Scientific work generally entails sifting through massive volumes of data, which humans find tedious but AI finds easy. BacterAI, a new platform, can do 10,000 trials each day to learn about bacteria.
Trillions of microorganisms cover nearly every surface of the human body. Many are essential to body processes, while others make you sick. Managing and analyzing the data from microbiome research is difficult.
“We know almost nothing about most of the bacteria that influence our health,” stated new research corresponding author Paul Jensen. “Understanding how bacteria grow is the first step toward reengineering our microbiome.”
To hasten findings, AI performs 10,000 trials each day on microorganisms.
AI is great at discovering patterns in large datasets, so scientists use it to analyze bacterium data. This usually entails putting existing datasets into machine-learning models, but it doesn’t assist species with little data, which is a lot since 90% of bacteria have never been studied.
BacterAI, created by Michigan researchers, can analyze microorganisms without prior information. It builds its own dataset by proposing experiments for laboratory robots to conduct one after another, with each experiment influencing the next. It can eventually synthesize its discoveries into logical principles that human scientists may examine.
BacterAI was used to study the metabolism of two common oral bacteria, Streptococcus gordonii and S. sanguinis. This entails screening through over a million amino acid combinations to find a bug-friendly combination from a list of 20 that support life.
BacterAI tested several hundred amino acid combinations every day and followed up on the most promising ones. After nine days of 10,000 trials every day, it made 90% correct predictions.
“When a toddler learns to walk, they don’t merely observe grownups walk and say ‘Ok, I got it,’ get up, and start walking. “They stumble and try,” Jensen added. “We wanted our AI bot to take steps and fall down, to have its own thoughts and make mistakes. It improves daily.”
BacterAI may accelerate bacteria research, which might lead to novel medications or other beneficial chemicals.
