Flexible Kernels for Protein Property Prediction
The paper introduces a new class of sequence kernels that use evolutionary substitution matrices and local linearity to predict protein properties from sparse data.
The paper introduces a new class of sequence kernels that use evolutionary substitution matrices and local linearity, which is novel compared to prior work that relies on foundation model embeddings.
Keywords
Before reading this…
Applications
- →Protein design
To understand this paper, make sure you know these concepts first:
- Protein propertiesfind papers →
- Sequence kernelsfind papers →
- Gaussian processesfind papers →
Abstract
More Like ThisDespite its importance to applications in protein design, predicting protein properties like binding affinity and thermostability from sparse experimental data remains a significant challenge. Accordingly, we introduce a class of sequence kernels that exploit evolutionary substitution matrices as well as local linearity and demonstrate that the resulting Gaussian processes provide data-efficient models of protein property landscapes, frequently outperforming alternatives that rely on foundation model embeddings. Furthermore--by learning what are in effect structure-aware substitution matrices--we show that our kernels can readily incorporate structural information from foundation models. We demonstrate that these structure-conditioned kernels are well suited to multi-task learning across multiple protein property landscapes and can decisively outperform local supervised learning methods.