Scientists have unveiled ProtET, a groundbreaking AI-driven tool that translates text-based instructions into precise protein modifications, offering a transformative leap for genetic therapies, enzyme design, and biopharmaceutical innovation.
Research: Multi-Modal CLIP-Informed Protein Editing. Image Credit: S. Singha / Shutterstock
Researchers from Zhejiang University and HKUST (Guangzhou) have developed a cutting-edge AI model, ProtET, that leverages CLIP-informed multi-modal learning, combining protein sequences and natural language descriptions, to enable controllable protein editing through text-based instructions. This innovative approach, published in the journal Health Data Science, bridges the gap between biological language and protein sequence manipulation, enhancing functional protein design across domains like enzyme activity, stability, and antibody binding.
Proteins are the cornerstone of biological functions, and their precise modification holds immense potential for medical therapies, synthetic biology, and biotechnology. While traditional protein editing methods rely on labor-intensive laboratory experiments and single-task optimization models, ProtET introduces a transformer-structured encoder architecture, which integrates two advanced language models—ESM-2 for protein sequences and PubMedBERT for biological text—to improve protein editing accuracy. Using contrastive learning, this model aligns protein sequences with natural language descriptions, enabling intuitive, text-guided protein modifications.
The research team, led by Mingze Yin from Zhejiang University and Jintai Chen from HKUST (Guangzhou), trained ProtET on a dataset of over 67 million protein–biotext pairs, extracted from Swiss-Prot and TrEMBL databases. The model demonstrated exceptional performance across key benchmarks, improving protein stability by up to 16.9% and optimizing catalytic activities and antibody-specific binding.
"ProtET introduces a flexible, controllable approach to protein editing, allowing researchers to fine-tune biological functions with unparalleled precision," said Mingze Yin, the study's lead author.
The workflow and framework details of ProtET. (A) A CLIP-like contrastive pretraining aligns features of protein sequences and biotext descriptions. (B) FiLM module and transformer decoders for protein editing. The FiLM module integrates multi-modal features from the original protein sequences and the editing instruction texts, serving as the editing condition. Based on this condition, transformer-decoders design edited protein sequences through an autoregressive generation process. (C) Details of the FiLM module. It extracts multiplicative and additive factors from text features using linear mappings. These factors conditionally optimize protein features through addition and multiplication to create fused features. (D) Details of the transformer decoder. It uses a multi-head self-attention module to learn comprehensive residue-residue interactions and predicts the next residue based on the previous ones.
The model successfully optimized protein sequences across different experimental scenarios, including enzyme catalytic activity, protein stability, and antibody-antigen interaction binding. In zero-shot tasks—where the model optimizes antibody sequences without additional fine-tuning on specific antibody data—ProtET designed SARS-CoV antibodies that formed stable and functional 3D structures, demonstrating its real-world applicability in biomedical research.
The team also conducted an ablation study to validate the importance of key components, showing that multi-modal pretraining and the Feature-wise Linear Modulation (FiLM) module were essential for improving editing accuracy. However, the researchers acknowledged some limitations, including challenges in precisely interpreting natural language instructions and controlling sequence length in the generated proteins, which they plan to address in future work.
Looking ahead, the team envisions ProtET becoming a standard tool in protein engineering, paving the way for breakthroughs in synthetic biology, genetic therapies, and biopharmaceutical manufacturing.
This study marks a transformative step in AI-driven protein design, showcasing how cross-modal integration can unlock new horizons in scientific discovery and innovation.
Source:
Journal reference:
- Mingze Yin, Hanjing Zhou, Yiheng Zhu, Miao Lin, Yixuan Wu, Jialu Wu, Hongxia Xu, Chang-Yu Hsieh, Tingjun Hou, Jintai Chen, et al. Multi-Modal CLIP-Informed Protein Editing. Health Data Sci. 2024;4:0211. DOI:10.34133/hds.0211, https://spj.science.org/doi/10.34133/hds.0211