Artificial intelligence (AI) has taken on an increasingly central role in chemistry research and industrial development in the last few decades. In particular, analytical chemistry and biochemistry for life sciences have integrated tools such as machine learning algorithms and artificial neural networks (ANNs) to process large datasets and gain new knowledge in the field of chemistry. This article examines the role that AI plays in chemistry today.
Image Credit: metamorworks/Shutterstock.com
The Recent Growth of AI
AI systems make predictions based on data models (i.e., if the data has these features and relationships, it is likely to behave in this way in the future). Generally, AI systems need to be “trained” on datasets with known values, to compare predictions with reality and refine the data models underpinning those predictions.
AI is particularly suited for solving problems that involve a lot of data, or data with complicated input and output relationships. Data like this is difficult or impractical to model with traditional manual procedures, so as AI has developed in the last few years, new modeling applications have risen alongside it.
How Do Chemists Use AI?
There are numerous chemical tasks and processes where data sets often feature complicated input and output relationships. For example, it is possible to predict how soluble a new compound will be with theoretical calculations or calculations with empirical data. AI programs that have learned about structure/solubility relationships from training data with known solubilities can also predict solubility in new compounds.
In general, chemists use AI to help reduce the effort required to design and perform experiments. This is achieved in a number of AI application areas: laboratory automation, predicting new drugs’ bioactivity, optimizing conditions for chemical reactions, and making suggestions for how to synthesize complicated target substances.
AI Applications Are Growing in Chemistry, Too
Increasingly, applications for AI in chemistry research are growing. This is due to a few complementary factors: computer processing capacity is rapidly expanding, open source machine learning frameworks and AI algorithms are being developed and shared among scientists, and chemists are becoming increasingly data literate, with university courses and postgraduate programs now introducing students to data science at an early stage.
Large public data challenges have also driven the growth of AI in chemistry, such as the ImageNet competition and Merck Molecular Activity Challenge.
Examples of open source frameworks that have contributed to the growth of AI in chemistry include TensorFlow (developed in 2015) and PyTorch (which was released in the following year).
This growth is evident in the number of articles published detailing the chemistry applications of AI. As of 2020, over half of the documents in this area were published in the preceding four years.
What Areas of Chemistry Most Rely on AI?
While all branches of chemistry – indeed, pretty much all branches of science in 2022 – have benefitted from AI technology, some have become much more intertwined with AI than others.
Going by the number of journal publications and patent documents published, analytical chemistry uses AI more and more often than any other area of chemistry.
After analytical chemistry, the next two areas using AI the most are environmental chemistry and industrial chemistry or chemical engineering.
Biochemistry is strongly represented in terms of patent applications around chemistry and AI, but less so in terms of journal publications. This is likely due to the strong financial incentive to develop and patent technologies in medical industries.
Notable Chemistry Research Using AI
In a recent literature review published in the Journal of Chemical Information and Modeling, researchers selected the 34 most influential journal publications using AI for chemistry. These articles have over 100 citations each and demonstrated novelty in their fields.
The articles used a mixture of AI techniques, especially relying on machine learning, ANNs, density functional theory, and random forest classification.
The life sciences-related chemistry articles selected often used AI to understand high throughput drug screening, analyze nucleic acid sequences, and predict protein structures.
Materials sciences applications included predicting structure/property relationships to discover new functional materials and creating memristors for neuromorphic computing.
Analytical chemistry papers used AI to develop new methods for automating flow chemistry, improving retrosynthetic planning, and predicting outcomes from chemical reactions.
In the Future, AI Will Be Even More Accessible for Chemists
As well as these cutting-edge applications for AI in chemistry, the researchers also found that notable research using AI also tended to develop user-friendly and accessible computational tools.
As AI – and computing in general – becomes more powerful, it can also become more usable. This is because developers can wrap up AI tasks based on complicated data science in user-friendly packaging, essentially abstracting the mathematical workings of the algorithm into a more intuitive language for people to navigate.
More from AZoM: A to Z of Electromagnetic Shielding
References and Further Reading
Baum, Z. et al (2021). Artificial Intelligence in Chemistry: Current Trends and Future Directions. Journal of Chemical Information and Modeling. doi.org/10.1021/acs.jcim.1c00619.
Pilkington, B., (2022). Can Machine Learning Reduce AFM Uncertainty? [Online] AZO Materials. Available at: https://www.azom.com/article.aspx?ArticleID=21867
---. (2022). How Quickly Can AI Process Data? [Online] AZO Robotics. Available at: https://www.azorobotics.com/Article.aspx?ArticleID=510
---. (2022). Why Do We Need Lab Automation? [Online] AZO Robotics. Available at: https://www.azorobotics.com/Article.aspx?ArticleID=542