Advances in computer hardware and computer technology are aimed at accelerating and improving all aspects of the classic design, synthesis, testing, and analysis cycle of medicinal chemistry. The use of data-driven synthetic prediction tools to reduce failures in the synthesis of new molecular entities has attracted more and more attention. Combining with artificial intelligence (AI) design synthetic routes, scientists use existing chemical reactions in the database to train AI algorithms to propose synthetic routes for a given molecule, including reaction conditions, and evaluate which route is best based on the number of steps and yield predictions. Building upon recent advances in machine learning, cheminformatics, and computational chemistry, Alfa Chemistry has developed a knowledge-based, computational synthesis route design platform for reaction pathway identification, scoring, and selection.
Figure 1. Computational synthesis of CuI nanoparticles immobilized on modified poly (styrene-co-maleic anhydride). (Heravi, M. M.; et al.)
At Alfa Chemistry, synthesis route can be designed in a short time according to the search and analysis algorithm. Each route is evaluated in terms of cost, availability of raw materials, the number of reaction steps, and the operational difficulty of the reaction. The optimal synthesis route is comprehensively evaluated by our experts. We provide the following services:
The bottleneck of manual synthetic evaluation can be solved by generating a hypothetical synthetic route, which is used to quickly determine the priority of compounds through a easy synthesis process. We use this method to offer a compound set as a starting point for synthetic route planning.
The general steps of extracting templates from the reaction data set through algorithms are:
1) Identify the reaction center.
2) Identify the atoms adjacent to the reaction center.
3) Add the functional groups involved in the reaction.
Models based-machine learning has the potential to provide better prediction of performance and certainty of chemical reaction, thereby speeding up searches.
At Alfa Chemistry, we are capable of using model-based mechanical learning technology which constructs an alternative model of reaction performance based on reaction conditions. Our teams hierarchically place various search strategies (for example, Bayesian optimization) on these models to help select the next set of conditions to optimize the model.
Compared with the retrosynthesis model, it's easier to use the forward reaction model to evaluate quantitatively. One of the goals of computational synthesis planning is to ensure that the results obtained through algorithm design are reliable and feasible by performing accurate prediction. Our machine learning methods applied for reaction prediction include deriving reaction rules from a predefined list of templates, graph convolutional neural networks that predict changes in atoms and bonds from the starting material to the product, and predicting the sequence of the product.
Many reactions can lead to a variety of regio or stereoisomeric compounds. The information about the by-products related to the selectivity of the reaction is of importance which can guide the priority of synthesis. We use forward reaction prediction method to analyze by-products and apply it in the design of purification strategies.
Computational synthesis planning provides an effective way to optimize the chemical process. Our computational synthesis planning services remarkably reduce the cost, promote further experiments, and enhance the understanding of chemical process for customers worldwide. Our personalized and all-around services will satisfy your innovative study demands. If you are interested in our services, please don't hesitate to contact us. We are glad to cooperate with you and witness your success!