Choosing the Strategic Disconnection: How to Choose a Disconnection in Retrosynthesis

You have the target drawn: a substituted ring, an amide on one arm, a secondary alcohol near a quaternary center, an aryl group off the side. Six or seven bonds look breakable, and the hard part of retrosynthesis is rarely seeing a disconnection — it is choosing which one to commit to first. This post treats how to choose disconnection in retrosynthesis as a ranking procedure: score each candidate bond against a handful of factors, then take the one that wins on the most of them. It is a principled first cut, not a single rule that settles every target.

Why "how to choose disconnection in retrosynthesis" is a ranking problem

Most disconnection guides hand you a list of rules — disconnect at the carbon bearing the heteroatom, break to a known reaction, use symmetry — and stop there. The trouble is that on a real target, three or four of those rules fire at once and point at different bonds. The skill is weighing them. Treat each plausible disconnection as a candidate and run it past the factors below in priority order. The early factors are the heaviest: a candidate that wins on skeleton simplification usually wins overall, even if it loses a tiebreaker later.

Factor 1 ⇒ which disconnection simplifies the skeleton most

Corey's first heuristic still dominates: prefer the disconnection that removes the most molecular complexity. A bond whose cleavage strips off a whole appendage, collapses a ring, or removes a stereocenter buys more than one that trims a methyl. If one candidate halves the carbon count of the largest fragment and another shaves three atoms, the first one wins this factor outright.

Factor 2 ⇒ does the bond map to a reliable named reaction

A disconnection is only useful if the forward reaction actually runs, so look for a recognizable retron — a structural pattern that signals a dependable transformation. A β-hydroxy carbonyl (a 1,3-oxygenation pattern) points to an aldol; an α,β-unsaturated carbonyl, to a Michael addition; a biaryl or styryl bond, to a cross-coupling. A suitably substituted cyclohexene points to a Diels–Alder, which builds the ring and sets up to two stereocenters in one step.

When two candidates simplify the skeleton about equally, the one with the cleaner retron-to-reaction map should win. A bond that maps to a coupling you have run before beats a bond that needs a reaction you would have to go read up on. If you are unsure a fragment is even a sensible structure, sketch it in a SMILES-to-structure tool before you build the route around it.

Factor 3 ⇒ will the disconnection let you converge

Prefer a disconnection that splits the target into two fragments of comparable size over one that peels off a small piece and leaves a long linear chain. Convergent routes carry material forward more efficiently: a step late in a linear sequence costs you everything you invested upstream, while a convergent coupling joins two pieces built in parallel. So when a central bond — an amide, an ether, a C–C coupling near the middle — cuts the target roughly in half, weight it heavily. Two fragments of 8 and 9 carbons are a better starting position than fragments of 2 and 15.

Factor 4 ⇒ is the bond at a branch point or next to a functional group

Carbons that carry a heteroatom or sit at a branch point are the natural disconnection sites, because that is where polarity gives you a handle. A C–N bond at an amide, a C–O at an ester or ether, a C–C next to a carbonyl — each maps to a forward bond-formation with well-understood reactivity. Disconnecting an unactivated C–C bond in the middle of an alkyl chain, by contrast, leaves you with two fragments that have no obvious way to be rejoined.

Factor 5 ⇒ ring construction vs ring functionalization

When the target contains a ring, decide early whether to build the ring or build onto it. If a ring-forming reaction would set up its substituents cleanly — a cyclohexene from a Diels–Alder, a saturated N-heterocycle from a reductive amination or lactamization — that disconnection often simplifies more than picking off substituents one at a time. If the ring is a plain benzene or a common heterocycle, do the opposite: keep it intact, buy it, and disconnect the substituents.

Factor 6 ⇒ symmetry and commercial availability

Two cheaper factors break ties. Symmetry: if the target has a local two-fold symmetry, a disconnection that exploits it lets you make two identical fragments from one precursor. Availability: fragments that are catalog compounds beat fragments you would have to synthesize. A precursor you can order tomorrow outranks an elegant one you would spend three steps building.

Putting the factors in order

The factors are not equal. Run candidates through them in roughly this priority:

• Skeleton simplification — the heaviest factor; the most-simplifying disconnection usually wins.
• Reliable reaction — the simplification only counts if the forward step runs.
• Convergence — favor splits into two comparable fragments.
• Functional handle / branch point — cut where polarity gives you reactivity.
• Ring strategy — build the ring or build onto it, decided early.
• Symmetry and availability — tiebreakers, not drivers.

Sometimes the highest-ranked disconnection produces a fragment with the wrong polarity for any standard reaction — two electrophilic carbons that need to be joined, say. That is the signal the disconnection needs umpolung to invert the polarity before it is practical. And once the bonds are settled, your fragments are still synthons; mapping them to real reagents is the difference between synthons and their synthetic equivalents.

Where this fits, and how to test a candidate

Choosing a disconnection is one decision inside route planning. For the broader sequence — setting up the target, working back through layers, pruning the tree — see the full step-by-step retrosynthesis workflow. Once a route is on paper, the chosen synthons turn into amounts and limiting reagents; that work feeds turning synthons into a reagent table. When the candidate tree grows large, a survey of retrosynthesis software covers where the tools help and where they still need a chemist's judgment.

The disconnection approach traces to E.J. Corey, whose work on the logic of chemical synthesis won the 1990 Nobel Prize in Chemistry. Its retron-and-transform vocabulary is summarized in the retrosynthetic analysis reference, and the cycloaddition behind so many ring disconnections has its own Diels–Alder reaction entry.

For a fast check on any candidate, sketch the two fragments in the SMILES-to-structure tool and ask whether each piece is something you would want to make. A disconnection can look clean on paper yet leave a precursor harder to source than the target. The one that gives you two precursors you would happily start from is usually the one to commit to.