Researchers from the Massachusetts Institute of Technology (MIT) in Cambridge have used machine learning to uncover a potential improvement to a gene therapy treatment for Duchenne muscular dystrophy (DMD).
The treatment uses antisense phosphorodiamidate morpholino oligomers (PMO), a synthetic molecule that enters cells and modifies the dystrophin gene to produce a key protein that is usually absent in DMD patients. Through the use of machine learning, the researchers discovered highly effective peptide sequences that can be attached to PMO to boost its delivery into the cell.
Duchenne muscular dystrophy is a rare genetic disease that causes the progressive weakening of muscles, eventually leading to cardiac and respiratory failure. It is usually diagnosed in young boys, and the prognosis is very poor; patients tend to lose their ability to walk by 12 years, and they have a life expectancy of only 26 years.
In 2016, Sarepta Therapeutics, a biopharmaceutical company for rare diseases, announced the approval of a breakthrough DMD gene therapy drug, Exondys 51. It works by directly targeting the mutated gene that causes DMD, but it has one major flaw — it is not very adept at crossing the cell and nuclear membranes. A proposed method of addressing this problem is to affix cell-penetrating peptides (CPPs) to the drug to boost its efficacy in entering cells.
Read more about DMD experimental therapies
To discover which peptide sequence could do this best, MIT researchers mixed and matched 57 different peptides to create a library of 600 miniproteins that were each attached to PMO. Using an assay, they were then able to quantify how well each miniprotein performed, and, for simplicity, they chose to represent each miniprotein as a barcode. They used machine learning to process the miniprotein library to propose novel miniprotein sequences that were most effective at aiding the cell delivery of the drug.
The machine-learning model was highly successful in proposing sequences that were more effective than any previous variant, with 1 sequence achieving an impressive result of boosting PMO delivery by 50-fold. Researchers validated the results by injecting mice with the proposed sequences.
Although it is too early to predict how this work will impact patients with DMD, the method of attaching CPPs to drugs could potentially lower patient exposure to the drug without losing any of its benefits. An MIT News article on the researchers’ work notes: “As a testament to the concept, recent clinical trials demonstrated that a proprietary CPP from Sarepta Therapeutics could decrease exposure to PMO by 10-fold.” The novel method developed by MIT researchers could be used to improve cell delivery of other drugs as well, potentially speeding up the development of powerful new drugs for treating other rare diseases.
Machine learning discovers new sequences to boost drug delivery. MIT News. August 10, 2021. Accessed August 11, 2021.