One problem with gene editing technology is the extreme difficulty associated with designing a guide RNA sequence to guide the Cas9 nuclease to the correct target sequence. The reason behind this difficulty is the fact that many gRNA sequences may have severe off-target effects, since the CRISPR system can sometimes accept sites that have slight mismatches to the gRNA sequence. The problem is, current methods to design optimal sequences rely on statistical probabilities which are not very accurate, and different methods can give different results for the same sequence. Due to this, designing a CRISPR experiment, including the gRNA sequence, can take days, weeks, or even months. As such, my innovation uses machine learning and deep learning algorithms to improve our methods of designing the gRNA sequences, to design more accurate CRISPR experiments with minimal off-target effects. This is innovative as deep learning and machine learning have not yet been applied in this problem, or even field, essentially wasting resources. The program builds on previous machine learning models which predict off-target effects, such as the one designed in the study titled “Off-target predictions in CRISPR-Cas9 gene editing using deep learning” (http://doi.org/10.1093/bioinformatics/bty554). The program also contains a type of machine learning algorithm known as a genetic algorithm, which optimizes the gRNA sequence based on scoring from the other model. This can drastically improve the time and effort that goes into designing a CRISPR experiment, as well as make the mechanism more accurate, making it viable for therapeutic purposes.

What inspired you (or your team)?

I have always been passionate about medicine as well as technology, and how far medicine and healthcare can move forward with the application of upcoming technologies such as artificial intelligence, nanotechnology, in conjunction with biomedical innovations such as gene editing. As such, I was discussing gene editing with a researcher at Synthego, who discussed that one of the major challenges with gene editing is the design of the gRNA molecule, to minimize off-target effects. Motivated to solve this problem, I soon found out about genetic algorithms for text generation as well as previous research on off-target effect predictions. I combined all these areas to innovate a more effective way of performing this function. If this technology is applied successfully, it can reduce the effort required in CRISPR experiment design, bringing us closer to the ability to use CRISPR in therapeutic purposes. This can impact the healthcare system very beneficially, allowing us to cure or treat many of the disorders we consider incurable today, including some of the worst disorders today such as cancer.