To test a molecule one must make it: therefore synthesis planning is a key part of molecule discovery. In this talk I will describe two projects to improve evaluation of synthesis planning. The first, syntheseus, is a python package enabling easy use and evaluation of established retrosynthesis algorithms. The second uses a probabilistic formulation of uncertainty and is described in our recent "Retro-fallback" paper (along with a new search algorithm). I'll end with some speculation about future directions in this field.
Retro-fallback:
Syntheseus: