Discovery of CRISPR-Cas12a clades using a large language model

Area of Science:

• Genomics
• Molecular Biology
• Bioinformatics

Background:

• CRISPR-Cas systems are powerful tools in life sciences, but metagenomes harbor vast numbers of uncharacterized Cas proteins.
• Traditional methods for identifying Cas proteins rely on sequence alignments, limiting discovery.
• Exploring novel Cas protein functions is crucial for advancing gene editing technologies.

Purpose of the Study:

• To leverage evolutionary scale language models (ESM) for identifying and characterizing unknown Cas proteins from metagenomic data.
• To predict the trans-cleavage activity of uncharacterized Cas12a proteins by integrating ESM with machine learning.
• To explore the structural and functional diversity of newly discovered Cas protein subtypes.

Main Methods:

• Employing an evolutionary scale language model (ESM) trained on CRISPR-Cas data for alignment-free Cas protein identification.
• Integrating machine learning with ESM for predicting trans-cleavage activity of uncharacterized Cas12a.
• Utilizing structural analyses and Cryo-electron microscopy (CryoEM) to investigate protein structures and RNA interactions.
• Classifying gene clusters to explore undocumented Cas protein functions.

Main Results:

• Accurate identification of Cas proteins using ESM without relying on sequence alignments.
• Discovery of 7 undocumented Cas12a subtypes with unique CRISPR loci and distinct 3D-folds.
• Identification of 8 subtypes of Cas1, Cas2, and Cas4 proteins.
• Unveiling unique RNA interactions and distinct DNA cleavage preferences (double-strand and single-strand) with broad PAM recognition for novel Cas12a subtypes.
• Development of a specific detection strategy for an oncogene SNP using a novel Cas12a PAM.

Conclusions:

• Evolutionary scale language models offer a powerful approach for discovering and functionally annotating uncharacterized Cas proteins in metagenomes.
• The discovered Cas12a subtypes exhibit unique biochemical properties and structural features, expanding the CRISPR toolkit.
• This work demonstrates the potential of AI-driven methods in accelerating the exploration of microbial diversity and its biotechnological applications.
• The developed detection strategy highlights the potential for targeted disease diagnostics using novel CRISPR-Cas systems.