Machine Learning Can Improve The Success Rate of Genome Editing

Researchers were able to use these molecular scissors to remove, modify or add DNA to any part of the genome. This technology was used to determine which genes are essential for different conditions, such as cancer and rare diseases. It also allowed researchers to create treatments to fix or eliminate harmful mutations.

Base editors are an innovation that expands on CRISPR/Cas9, and were called “molecular pencils” for their ability to replace single bases of DNA. Prime editors are the latest generation of gene editing tools. They were created in 2019. They are known for their ability to search and replace the genome directly with high precision, which has earned them the nickname “molecular word processors”.

These technologies are designed to correct genetic mutations that cause disease in humans

Treatment for sickle cell disease is the most advanced CRISPR/Cas9 clinical trial. Patients’ red blood cells are modified to activate the foetus haemoglobin gene. This gene is not affected by the harmful sickle cell mutation, unlike adult haemoglobin. You can find more information about current clinical trials on the Innovate genomics website.

There have been more than 16,000 small deletion variants, which are small deletions of DNA bases from the genome that have been linked to diseases. Cystic Fibrosis is one example. 70% of cases are due to the deletion of three DNA bases. Base edited T-cells, which were created from cells that had been deleted by chemotherapy, were used successfully to treat leukemia in a patient who had previously failed bone marrow transplant and chemotherapy.

Researchers at the Wellcome Sanger Institute created 3,604 DNA sequences ranging in length from one to 69 DNA bases. These sequences were then inserted in three different human cell types using different prime editor delivery systems and different DNA repair contexts. The cells were then genome sequenced after a week to determine if the edits were successful.

To determine the common factors that contributed to each edit’s success, we evaluated each sequence’s insertion efficiency (or success rate). The sequence length was a significant factor as well as the type of DNA repair method involved.

Jonas Koeppel was the first author of the Wellcome Sanger Institute study. He stated: “The variables involved with successful prime edits are many, but it’s beginning to be discovered what factors increase the chance of success. Although the length of a sequence is one factor, it’s not easy to insert long sequences. Another type of DNA repair was able to prevent the insertions of short sequences. However, it prevented the insertions of long sequences.

The researchers used machine learning to help them make sense of the data. Machine learning can detect patterns that indicate success in insertion, such as the length of the DNA repair and the type of DNA. After being trained using the existing data, the algorithm could be tested with new data to confirm its accuracy in predicting success.

Juliane Weller, the first author of the Wellcome Sanger Institute study, stated: “Put simply. Several different combinations of DNA letters can encode the same amino acid within a protein. There are many ways to edit a gene in order to get the same result at the protein level. We have developed a machine-learning algorithm that ranks potential gene edits based on their likelihood of working. This will eliminate much of the trial-and-error involved in prime editing, and help speed up progress.

Next, the team will create models of all human genetic diseases. This will allow them to understand whether and how prime editing can fix them. Other research groups from the Sanger Institute, as well as its collaborators, will be involved in this endeavor.