Artificial intelligence (AI) offers several advantages to the healthcare sector. It provides precise and accurate results to complex and time-consuming diagnostics to get unprecedented insights that can sometimes be hard for humans to notice.
Recently, researchers from the Nara Institute of Science and Technology (NAIST) introduces a new machine learning program to predict protein (related to actin) location in cells based on actin location.
Early studies have shown the use of AI in predicting the direction of cell migration. However, this is the first known machine learning algorithm developed to determine protein location in cells based on their relationship with other proteins.
Actin is vital for cell shape and structure, and it aids in the formation of lamellipodia, fan-shaped structures that cells use to “walk” forward. Other proteins that bind to actin in lamellipodia maintain the fan-like form and keep the cells moving.
The researchers used images of cells in which the proteins were labelled with fluorescent markers to show where they were positioned to train an AI system. They next showed the programme photos in which only actin was labelled and asked it to locate the associated proteins.
They found remarkable similarity in the predicted images with the actual images. The program correctly predicted the location of three actin-associated proteins within lamellipodia. In addition, it could locate one of the proteins in other structures within the cell.
The team mentions that the program does not perform well when asked to predict tubulin location, which is not directly to actin. The researchers infer that their algorithm can only accurately predict the location of functionally related proteins and even describe the physical relationships between them.
Given how difficult it is for non-experts to see lamellipodia, the algorithm created in this study could be utilized in the future to quickly and accurately identify these structures from cell images. Furthermore, this method could be used as an artificial cell staining method to get around the constraints of present cell staining methods.