Behind IBM’s Autonomous Chemistry Lab

With no stoppage of the COVID pandemic in the foreseeable future, drug discovery has been a critical concern for the healthcare industry. But, in reality, it takes around a decade with approximately $10 million to discover a new drug, get it tested and bring it to the market. Such a delayed process can be majorly attributed to the tedious process of repetitive experiments to bring out the drug. Thus, to omit this significant hurdle in drug discovery, IBM has launched an autonomous chemistry lab combining AI, robotics and cloud technology — RoboRXN.

A cloud-based AI model, RoboRXN has been developed to perform autonomously, like a robot, using machine learning algorithms to produce chemicals and drugs in a laboratory with little to no human intervention. Not only a development like that can optimise the drug discovery process by reducing the production time, but can also lower down the cost, allowing scientists to bring out new drugs in health crises, keeping the norms of social distancing in place.

Teodoro Laino, a researcher at IBM, stated in the company blog that chemists and scientists rely on traditional processes in discovering drugs and other chemicals, and little progress has been made to modernise those processes of trial and error. And therefore, IBM has automated this process to accelerate material discovery.

Also Read: Google’s New Machine Learning-Based Initiative For Accelerated Drug Discovery

Overview Of IBM’s Autonomous Lab

According to the blog post by Laino, the development of RoboRXN initiated when IBM began working on developing ML models to predict chemical reactions and offered it as a free service — called it as RXN for Chemistry. The tool leveraged neural machine learning translation method for predicting the product of the chemical reactions, using the SMILE representation to identify chemical entities.

Download our Mobile App

Using SMILE representation, this molecule is translated into BrCCOC1OCCCC1

However, mixing all those ingredients and experimenting with the trial and error methods with all the components was still time-consuming. And thus, in an attempt to reduce that, IBM combined AI with cloud and automation technology to bring our RoboRXN, as a remotely accessible lab for better production of drugs. 

Launched as an online tool — RoboRXN works through a web browser, where the scientists can use the blank canvas to type and draw the required “skeletal structure of the molecular compounds.” With that information, the online platform uses ML algorithms to predict the components needed and learn the right process of synthesising the elements. To facilitate this, the company designed the algorithm to precisely extract the necessary synthesis information and converted the same in a machine-readable format. This information is then sent to a robot placed in the independent lab to execute the formula. The platform also sends a real-time report to the scientists with the results.

PC: IBM Research

RoboRXN further claimed on the company blog that it does not rely its workings on individual entities of sentences, nor does it require to be mentioned on the words that need corresponding actions. This, in turn, allows the model to be more flexible and accurate. Furthermore, in order to provide data input for the ML model, the company also developed an annotation framework that generated examples on a variety of sentences with “synthesis procedures and corresponding operations.”

Moreover, IBM claims that RoboRXN for predicting reactions can easily outstrip all other data-driven models with 90% accuracy. To build the reaction dataset, the company extracted the chemical information using NLP called Molecular Transformer. When asked, Laino explained, in the company blog, the importance of cleaning up the data in order to achieve an accurate prediction model.

The model has been trained on a large number of chemical recipes and also been designed to learn the specifics of chemicals in order to recommend the correct sequence of the components. Not only RoboRXN is hardware agnostic, but also envelopes AI technology to execute a complex series of chemical procedures. With its cloud-native design, it has been designed to be accessed from anywhere remotely.

Also Read: Artificial Intelligence swarms drug discovery. Is India catching up too?

Wrapping Up

Claiming to revolutionise industrial chemistry, the launch of RoboRXN, with its reliable results, can be extremely beneficial for drug discovery amid this COVID pandemic. The combination of robotics, along with artificial intelligence and cloud, can not only reduce the time of the discovery process but also create more emphasis on design and innovation. With this development, it is hoped to be beneficial for researchers and chemists with the new treatment for COVID-19 or any future viruses possible.