Computational Biology Modeling with Tree Search and Learning
- Tuesday 5 December 2023
Niels Bohrweg 1
2333 CA Leiden
Computational biology problems such as RNA Inverse Folding, Stochastic Inverse Models, Protein Folding, De novo Molecular Design and Retrosynthetic Planning have very large search spaces leading to a combinatorial explosion. Based on this, solving these problems using classical AI search methods is not feasible.
Combining tree search with deep learning (DL) models as function or policy functions can reduce the search space as a result of robust generalization. While this approach is primarily mainstream in solving games such as Go and Chess, it can also be applied to solve computational biology problems. Accordingly, in this work, we show how RNA Inverse Folding, Retrosynthetic Planning and Stochastic Inverse Models can be solved and optimized using tree search and DL.