BacterAI: New AI system enables robots to conduct 10,000 scientific experiments a day

However, determining the precise combination of amino acids for each bacterial species was difficult due to the large number of possible combinations resulting from the 20 available amino acids.

To overcome this challenge, BacterAI was developed to test hundreds of combinations of amino acids daily and modify the combinations based on the previous day's outcomes. Within nine days, BacterAI was accurately predicting the necessary amino acid combinations for growth 90% of the time.

Automated experimentation to speed up discoveries

Automated experimentation and the use of BacterAI have the potential to extend far beyond microbiology, allowing researchers from different fields to frame questions as puzzles for AI to solve through trial and error.

It is worth noting that approximately 90% of bacteria remain understudied, and conventional methods of obtaining basic scientific knowledge about them require a significant amount of time and resources. The rapid pace of automated experimentation can significantly speed up these discoveries, opening doors for future breakthroughs in various fields.

BacterAI is a remarkable innovation in the field of AI, thanks to its ability to create its data set via a series of experiments, as opposed to the conventional approach of feeding labeled data sets into a machine-learning model.

The platform utilizes the results of prior trials to make predictions about the experiments that can provide the most significant amount of information. With fewer than 4,000 experiments, BacterAI has figured out most of the rules for feeding bacteria accurately.

Led by Professor Paul Jensen, the team's development of BacterAI is a remarkable leap forward in the utilization of AI to expedite scientific discovery. As demonstrated by the team's research, BacterAI can provide crucial information on the microbial world, including those that influence our health.

With robots conducting up to 10,000 experiments daily, the platform's potential for accelerating discoveries is immense. The National Institutes of Health funded the study with support from NVIDIA.

Study Abstract: 

Training artificial intelligence (AI) systems to perform autonomous experiments would vastly increase the throughput of microbiology; however, few microbes have large enough datasets for training such a system. In the present study, we introduce BacterAI, an automated science platform that maps microbial metabolism but requires no prior knowledge. BacterAI learns by converting scientific questions into simple games that it plays with laboratory robots. The agent then distils its findings into logical rules that can be interpreted by human scientists. We use BacterAI to learn the amino acid requirements for two oral streptococci: Streptococcus gordonii and Streptococcus sanguinis. We then show how transfer learning can accelerate BacterAI when investigating new environments or larger media with up to 39 ingredients. Scientific gameplay and BacterAI enable the unbiased, autonomous study of organisms for which no training data exist.