More and more often, chemists are looking to advanced computer technology — like artificial intelligence in chemistry — to help them manage growing amounts of chemical information and make breakthroughs where traditional approaches have failed.
The pattern-finding ability of AI makes the technology particularly good at discovering relationships in large data sets — like those describing complex chemical substances. This means that AI and machine learning tools can often make surprising discoveries that human researchers and traditional analytics methods may not have found otherwise.
As a result, artificial intelligence and machine learning may revolutionize how chemists search for new compounds.
Here is how artificial intelligence is being applied to chemistry right now — and how it may soon change the field.
1. Machine Learning in Drug Discovery
Drug discovery is almost always a costly and time-consuming process. On average, it takes an estimated $2.6 billion to discover a new treatment as researchers must work over months and years to discover, synthesize and test new chemicals.
Part of the price tag, however, is due to how many drugs never make it past the first phase of testing. For every ten drugs that go Phase I testing, just one will make it past regulators and to the market.
Some researchers and major biopharmaceutical companies believe that with improved drug discovery methods, they can reduce the rate of ineffective drugs that make it to trial — making it much cheaper to develop new drug therapies.
Some of these companies are looking to the pattern-finding abilities of AI, which may be able to more accurately pinpoint compounds likely to make effective therapies.
Recently, AI technology was used by a team of researchers at the Leiden Academic Centre for Drug Research (LACDR) to find five new “substances with an inhibitory effect on a specific type of kinase.” Kinases, which are enzymes that manage the function of proteins in the body, play an important role in the development of cancer.
To discover the new substances, the team created a massive database of known compounds and their properties. A machine learning model, trained on this data, then sifted through a much larger database of compounds with unknown properties, identifying those most likely to be effective as drug therapies.
Another AI model was then used to test these flagged compounds, predicting whether or not they would be able to inhibit the kinases targeted by the research team.
2. Using AI to Understand Quantum Chemistry
AI has also been effective at helping leading chemists understand more about higher-level and more theoretical aspects of chemistry that have proven difficult to break down.
Early in May 2020, researchers announced that, for the first time, a new machine learning tool had been used to more accurately predict the energy needed to make or break certain molecules than conventional prediction methods.
Determining a molecule’s electronic structure is tough because your models need to account for all the potential states that a molecule’s electrons can be in. You also have to consider the probability of being in those states for each electron — the quantum uncertainty of a molecule’s structure.
Worse, these electrons can interact with each other — meaning the more complicated a molecule is and the more electrons it has, the greater the range of possibilities. All of this uncertainty makes the energy needed to break a molecule’s bonds difficult — unless you have the right tools.
The predictive model may help researchers better understand why molecules and electrons behave the way they do — paving the way for future quantum chemistry discoveries.
3. The Growing Impact of Artificial Intelligence in Chemistry
It’s likely that AI will become even more widely used in chemistry over the next few years. In late May 2020, two new AI drug discovery startups received a combined $200 million in investor funding — a sign that we’ll probably see new applications of AI in drug discovery in the near future.
At the same time, the researchers behind the new quantum chemistry model are already planning to test its effectiveness on more complex molecules. The team hopes to soon predict the electronic structure of molecules in the nitrogen cycle, in which nitrogen-based molecules are broken down and built up into compounds necessary for life.
Follow Us On
Get the latest tech stories and news in seconds!
Sign up for our newsletter below to receive updates about technology trends