I know the robot revolution is becoming a played out meme on the internet these days, but that’s mostly because the world around us is actually becoming more and more automated. Our GPS units are learning about us as we drive. In February, the L train in New York started overnight operations with a fully computer-controlled conductor. Now, my day job as a scientist could be replaced soon by automated workstations.
A group at Aberystwyth University in the UK has developed a completely automated workstation, which is capable of developing scientific hypotheses and then designing future experiments to verify them, all without any intervention from scientists. Using this workstation, named “Adam,” the scientists discovered the genetic coding for an orphan enzyme than had no known parent gene. While this may not sound like quite an accomplishment, this is actually an impressive feat that would have been rather labor-intensive and time-consuming for the scientists to carry out by hand.
The workstation is basically a fully fleshed-out molecular biology laboratory. It’s controlled by four computers, and comes with centrifuges, spectrophotometers to measure cell growth, automated liquid handlers and freezers among many different tools. The workstation can carry out 1,000 experiments simultaneously, each lasting five days, while making measurements every thirty minutes on each sample. The software then compiles all of the data it collects, makes statistical inferences, and then designs future experiments, and again carries them out in an iterative process. The human scientists simply added laboratory consumables and removed waste — Adam carried out the rest of the work.
Science is no stranger to automation. The vast majority of drug discovery work is done on liquid handing robot. However, the vast majority of this work is just brute force. For example, in the case of drug discovery, each well of a 384 well plate is loaded with a slightly different version of a molecule to see which version has the most activity. The most promising candidates are selected (in a process called hit-picking), and then further developed. Adam performed a similar process, determining which of the 1,200 known coding regions in the yeast genome actually coded a certain gene. But whereas, in drug development, scientists try every version of a drug they can synthesize, Adam was able to intelligently select which avenues of development were the most promising, eventually determining that three separate genes actually coded together for the one final product.
Clearly a lot of human work went into developing the software and logic algorithms that controlled Adam, so scientists aren’t going to be replaced by robots anytime soon. Adam didn’t design himself, after all. However, work of this kind could accelerate scientific discovery and development. As computers become more powerful, we’ll be able to analyze larger and larger data sets, finding patterns that would otherwise be too difficult for the human mind to tease out. The work done by this group is pioneering, and could be changing the face of science as we know it.
The problem is when to draw the line between having machines doing everything, and machines aiding in the things that us humans do.