CRISPR-GPT

LLM Agents for Automated Design of Gene-Editing Experiments

Authors

Yuanhao Qu^1,&, Kaixuan Huang^2,&, Ming Yin², Kanghong Zhan³, Dyllan Liu⁴, Di Yin¹, Henry C. Cousins^5,6, William A. Johnson¹, Xiaotong Wang¹, Russ B. Altman^4,7, Denny Zhou⁸, Mengdi Wang^2,*, Le Cong^1,*

¹Department of Pathology, Department of Genetics, Cancer Biology Program, Stanford University School of Medicine, Stanford, CA 94305, USA

²Center for Statistics and Machine Learning, Department of Electrical and Computer Engineering, Princeton University, Princeton, NJ 08544, USA

³Department of Computing, Data Science, and Society, University of California, Berkeley, Berkeley, CA 94720, USA

⁴Department of Computer Science, University of California, Berkeley, Berkeley, CA 94720, USA

⁵Department of Medicine, Stanford University School of Medicine, Stanford, CA 94305, USA

⁶Medical Scientist Training Program, Stanford University School of Medicine, Stanford, CA 94305, USA

⁷Department of Bioengineering, Department of Genetics, Stanford University, Stanford, CA 94305, USA

⁸Google DeepMind, Mountain View, CA 94043, USA

^&These authors contributed equally

^*Corresponding authors: mengdiw@princeton.edu (M.W.), congle@stanford.edu (L.C.)

Abstract

Genome engineering technology has revolutionized biomedical research by enabling precise genetic modifications. However, designing effective gene-editing experiments requires a deep understanding of both the CRISPR technology and the biological system involved. In this work, we present CRISPR-GPT, an LLM agent system to automate and enhance the CRISPR-based gene-editing design process. CRISPR-GPT leverages the reasoning capabilities of LLMs for complex task decomposition, decision-making, and interactive human-AI collaboration. This system is driven by multi-agent collaboration, and it incorporates domain expertise, retrieval techniques, external tools, and a specialized LLM fine-tuned with a decade's worth of open-forum discussions among gene-editing scientists. CRISPR-GPT assists users in selecting CRISPR systems, experiment planning, designing gRNAs, choosing delivery methods, drafting protocols, designing assays, and analyzing data. We showcase the potential of CRISPR-GPT in assisting beginner researchers with gene-editing from scratch, knocking-out four genes with CRISPR-Cas12a in a human lung adenocarcinoma cell line and epigenetically activating two genes using CRISPR-dCas9 in human melanoma cell line, both successful on first attempt.

How CRISPR-GPT Works

CRISPR-GPT leverages the reasoning capabilities of LLMs for complex task decomposition, decision-making, and interactive human-AI collaboration

Multi-Agent Collaboration

Intelligent task decomposition

CRISPR-GPT uses a team of specialized agents including an LLM Planner, Task Executor, and User-Proxy to break down complex gene-editing tasks into manageable steps.

Domain Expertise

Specialized knowledge integration

The system incorporates domain expertise from leading practitioners and peer-reviewed literature, enhanced by CRISPR-Llama3, a specialized LLM fine-tuned on expert discussions.

Integrated Tools

Comprehensive bioinformatics suite

CRISPR-GPT integrates search capabilities and bioinformatics tools like Primer3, CRISPRitz, and CRISPresso2 for comprehensive gene-editing design and analysis.

Flexible Interaction

Multiple operation modes

The system offers three interaction modes: Meta Mode for step-by-step guidance, Auto Mode for customized workflows, and QA Mode for on-demand scientific inquiries.

Key Features

CRISPR-GPT combines the strengths of LLMs with domain-specific knowledge, chain of thought reasoning, and specialized tools

Optimal Guide RNA Design

Maximize editing efficiency and specificity

Design highly specific guide RNAs with minimal off-target effects, optimized for your experimental system and target gene.

Complete Protocol Generation

Step-by-step experimental workflows

Generate detailed, ready-to-execute protocols for cloning, transfection, validation, and analysis tailored to your specific experimental goals.

Validation Strategy Design

Comprehensive quality control

Design robust validation strategies including sequencing primers, PCR assays, and functional tests to confirm successful editing.

See CRISPR-GPT in Action

Watch our demo showcasing CRISPR-GPT's capabilities for automated gene-editing experiment design

Watch Demo

CRISPR-GPT: LLM Agents for Automated Design of Gene-Editing Experiments

See how CRISPR-GPT guides researchers through the complete gene-editing workflow, from CRISPR system selection and guide RNA design to protocol generation and validation.

Real-World Applications

CRISPR-GPT has been validated in real laboratory settings, enabling successful gene editing experiments by researchers with minimal prior experience

Multi-Gene Knockout

CRISPR-Cas12a knockout of four genes in human lung adenocarcinoma cells

A junior researcher with no prior gene-editing experience successfully knocked out TGFBR1, SNAI1, BAX, and BCL2L1 genes in A549 cells with ~80% efficiency using CRISPR-GPT guidance.

The system provided complete guidance from CRISPR system selection and gRNA design to protocol generation and validation strategy, resulting in successful editing confirmed by next-generation sequencing.

Epigenetic Activation

CRISPR-dCas9 activation of two genes in human melanoma cells

An undergraduate student with no prior CRISPR experience successfully activated NCR3LG1 and CEACAM1 genes in A375 melanoma cells with up to 90% efficiency using CRISPR-GPT.

The AI agent guided the entire process from system selection to validation, demonstrating the potential of AI-guided gene editing to accelerate research and make advanced techniques accessible to beginners.