Last week, scientists in the U.K. announced that they have developed a robot capable of making simple scientific discoveries on its own.
The robotic system, dubbed Adam, hypothesizes about which genes in yeast code for the enzymes responsible for catalyzing certain biochemical reactions. It then carries out experiments designed to prove its hypotheses right or wrong.
The researchers described the results of Adam’s experiments in the journal Science. The group is developing Eve, another robotic lab system, which will use a similar approach for drug discovery and should be functioning this summer.
Sophisticated robotic laboratory systems such as Adam and Eve could have a big impact on future drug and genomic testing, experts say. For years, a technique called high-throughput screening (HTS) has allowed companies to rapidly perform experiments using automated devices. An HTS system can rapidly and precisely inject hundreds of thousands of compounds into samples. These automated systems can do in a few days what a team of scientists would need much longer to complete by hand.
But researchers must still evaluate the results of experiments themselves, even though they often employ software programs to analyze huge amounts of data. Adam and Eve aim to integrate the evaluation processes to make laboratory testing even faster and more efficient.
“Laboratory robotics test thousands and tens of thousands of compounds in a brute-force way,” says Ross King, a professor of computer science at Aberystwyth University and a main researcher of the intelligent robotic systems. “Our idea is to try to do this in a more intelligent way.” King says that Adam and Eve are the first systems to combine large libraries of scientific data with machine-learning techniques and sophisticated lab hardware.
“The kinds of experiments robots can carry out are very simple,” says Jim Inglese, the deputy director of the National Institute of Health’s Chemical Genomics Center. “They can’t do multiple steps; they have to have very straightforward protocol.” Smarter robotic helpers “might be part of what’s going to bring the new era of robotics into science,” he says. “We can’t take 10 years to develop a drug anymore. We’ve got to be able to do all this stuff faster.”
Eve will eventually test drugs for treating malaria and schistosomiasis (an infection caused by several kinds of parasitic worm). Eve will do this by predicting how drug molecules should interact with laboratory samples. The robot has a large data bank on chemical compounds; machine learning will be used to teach it how different shapes relate to chemical activity.
After Eve has discovered a few key compounds–ones that generate some desired activity or reactions in the laboratory–it will “make hypotheses about what could be important about the shape of the chemical that’s causing the activity,” says King. Then it will perform further experiments based on those assumptions.
Such an automated system “could definitely be useful, because there’s so much data out there,” says Anthony Johnson, CEO of Empire Genomics, a service lab that performs high-throughput testing. “If you had a machine that could understand the end objective of a given research project and design a series of experiments, that would be great,” he says.
However, because genomic testing is so complex, Johnson believes that such a system could prove most useful as a way to make current research more efficient, rather than as a means for making new discoveries on its own.
Eventually, the two robots will work together: Adam will create yeast cultures that Eve will use in its experiments. “In 10 to 20 years, I would think that such systems would be relatively common in labs,” says King.