A Text-guided Protein Design Framework

TL;DR

This paper introduces ProteinDT, a novel multi-modal framework that incorporates textual descriptions to improve protein design, achieving high accuracy and zero-shot editing capabilities, and enhancing property prediction performance.

Contribution

The paper presents ProteinDT, the first framework to leverage text descriptions for protein design, integrating a large dataset and demonstrating superior results across multiple tasks.

Findings

Over 90% accuracy in text-guided protein generation

Best hit ratio in zero-shot protein editing tasks

Outperforms existing methods on property prediction benchmarks

Abstract

Current AI-assisted protein design mainly utilizes protein sequential and structural information. Meanwhile, there exists tremendous knowledge curated by humans in the text format describing proteins' high-level functionalities. Yet, whether the incorporation of such text data can help protein design tasks has not been explored. To bridge this gap, we propose ProteinDT, a multi-modal framework that leverages textual descriptions for protein design. ProteinDT consists of three subsequent steps: ProteinCLAP which aligns the representation of two modalities, a facilitator that generates the protein representation from the text modality, and a decoder that creates the protein sequences from the representation. To train ProteinDT, we construct a large dataset, SwissProtCLAP, with 441K text and protein pairs. We quantitatively verify the effectiveness of ProteinDT on three challenging tasks: (1) over 90% accuracy for text-guided protein generation; (2) best hit ratio on 12 zero-shot text-guided protein editing tasks; (3) superior performance on four out of six protein property prediction benchmarks.