AI is currently at the height of its popularity. However, AI’s applications go beyond answering questions and generating art. In 2022, Demis Hassabis and John Jumper, lead researcher and primary researcher respectively, were awarded 3 million dollars for making one of the smartest AI programs in recent years. They had successfully used AI to predict around 200 million different protein structures. As of now, they have predicted nearly 215 million different protein structures.
This AI started as AlphaGo, an artificial intelligence program that played a game called “Go.” In 2016, this program beat the world champion of Go, a board game considered to be more complex than chess. Hassabis and Jumper saw the potential of this program and wanted to use it for more than games. After returning from South Korea, work on the new AI began.
The name of their AI program is called AlphaFold 2. This AI predicts the many different shapes of a protein. The AI works by being able to limit how much it searches for different proteins. It uses a process called multisequence alignment, the process of aligning three or more proteins, it creates limits which help to narrow the range of plausible structures. This helps the AI to narrow down the list of possible protein structures and correctly predict more.
Currently, there have been many plans for AlphaFold. Hassabis mentions talking to Doudna, pioneer of CRISPR gene-editing, about making crops more sustainable when faced with climate change. In a previous project, AlphaFold helped determine the structure of the nuclear pore complex, a protein complex which controles molecules exchange between the cell’s nucleus and cytoplasm. AlphaFold provided predictions that were needed to determine the structure. Overall, the future for AlphaFold and its future projects is bright.
Comments are closed.