In a science lab, robots carry out scientific experiments at lightning speed. Thanks to artificial intelligence, they are able to carry out up to 10,000 experiments a day that could revolutionize research in fields such as medicine, agriculture and environmental sciences.
The artificial intelligence platform, called BacterAI, was developed by a team of researchers from the University of Michigan in the United States. It determined the metabolism of two microbes associated with oral health without initial information. Bacteria use combinations of the 20 essential amino acids, but each species needs specific nutrients to grow. The team wanted to know which amino acids are needed by beneficial microbes in our mouths to help them grow.
The challenge was to find the combination of amino acids that the bacteria preferred. These 20 amino acids create more than a million possible combinations. However, BacterAI managed to discover the amino acid requirements for the growth of two bacteria: Streptococcus gordonii and Streptococcus sanguinis. BacterAI tested hundreds of amino acid combinations daily and adjusted its combinations each morning based on the previous day’s results. In just nine days, it provided accurate predictions 90% of the time.
Unlike traditional approaches that use labeled datasets for a machine learning model, BacterAI creates its own dataset through a series of experiments. By analyzing the results of previous experiments, she makes predictions about which new experiments could give her the most information. With less than 4,000 experiments, she discovered most of the rules for feeding bacteria.
About 90% of these bacteria are poorly understood, and the investment of time and resources to obtain basic scientific information about them using traditional methods is indeed enormous. Automated experimentation can dramatically accelerate these discoveries. In a single day, the team performed up to 10,000 experiments.