What is retrosynthesis? A working chemist’s guide
Retrosynthesis is how chemists work backward from a target molecule to buyable starting materials. Here is the method, the logic, and how AI planners actually help.
What is retrosynthesis?
Retrosynthesis is the technique of designing a synthesis by working backward from a target molecule to simpler, commercially available starting materials. Introduced by Nobel laureate E. J. Corey in the 1960s, it breaks a complex structure into smaller fragments through a series of "disconnections" until every piece can be bought or easily made. It is the planning language every synthetic organic chemist uses before touching a single reagent.
Why chemists work backward
Working forward from starting materials means guessing which of thousands of reactions to try. Working backward constrains the problem: each disconnection corresponds to a known, reliable reaction run in reverse. This turns an open search into a structured tree of choices, which is exactly why retrosynthesis maps so well onto software.
How to do a retrosynthetic analysis
- Identify the target: Draw the full structure of the molecule you want to make.
- Find strategic bonds: Locate bonds whose disconnection simplifies the skeleton the most — usually next to functional groups, rings, or branch points.
- Disconnect to synthons: Break a strategic bond to produce idealized fragments called synthons, then map each synthon to a real reagent (its synthetic equivalent).
- Repeat until buyable: Apply disconnections recursively until every fragment is a commercially available compound.
- Reverse the tree: Read the disconnections forward as a forward synthesis, then check feasibility, selectivity, and yield at each step.
Synthons and synthetic equivalents
A synthon is an idealized fragment — often a charged species like an acyl cation — that captures the bonding logic of a disconnection. Its synthetic equivalent is the real, stable reagent you actually use. For example, the acyl cation synthon corresponds to an acyl chloride in the flask. Keeping the two separate is what stops a paper plan from failing at the bench.
How AI retrosynthesis planners work
Modern tools such as AiZynthFinder search the disconnection tree automatically, scoring routes against libraries of known reactions and stocks of purchasable building blocks. They return ranked synthetic routes in seconds. The catch: a planner can propose a step that violates atom conservation or chemical reality. That is why Cheemly runs every proposed structure through a deterministic Critic Gate — RDKit parsing plus atom-balance checks — before showing it to you.
Frequently asked questions
- What is the difference between synthesis and retrosynthesis?
- Synthesis builds a molecule forward from starting materials to product. Retrosynthesis is the planning step that works backward from the product to those starting materials, deciding which reactions to run before any are performed.
- Who invented retrosynthesis?
- E. J. Corey formalized retrosynthetic analysis in the 1960s and received the 1990 Nobel Prize in Chemistry partly for this contribution. It is now the standard framework for planning organic syntheses.
- Can AI do retrosynthesis reliably?
- AI planners like AiZynthFinder propose ranked routes quickly, but they can suggest chemically invalid steps. Reliable use pairs the planner with deterministic verification — atom conservation and structure parsing — so impossible routes are filtered out before a chemist trusts them.