Proteins are essential for cell function, each with a specialized role. Scientists have long known that a protein’s structure determines its function. Recently, they found that a protein’s localization within a cell is also crucial.
Cells contain compartments that organize proteins. Knowing where a protein is and which proteins it interacts with helps us understand its role, but predicting this information has been challenging.
AlphaFold, an AI tool, predicts protein structure from amino acid sequences. MIT researchers, led by Professor Richard Young, have developed ProtGPS, a model that predicts where proteins localize within a cell and how mutations affect this. ProtGPS can also design new proteins for specific compartments.
The researchers demonstrated that ProtGPS could predict which of 12 known types of compartments a protein will localize to and whether a disease-related mutation will alter this localization. They also created a generative algorithm to design new proteins to target specific compartments.
ProtGPS was validated with lab tests, bridging computational design and real-world application. This research helps us understand how proteins function, how mutations disrupt processes, and how to design therapies for cell dysfunction.
Revealing the Mystery of Protein Function
Regina Barzilay, the School of Engineering Distinguished Professor for AI and Health in MIT’s Department of Electrical Engineering and Computer Science and principal investigator in the CSAIL, said, “It really excited me to be able to go from computational design all the way to trying these things in the lab. There are a lot of exciting papers in this area of AI, but 99.9 percent of those never get tested in real systems. Thanks to our collaboration with the Young lab, we could test and really learn how well our algorithm is doing.”
Researchers trained and tested ProtGPS on two groups of proteins with known locations and found it could accurately predict where proteins go. They also tested how ProtGPS could predict changes in protein location based on disease-related mutations. These mutations can cause diseases, but the exact mechanisms are often unknown.
Understanding how mutations lead to disease is essential for developing therapies. The researchers suspected that many disease-related mutations might change protein localization. For example, a mutation could prevent a protein from joining essential compartments.
They tested this using ProtGPS to analyze over 200,000 proteins with disease-related mutations. ProtGPS predicted where mutated proteins would go and measured changes from their normal locations. Significant changes indicated potential mis-localization.
The researchers found many cases where mutations appeared to change protein localization. They tested 20 examples in cells using fluorescence, confirming ProtGPS’s predictions. This supports the idea that mislocalization might be an overlooked cause of diseases and demonstrates ProtGPS’s value in understanding diseases and finding new therapies.
“The cell is complex, with many components and interactions,” says Mitnikov. “We can study these systems and develop new therapies with this approach.”
The researchers hope others will use ProtGPS, such as predictive structural models, to advance protein function, dysfunction, and disease research.
Researchers wanted their prediction model, ProtGPS, to predict protein localizations and design new proteins. The goal was to create new amino acid sequences that would localize to specific cellular compartments. They restricted the algorithm to designing proteins similar to those found in nature, as nature has optimized protein sequences over billions of years.
Collaborating with the Young lab, the team tested their protein generator. The model performed well, creating 10 proteins aimed at the nucleolus. Four are strongly localized there, with others showing some bias toward the location.
Generating functional proteins could enhance drug development by designing drugs that localize to specific compartments, improving efficacy and reducing side effects.
The researchers are excited about using their model to design proteins with other functions, expanding therapeutic possibilities. They see ProtGPS as a promising tool for understanding protein localization and mislocalization in diseases and developing new therapies.
With this advancement, they aim to expand the model’s predictions, test more therapeutic hypotheses, and design more functional proteins for various applications.
Journal Reference:
- Henry Kilgore, Itamar Chinn, et al. Protein codes promote selective subcellular compartmentalization. Science. DOI: 10.1126/science.adq2634
Source: Tech Explorist
