Every synthesis project starts with the same question: we know the molecule we want, so how do we make it? For anything beyond trivial structures, answering that question forward, starting from raw materials and guessing toward the product, is hopeless. There are simply too many possible reactions, reagents, and intermediates.
Retrosynthesis inverts the problem. Instead of asking "what can I build from these starting materials?", the chemist asks "what could this molecule have come from?", and keeps asking, step by step, until the answer is a set of simple, commercially available compounds.
This backwards logic has been the backbone of organic chemistry for over half a century. What has changed recently is who, or what, can perform it. Advances in artificial intelligence have turned retrosynthesis from a skill that took a decade to master into a capability any chemist can access in minutes. This guide covers what retrosynthesis is, the concepts behind it, where it is applied across the chemical industry, and how AI tools like Scria are changing the economics of route design.
What Is Retrosynthesis?
Retrosynthesis (or retrosynthetic analysis) is a problem-solving technique in which a target molecule is transformed, step by step, into progressively simpler precursor structures, until the analysis reaches starting materials that are simple, known, or commercially available. The plan is then executed in the lab.
E. J. Corey developed the formalism at Harvard in the 1960s, whose systematisation of "the logic of chemical synthesis" earned the 1990 Nobel Prize in Chemistry. Before Corey, synthesis design was largely an intuitive process passed from mentor to student. After Corey, it became a teachable, repeatable discipline, and, decades later, one that software can learn.
The direction of reasoning is what defines the method. Synthesis works forward: from starting materials toward the product, executed at the bench. Retrosynthesis works backward: from the target toward starting materials, on paper or in software. A complete project uses both. Plan backward, run forward.
Key Concepts and Terminology
Disconnection: The core mental move: an imaginary bond cleavage that breaks the target into two (or more) simpler fragments. Disconnections are written with a special open-bodied arrow (⇒) to signal that this is analysis, not a real reaction.
Synthon: The idealised, often charged fragment produced by a disconnection, such as an acyl cation or a carbanion. Synthons are thinking tools. They may not be stable or even isolable species; their job is to point the chemist toward a real reagent.
Synthetic equivalent: The actual purchasable or preparable reagent that behaves like a given synthon in the flask. A carbanion synthon, for instance, is typically realised in practice with a Grignard or organolithium reagent.
Functional group interconversion (FGI). A retrosynthetic step that swaps one functional group for another (for example, viewing an amine as coming from a nitro group) to unlock a better disconnection further back in the analysis.
Retron: The structural pattern in a molecule that signals a particular transform can be applied. An ester group is the retron for an esterification transform. Recognising retrons is what lets an experienced chemist, or a trained model, spot a disconnection instantly.
Transform: The reverse of a reaction. Where a reaction converts A into B, a transform converts B back into A on paper. Retrosynthesis is the systematic application of transforms.
Retrosynthesis Example: Aspirin
Aspirin (acetylsalicylic acid) is the classic teaching example because a single, obvious disconnection solves it.
Looking at the aspirin molecule, the ester linkage is the natural place to cut. Disconnecting the ester C–O bond reveals two fundamental precursors:
- Salicylic acid, a cheap, commodity chemical available from any supplier
- Acetic anhydride, a standard acylating agent
Forward synthesis is then a single acetylation step: Treat salicylic acid with acetic anhydride, and aspirin forms in high yield. One disconnection, one forward reaction, two purchasable starting materials. That is exactly why aspirin has been “manufactured at scale for over a century”
Reading a retrosynthetic scheme takes a moment of adjustment: the open arrow (⇒) means "can be disconnected to," not "reacts to give." The analysis runs backward in time; the lab work runs forward.
Real targets are rarely this kind. A molecule with five stereocentres, a fused ring system, and sensitive functional groups might need 10 to 20 forward steps, each chosen from among many plausible disconnections. That is where the retrosynthetic tree comes in.
The Retrosynthetic Tree
Each disconnection of a target produces precursors, and each precursor can itself be disconnected further. Repeating this recursively generates a branching structure called the retrosynthetic tree: the target at the root, purchasable materials at the leaves, and every internal node a hypothetical intermediate.
A route "wins" when every one of its branches terminates in a compound that is commercially available or already known. Because most targets admit several sensible first disconnections, and each resulting intermediate admits several more, the tree exposes multiple complete routes side by side. A 6-step route using an expensive catalyst can be compared directly against an 8-step route using commodity reagents.
The tree is what makes retrosynthesis both powerful and painful. Powerful, because alternatives become visible and comparable. Painful, because the search space grows exponentially: if each intermediate admits even 10 plausible disconnections, a 5-level analysis already implies on the order of 100,000 branches. No human evaluates that exhaustively.
AI-Powered Retrosynthesis: From Art to Search Problem
Because retrosynthesis is fundamentally a tree search over a learned set of transformations, it is exceptionally well suited to machine intelligence. Modern (CASP) systems combine three ingredients:
Reaction knowledge at scale. Models trained on millions of published and proprietary reactions encode transformations no single chemist could hold in memory.
Efficient tree search. Algorithms such as “Monte Carlo tree search,” guided by neural policy networks, explore the exponential route space intelligently rather than exhaustively, spending compute on the most promising branches.
Real-world constraints. Good systems score routes on purchasable starting materials, step count, expected yield, hazard profile, and cost, not just on-paper feasibility.
The result is a change in the chemist's job rather than a replacement. Instead of spending weeks generating three candidate routes, the chemist spends hours evaluating fifty machine-generated ones, applying judgment where it actually matters: feasibility of tricky steps, scalability, freedom to operate, and plant fit.
Where our Scria deep research module fits
Scria is Scinode's AI retrosynthesis engine. Enter the target product, and Scria generates ranked synthetic routes in minutes, scored on step count, starting-material availability, and stoichiometry.
What makes Scria different from standalone retrosynthesis software is what happens after the route is generated. A standalone tool ends at a proposed route; you still have to source the starting materials, find a manufacturer with the right reaction capabilities, and manage compliance separately. Because Scria runs inside the Scinode platform, every route connects directly to what comes next:
- Live supplier availability for the starting materials the route requires
- Vetted manufacturing partners matched to the reaction capabilities each step demands
- The compliance pathway for your target market, from the same platform.
A workflow looks like this:
Input the target: Search by name or CAS number.
Generate the tree. Scria explores disconnections against its reaction knowledge base, and returns ranked routes, each fully expanded down to purchasable starting materials.
Compare what matters. Filter routes by step count, estimated cost, hazardous-reagent avoidance, or preferred chemistry classes
Act on the route. Source the starting materials through Scinode's supplier network, or hand the route to a manufacturing partner for quotation, without leaving the platform.
A retrosynthesis tool that ends at a PDF of routes still leaves you weeks of sourcing and quoting. A retrosynthesis tool wired into a live supplier and manufacturing network turns a route into an executable plan.
Conclusion
Retrosynthesis is being reshaped by AI. Computer-aided retrosynthesis delivers tangible benefits across pharma, agrochemicals, and academic research. Tools like Scria empower chemists to explore routes with speed and confidence, reducing the barriers to innovation. And because Scria lives inside Scinode, every route connects straight to sourcing, manufacturing partners, and compliance. This synergy of human insight and machine intelligence represents the future of chemical synthesis, and Scinode ensures that future is within reach for chemists and manufacturers everywhere.
Learn how Scria supports route design on Scinode. Sign up now.