Chemoenzymatic multistep retrosynthesis with transformer loops.

Area of Science:

• Computational Chemistry
• Biocatalysis
• Synthetic Chemistry

Background:

• Computer-aided synthesis planning (CASP) aims to develop selective, economical, and sustainable synthetic routes.
• Integrating enzymatic reactions into CASP offers significant advantages for green chemistry.

Purpose of the Study:

• To introduce the triple-transformer loop algorithm with biocatalysis (TTLAB) as a novel CASP tool.
• To enable chemo-enzymatic multistep retrosynthesis for designing efficient synthetic pathways.

Main Methods:

• TTLAB utilizes two triple transformer loops (TTLs): USPTO-TTL for chemical reactions and ENZR-TTL for biotransformations.
• Each TTL independently performs single-step retrosynthesis by predicting starting materials and reagents/enzymes, validated by a forward transformer.
• A heuristic best-first tree search combines predictions for multistep sequence exploration.

Main Results:

• TTLAB successfully explores multistep sequences by integrating predictions from both chemical and enzymatic reaction models.
• The algorithm proposes short synthetic routes from commercially available building blocks.
• Enantioselective biocatalytic steps are effectively incorporated into the designed routes.

Conclusions:

• TTLAB represents a significant advancement in CASP for chemo-enzymatic synthesis.
• The tool assists researchers in designing efficient and greener synthetic routes by leveraging biocatalysis.
• This approach facilitates the development of novel and sustainable chemical manufacturing processes.