Photo credit: www.sciencedaily.com
Genome editing continues to evolve rapidly, showing promise for the treatment of genetic disorders, though significant challenges remain. A recent publication by researchers from Mass General Brigham in the journal Nature explores the integration of scalable protein engineering and machine learning to advance gene and cell therapy. The study introduces a machine learning algorithm named PAMmla, capable of predicting the characteristics of approximately 64 million genome editing enzymes. This development aims to minimize unwanted off-target effects, enhance the safety of gene editing, improve efficiency, and facilitate the customization of enzymes for novel therapeutic applications.
“We believe our research is a key step toward significantly broadening the range of effective and safe CRISPR-Cas9 enzymes available for use. Our findings illustrate how PAMmla-predicted enzymes can accurately target and edit disease-related sequences in primary human cells and in a mouse model,” stated Ben Kleinstiver, PhD, a key investigator at Massachusetts General Hospital (MGH) and a member of the Mass General Brigham healthcare system. “We are eager to see these tools embraced by the research community and to extend this approach to other enzymatic properties in the field of genome editing.”
While CRISPR-Cas9 holds substantial potential for gene editing across diverse genomic locations, its conventional applications face certain limitations. A significant concern is the occurrence of off-target effects, where the enzymes inadvertently modify DNA at unintended sites. The new study seeks to address these challenges by leveraging machine learning to tailor enzymes for improved specificity, offering a scalable method that surpasses prior attempts, which often yielded far fewer engineered enzymes.
Central to the efficacy of CRISPR-Cas9 is the ability of these enzymes to recognize and bind to short DNA sequences known as protospacer adjacent motifs (PAMs). Utilizing machine learning, the research team predicted PAMs for millions of Cas9 enzymes, identifying a selection of novel engineered variants that exhibit optimal on-target activity and specificity. Initial experiments conducted in human cells and in a mouse model of retinitis pigmentosa demonstrated that these bespoke enzymes provided enhanced specificity in targeting.
A notable achievement from this research is the establishment of the PAMmla model, which researchers can now use to forecast customized enzymes tailored to their specific experimental needs. “With this model, we have created a robust toolbox of precise and safe Cas9 proteins that can be harnessed for various research and therapeutic purposes,” remarked Rachel A. Silverstein, PhD candidate, NSERC postgraduate scholar, and 2024 Albert J. Ryan Fellow at MGH, who served as the lead author on the study.
To promote accessibility, the researchers have developed a web tool allowing users to engage with the PAMmla model, which can be accessed at https://pammla.streamlit.app/.
Source
www.sciencedaily.com