Abstract. Retrosynthetic planning is a critical task in organic chemistry which identifies a series of reactions that can lead to the synthesis of a target product. The vast number of possible chemical transformations makes the size of the search space very big, and retrosynthetic planning is challenging even for experienced chemists. However, existing methods either require expensive return estimation by rollout with high variance, or optimize for search speed rather than the quality. In this paper, we propose Retro*, a neural-based A*-like algorithm that finds high-quality synthetic routes efficiently. It maintains the search as an AND-OR tree, and learns a neural search bias with off-policy data. Then guided by this neural network, it performs best-first search efficiently during new planning episodes. Experiments on benchmark USPTO datasets show that, our proposed method outperforms existing state-of-the-art with respect to both the success rate and solution quality, while being more efficient at the same time.
Illustration of the proposed neurally-guided search algorithm framework.
We use circles to represent molecule nodes, and squares to represent reaction nodes. The search tree is an AND-OR tree \(T\), where each AND node needs all its children to be proved, while OR node requires at least one to be satisfied. This is because a molecule can be synthesized using any one of its children reactions (or-relation), and each reaction node requires all of its children molecules (and-relation) to be ready.
An iteration consists of three phases.
(a) In the selection phase, one of the frontier molecule nodes is selected according to the cost estimation \(V_t(m|T)\).
(b) Then the an AND-OR stump is expanded from the selected node. All the new reactions and molecules are added to the tree.
(c) Finally the values inside the tree are updated using the \(V_m\)s from the newly added molecules.
Short video describing the work. Recorded for ICML’20.