New sequence kernels improve protein property prediction

Researchers have developed a new class of sequence kernels for Gaussian processes to predict protein properties from limited experimental data. These kernels leverage evolutionary substitution matrices and local linearity, often outperforming models that rely on foundation model embeddings. The approach can also incorporate structural information from foundation models and is effective for multi-task learning across various protein property landscapes. AI

IMPACT Introduces a novel method for protein property prediction that may offer advantages over current foundation model approaches.

RANK_REASON The cluster contains an academic paper detailing a new method for protein property prediction. [lever_c_demoted from research: ic=1 ai=0.7]

Read on arXiv cs.LG →

• foundation model embeddings
• Gaussian processes

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →