Prediction of Organic Reaction Outcomes Using Machine Learning
Connor W Coley1, Regina Barzilay1, Tommi S Jaakkola1
1Department of Chemical Engineering and Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, Massachusetts 02139, United States.
This study introduces a new AI model for predicting chemical synthesis outcomes, improving the accuracy of computer-aided retrosynthesis planning. The model successfully identifies the correct major product in over 70% of cases, enhancing experimental success rates.
Area of Science:
- Computational Chemistry
- Artificial Intelligence in Chemistry
- Organic Synthesis
Background:
- Computer-assisted retrosynthesis planning tools have limited adoption due to unreliable reaction outcome predictions.
- A key challenge is the discrepancy between predicted and experimentally observed reaction results.
Purpose of the Study:
- To develop a novel model framework for accurately anticipating chemical reaction outcomes.
- To improve the reliability and adoption of computer-aided synthesis design software.
Main Methods:
- A hybrid approach combining reaction templates with neural network pattern recognition.
- Training a model on 15,000 experimental reaction records from US patents.
- Utilizing a unique edit-based representation for candidate reactions.
Main Results:
- The model achieved a rank 1 prediction for the major product in 71.8% of cases.
- Rank ≤3 and rank ≤5 predictions were achieved in 86.7% and 90.8% of cases, respectively.
- Demonstrated high accuracy in identifying experimentally observed major products.
Conclusions:
- The developed model framework significantly enhances the prediction accuracy of chemical reaction outcomes.
- This approach offers a promising solution to improve the reliability of retrosynthesis planning software.
- The findings pave the way for more effective computer-aided design in chemical synthesis.
Related Concept Videos
Predicting Reaction Outcomes
Predicting Products: SN1 vs. SN2
With increased substitution on the alkyl halide,...
Predicting Products: Substitution vs. Elimination
The following factors can influence the mechanisms competing against each other:
Classification of Titrimetric Analysis Based on Reaction Types
Titrations between an acid and a base lead to neutralization reactions that form...
E1 Reaction: Kinetics and Mechanism
E2 Reaction: Kinetics and Mechanism
