Machine Learning Algorithm Can Predict Where Proteins Go

Researchers at the Nara Institute of Science and Technology (NAIST) have developed a machine learning algorithm that can accurately predict the location of proteins related to actin, which is a crucial part of the cellular skeleton. The algorithm can predict the location of the proteins based on the actual location of actin. 

The study was published in Frontiers in Cell and Developmental Biology. 

The Importance of Actin

Actin is key to providing shape and structure to cells, and it plays a role in forming lamellipodia during cell movement. Lamellipodia are fan-shaped structures that enable cells to “walk” forwards, and they contain various proteins that bind to actin to keep the cells moving.

Shiro Suetsugu is the lead author of the study and came up with the idea during a conversation with Yoshinobu Sato at the Data Science Center in NAIST. 

“While artificial intelligence has been used previously to predict the direction of cell migration based on a sequence of images, so far it has not been used to predict protein localization,” says Suetsugu. “We therefore sought to design a machine learning algorithm that can determine where proteins will appear in the cell based on their relationship with other proteins.”

Developing the AI System

The researchers trained an artificial intelligence (AI) system to predict where actin-associated proteins would be in the cell. They did this by showing the AI pictures of cells with the proteins labeled with fluorescent markers, which indicated to the system where they were located. The system was then fed pictures in which only the actin was labeled, and it was asked to locate the associated proteins.

“When we compared the predicted images to the actual images, there was a considerable degree of similarity,” says Suetsugu. “Our program accurately predicted the localization of three actin-associated proteins within lamellipodia; and, in the case of one of these proteins, in other structures within the cell.”

Demonstrating the system’s specific abilities, the team then asked it to predict where tubulin was in the cell. Tubulin is not directly related to actin, and the program performed worse at this task.

“Our findings suggest that machine learning can be used to accurately predict the location of functionally related proteins and describe the physical relationships between them,” says Suetsugu.

According to the researchers, the program could be used to quickly and accurately identify the structures from cell images, and it could act as an artificial cell straining method, which would help avoid the limitations of current methods.