ProteinEngine: Empower LLM with Domain Knowledge for Protein Engineering

Yiqing Shen,Outongyi Lv,Houying Zhu,Yu Guang Wang

2024-04-21

Abstract:Large language models (LLMs) have garnered considerable attention for their proficiency in tackling intricate tasks, particularly leveraging their capacities for zero-shot and in-context learning. However, their utility has been predominantly restricted to general tasks due to an absence of domain-specific knowledge. This constraint becomes particularly pertinent in the realm of protein engineering, where specialized expertise is required for tasks such as protein function prediction, protein evolution analysis, and protein design, with a level of specialization that existing LLMs cannot furnish. In response to this challenge, we introduce \textsc{ProteinEngine}, a human-centered platform aimed at amplifying the capabilities of LLMs in protein engineering by seamlessly integrating a comprehensive range of relevant tools, packages, and software via API calls. Uniquely, \textsc{ProteinEngine} assigns three distinct roles to LLMs, facilitating efficient task delegation, specialized task resolution, and effective communication of results. This design fosters high extensibility and promotes the smooth incorporation of new algorithms, models, and features for future development. Extensive user studies, involving participants from both the AI and protein engineering communities across academia and industry, consistently validate the superiority of \textsc{ProteinEngine} in augmenting the reliability and precision of deep learning in protein engineering tasks. Consequently, our findings highlight the potential of \textsc{ProteinEngine} to bride the disconnected tools for future research in the protein engineering domain.

Biomolecules,Artificial Intelligence,Machine Learning

What problem does this paper attempt to address?

This paper introduces a platform called ProteinEngine aimed at addressing the limitations in the application of large language models (LLMs) in the field of protein engineering. Despite performing well in complex tasks, LLMs lack domain-specific expertise to effectively handle protein function prediction, protein evolution analysis, and protein design, which require specialized skills. ProteinEngine enhances the capabilities of LLMs in protein engineering by integrating various relevant tools, packages, and software through API calls. It categorizes LLMs into three roles: AI Project Manager, AI Domain Expert, and AI Presenter, responsible for task coordination, specialized task solving, and result communication, respectively, thus improving the scalability and flexibility of the system.

Through user studies, ProteinEngine has demonstrated improved reliability and accuracy in protein engineering tasks, as well as reduced burden and learning difficulty for users from different backgrounds. The paper mentions various AI models and APIs related to protein engineering, such as AlphaFold for protein structure prediction and various interaction prediction tools. User studies show that ProteinEngine outperforms traditional methods in terms of task completion time, accuracy, system usability, and workload but falls short in significantly improving the workload.

In conclusion, the problem addressed in this paper is how to leverage the domain-specific knowledge of LLMs to enhance the efficiency and accuracy of protein engineering tasks, and how to achieve this goal through the ProteinEngine platform.