which sequences to select for artificial evolution, leading to improved performance and faster convergence. We evaluate our method on a suite of in-silico protein sequence design tasks and demonstrate substantial improvement over baseline algorithms.