Artificial intelligence can predict chemical reaction performance in addition to creating new materials

【introduction】

Machine learning methods are becoming part of the scientific inquiry of many disciplines. Machine Learning (ML) is the study and construction of computer algorithms that can be learned from data. The discovery of new materials and the chemical synthesis in our minds are still basically in the scene where traditional researchers are wearing white coats and holding various chemical reagents in their hands. I don't know that the development of artificial intelligence has been applied to various fields in recent years. Artificial intelligence really makes people accept that it should be AlphaGo warfare genius Ke Jie and become famous. It can record the chess behavior of hundreds of top players through data, through the big data analysis, the background optimization algorithm to achieve the more competitive. In the field of material chemistry, artificial intelligence is also playing an increasingly important role. Often, researchers can't do anything they can't do. It can give the best answer after thousands of calculations.

[Introduction]

On April 13, 2018, Beijing time, Science published the Abigail G. of Princeton University online. Doyle, Merck Sharp & Dohme Spencer D. Dreher (Common Communications) et al. entitled "Predicting reaction performance in C-N cross-coupling using machine learning", the team demonstrated that machine learning can be used to predict the performance of synthetic reactions in multidimensional chemical spaces, using high throughput The data obtained from the experiment. The creation of scripts to calculate and extract atomic, molecular and vibrational descriptors for palladium-catalyzed cross-coupling of Buchwald-Hartwig aryl halides with 4-methylaniline in the presence of various potential inhibitor additives. Using these descriptors as inputs and reaction yields as outputs, the random forest algorithm provides significantly improved prediction performance over linear regression analysis.

[Graphic introduction]

Figure 1 ML application in response prediction

Figure 2 Test set performance chart

Figure 3 Addition prediction

Figure 4 Model analysis