Machine Learning Creates a Map of Smelly Molecules

Nov 27, 2023 ·

Machine learning has revolutionized the way scientists study molecules. A new study published in Science showed that machine learning can be used to map out a vast database of molecules with an unprecedented level of detail. The research team, led by Professor Vasant Dhar of New York University, used machine learning to create an enormous map of the chemical space of odor molecules.

The team began by training a deep learning algorithm on a large dataset of known molecules. This dataset was obtained from databases like PubChem and ZINC, which contain over 400,000 different molecules. Next, the researchers used this algorithm to generate a massive database of millions of potential odor molecules, many of which had never been seen before.

Using machine learning, the team was able to accurately identify each molecule's structural information, such as its shape, size, and composition. They were also able to classify each molecule as being either "odorless" or having an identifiable smell. Once these molecules were categorized, they could then be used to build the massive map of the chemical space of odor molecules.

One of the most remarkable aspects of the study is that it was able to accurately predict previously undiscovered odor molecules. By analyzing the data generated by the machine learning algorithm, the team was able to detect molecules that hadn't been observed before. In addition, the team was able to use the data to better understand how odor molecules interact with one another, providing valuable insights into the structure and characteristics of odors.

Overall, the study showed that machine learning is capable of creating large and detailed maps of the chemical space of odor molecules. This could have a wide range of applications, from helping develop more effective consumer products to aiding in the diagnosis of diseases. By exploring and utilizing this new technology, scientists may be able to gain a greater understanding of the complexities of smell.