Towards "AlphaChem": Chemical Synthesis Planning with Tree Search and Deep Neural Network Policies
This work addresses the challenge of efficiently planning chemical syntheses for molecules like drugs and agrochemicals, representing an incremental improvement over existing methods.
The authors tackled the problem of chemical synthesis planning by modeling retrosynthesis as a Markov Decision Process and using Monte Carlo Tree Search with a Deep Neural Network policy trained on extensive chemical knowledge, demonstrating that this approach outperforms traditional best-first search with hand-coded heuristics.
Retrosynthesis is a technique to plan the chemical synthesis of organic molecules, for example drugs, agro- and fine chemicals. In retrosynthesis, a search tree is built by analysing molecules recursively and dissecting them into simpler molecular building blocks until one obtains a set of known building blocks. The search space is intractably large, and it is difficult to determine the value of retrosynthetic positions. Here, we propose to model retrosynthesis as a Markov Decision Process. In combination with a Deep Neural Network policy learned from essentially the complete published knowledge of chemistry, Monte Carlo Tree Search (MCTS) can be used to evaluate positions. In exploratory studies, we demonstrate that MCTS with neural network policies outperforms the traditionally used best-first search with hand-coded heuristics.