One of the common, basic goals of most chemistry researchers is the need to predict the properties of the molecule, such as its cooking or melting point. When scientists can indicate this forecast, they are able to move forward to get discoveries that lead to medicines, materials and others. Historically, however, traditional methods of unveiling these forecasts are associated with a significant cost – spending time and consumption for equipment, in addition to funds.
Enter a branch of artificial intelligence known as machine learning (ML). ML has somewhat reduced the load on the forecasting of the particle properties, but advanced tools that are most effective in the process – learning from existing data to quickly predict new molecules – require a significant level of specialist programming knowledge from the user. This creates a barrier to the availability of many chemists who may not have significant computational experts required to move on a prognostic pipeline.
To alleviate this challenge, scientists in McGuire Research Group in myth created ChemxploremlA user -friendly computer application that helps chemists make these critical forecasts without requiring advanced programming skills. Free available, easy to download and functional on mainstream platforms, this application is also built for operation entirely offline, which helps maintain the ownership of research data. The exciting new technology is outlined in Article recently published in Journal of Chemical Information and Modeling.
One specific obstacle to chemical machine learning is the translation of molecular structures into a numerical language that computers can understand. Chemxplorl automates this complex process with powerful, built -in “molecular embedded”, which transform chemical structures into instructive numerical vectors. Then the software implements the latest algorithms for identifying patterns and accurate prediction of molecular properties, such as cooking and melting, through an intuitive, interactive graphic interface.
“The goal of Chemxplorl is to democratize the use of machine learning in chemical sciences,” says Aravindh Nivas Marimuthu, Postdoc in the McGuire group and the main author of the article. “By creating an intuitive, powerful and capable of offline computer application, we place the most modern predictive modeling directly into the hands of chemists, regardless of their programming program. This work not only accelerates the search for new drugs and materials, increasing the process of screening faster and cheaper, but also flexible design, but also flexible design also opens to future innovations.”
Chemxplorl is aimed at evolution in time, so as future techniques and algorithms are developed, they can be easily integrated with the application, ensuring that scientists are always able to access and implement the most current methods. The use of five key properties of molecular organic compounds has been tested – melting temperature, boiling point, steam pressure, critical temperature and critical pressure – and high accuracy results were achieved up to 93 percent for the critical temperature. Scientists also showed that the new, more compact method of representing molecules (Vicgae) was almost as accurate as standard methods such as MOL2VEC, but it was even 10 times faster.
“We anticipate the future in which each researcher can easily adapt and use machine learning to solve unique challenges, from developing sustainable materials to the examination of the complex chemistry of the interstellar space,” says Marimuthu. In the newspaper, he joins the senior author of the first class of assistant career development in 1943. Professor of chemistry Brett McGuire.
