[Paper Review] CRISPR-GPT for Agentic Automation of Gene-editing Experiments

Yuanhao Qu, Kaixuan Huang+12 more|arXiv (Cornell University)|Apr 27, 2024

CRISPR and Genetic Engineering9 citations

TL;DR

CRISPR-GPT is an LLM-based agent augmented with domain knowledge and tools to automate CRISPR gene-editing experimental design, supporting meta, auto, and Q&A modes with real-world validation and ethical safeguards.

ABSTRACT

The introduction of genome engineering technology has transformed biomedical research, making it possible to make precise changes to genetic information. However, creating an efficient gene-editing system requires a deep understanding of CRISPR technology, and the complex experimental systems under investigation. While Large Language Models (LLMs) have shown promise in various tasks, they often lack specific knowledge and struggle to accurately solve biological design problems. In this work, we introduce CRISPR-GPT, an LLM agent augmented with domain knowledge and external tools to automate and enhance the design process of CRISPR-based gene-editing experiments. CRISPR-GPT leverages the reasoning ability of LLMs to facilitate the process of selecting CRISPR systems, designing guide RNAs, recommending cellular delivery methods, drafting protocols, and designing validation experiments to confirm editing outcomes. We showcase the potential of CRISPR-GPT for assisting non-expert researchers with gene-editing experiments from scratch and validate the agent's effectiveness in a real-world use case. Furthermore, we explore the ethical and regulatory considerations associated with automated gene-editing design, highlighting the need for responsible and transparent use of these tools. Our work aims to bridge the gap between beginner biological researchers and CRISPR genome engineering techniques, and demonstrate the potential of LLM agents in facilitating complex biological discovery tasks. The published version of this draft is available at https://www.nature.com/articles/s41551-025-01463-z.

Motivation & Objective

• Lower the barrier for novices to design CRISPR experiments by providing an automated design assistant.
• Integrate domain knowledge, external tools, and structured task execution to design CRISPR experiments.
• Evaluate CRISPR-GPT against general LLMs and validate with a real wet-lab use case.
• Address ethical and regulatory considerations in automated gene-editing design.

Proposed method

• LLM-powered design and planning engine with four core modules: LLM planner, Tool provider, Task executor, and LLM Agent.
• State-machine-based Task Executor decomposes tasks into 22 sub-goals across 4 predefined meta-tasks (meta-tasks) and can chain tasks into a complete pipeline.
• Tool Provider wraps external APIs and tools (e.g., Google web search, Primer3, gRNA libraries) for LLM access via user-friendly prompts.
• Use of ReAct-style prompting to enable chain-of-thought reasoning and plausible action selection.
• Meta Mode, Auto Mode, and Q&A Mode to guide design workflows and interactive querying.
• Incorporation of safeguards and ethical considerations, including restrictions on human use and data privacy protections.

Experimental results

Research questions

• RQ1 How does CRISPR-GPT perform in designing CRISPR experiments compared to general LLMs like ChatGPT 3.5 and 4.0?
• RQ2 Can CRISPR-GPT assist non-experts in creating complete, accurate, and concise gene-editing design plans?
• RQ3 What is the impact of integrating domain knowledge and external tools on design quality and completeness?
• RQ4 What are the safety and ethical considerations and how can they be mitigated in automated gene-editing design?

Key findings

• CRISPR-GPT achieves significantly higher accuracy than general LLMs in gene-editing design tasks across Meta, Auto, and Q&A modes.
• CRISPR-GPT demonstrates better completeness and concise answers compared to general LLMs, particularly in Q&A mode.
• Human experts rated CRISPR-GPT higher in accuracy, reasoning, completeness, and conciseness than ChatGPT-3.5 and showed competitive or superior reasoning across modes.
• A real-world wet-lab demonstration in a human A375 cell line showed CRISPR-GPT guiding a multi-gene knockout workflow with design of gRNAs, delivery, cloning protocols, and validation steps.
• The study documents ethical and safety safeguards and discusses limitations and potential extensions for more complex or rare biological cases.