MIT researchers have developed a computational approach to help predict mutations that can create optimized versions of certain proteins, working with a relatively small amount of data. The team believes the system could lead to potential medical applications and neuroscience research tools.
Usually, protein engineering begins with a natural protein that already has a desirable function, like emitting fluorescent light. This protein is put through several rounds of random mutation to develop an optimized version. This process has resulted in optimized versions of numerous key proteins, like the green fluorescent protein (GFP). But for some proteins, creating an optimized variant has been a significant challenge.
The model developed by the MIT researchers provided proteins with mutations projected to create improved versions of GFP and a protein from adeno-associated virus (AAV), commonly used for delivering DNA in gene therapy. They hope the model could be used to develop additional tools for neuroscience research and medical applications.
The researchers also managed to enhance the viral capsid of AAV, optimizing the capsid for its DNA payload packaging ability. Shahar Bracha, an MIT postdoc, mentioned that GFP and AAV were used as proof-of-concept to demonstrate the method works on well-characterized data sets, suggesting its applicability to other protein engineering problems. Now, the researchers plan to apply this computational technique on data related to voltage indicator proteins.
The research was funded by various organizations, including the U.S National Science Foundation, the Machine Learning for Pharmaceutical Discovery and Synthesis consortium, the Abdul Latif Jameel Clinic for Machine Learning in Health, and the DTRA Discovery of Medical Countermeasure Against New and Emerging threats program.
