Chemistry researchers at the University of Amsterdam have created an automated chemical synthesis robot equipped with an artificial intelligence machine learning system, named "RoboChem". This compact laboratory instrument surpasses human chemists in speed and precision while also displaying a high degree of ingenuity. As the first of its kind, it could significantly accelerate the chemical discovery of molecules for pharmaceuticals and many other applications, and the first results from RoboChem were recently published in the journal Science.

RoboChem was developed by Professor Timothy Noël's research group at the Van't Hoff Institute of Molecular Sciences at the University of Virginia. Their paper shows that RoboChem is a precise and reliable chemist that can perform a variety of reactions while producing minimal amounts of waste. The system works autonomously around the clock, delivering results quickly and tirelessly.

Time-lapse photography of robotic chemistry. Source: University of Amsterdam

"Within a week, we were able to optimize the synthesis of around ten to twenty molecules, which would have taken a PhD student several months," says Noël. "Not only does the robot obtain optimal reaction conditions, it also provides a set-up for scale-up. This means we can produce quantities that are directly relevant to suppliers to the pharmaceutical industry."

Noël's research group specializes in flow chemistry, a novel chemical approach that replaces beakers, flasks and other traditional chemical tools with flexible small tube systems. In RoboChem, a mechanical needle carefully collects starting materials and mixes them in small volumes of just over half a milliliter.

RoboChem is based on flow chemistry principles. The reaction is carried out in a small tube with a volume of only 650 microliters. Source: University of Amsterdam

These then flow to the reactor through a piping system. There, light from high-power LEDs triggers molecular transformations by activating photocatalysts in the reaction mixture. The optical flow then continues to an automated NMR spectrometer to identify the converted molecules. This data is fed back in real time to the computer controlling RoboChem.

"This is the brain of RoboChem. It uses artificial intelligence to process information. We use a machine learning algorithm that autonomously decides which reactions to perform. It is always aiming for the best results and constantly improving its understanding of chemistry," says Noël.

The robotic needle sampler precisely selects the amounts of various reagents and deftly mixes them together to form a reaction solution. Source: University of Amsterdam

In order to confirm RoboChem's results, the research team made great efforts. All molecules included in Science papers were isolated and examined manually.

Noël said he was impressed by the system's ingenuity: "I have been working on photocatalysis for more than ten years. Nonetheless, RoboChem showed results that I could not have predicted. For example, it found reactions that require very little light. Sometimes I have to scratch my head and wonder what exactly it does. And then you think, would we do the same thing? Looking back now, you understand the logic of RoboChem. But I doubt we would have achieved the same results ourselves. At least not so soon'.

At the heart of RoboChem is a powerful photochemical reactor that features an array of extremely powerful LEDs that illuminate the reaction solution. Here, molecules are transformed according to instructions from an artificial intelligence controller. Source: University of Amsterdam

The researchers also used RoboChem to replicate previously published findings in four randomly selected papers. They then determined whether Robochem produced the same or better results. About 80 percent of the time, the system produced better results. Noël said: 'In the other 20 per cent of cases the results were similar. This leaves me with no doubt that AI-assisted methods will benefit chemical discovery in the broadest sense. "

The point of RoboChem and other "computerized" chemistry is also to generate high-quality data that will benefit future applications of artificial intelligence. In traditional chemical discovery, only a few molecules are studied in depth. The results are then extrapolated to seemingly similar molecules. The data sets generated by RoboChem are complete and comprehensive, in which all relevant parameters for each molecule are available. This provides more insight.

RoboChem uses machine learning algorithms to process data obtained from the system. It decides which reactions to perform, always aiming for the best results. Manual intervention occurs only at the beginning, when setting up the stock solution and starting the RoboChem session. Source: University of Amsterdam

Another feature is that the RoboChem system can also record "negative" data. In current scientific practice, most published data only reflect successful experiments. Failed experiments can also provide relevant data. But this data can only be found in the researchers' handwritten experimental notes. These data are not published publicly and therefore cannot be used for AI-driven chemistry research, and RoboChem will change this.

Compiled source: ScitechDaily