When Perfect Molecules Fail: How Retrosynthesis Predicts Success

From minor aches to major ailments, there is a vast variety of medicines for helping us heal. But did you ever wonder how much research goes into every single drug we take? Developing new drugs is a long and expensive process. It takes an estimated $2.8 billion and up to 15 years to take a new drug through to regulatory approval ^{[1] [2]}.

The overwhelming majority of today’s pharmaceuticals are ‘small molecule’ compounds, made of a few hundred atoms or less. Between 2010 and 2020, 76% of new drugs approved by the US Food and Drug Administration (FDA) were small molecules ^[3]. Compared to larger biological therapies, the extra-small size of these drugs makes it easier for them to enter cells and reach targets in the body.

The diverse compounds in this group share little in common except for their size and the fact that they’re made from synthetic chemical reactions. A popular example is aspirin – one of the oldest and most widely used medicines in the world. Many of the new cutting-edge drugs for treating illnesses, such as cancer, autoimmune diseases and depression, are also small molecules. Currently, countless more are being developed to prevent and treat different conditions.

At the heart of small molecule drug discovery is chemical synthesis, a complex, laborious  process for creating new molecules – and a key bottleneck for advancing new medicines to the clinic. But recent developments in AI-based software offer unprecedented opportunities for accelerating and facilitating this stage of drug discovery.

In organic chemistry, “forward synthesis” starts from known reagents and asks, “What do I get if I react A with B under these conditions?” Retrosynthesis turns this question around. You start from the target molecule and work backwards, conceptually “cutting” the molecule into simpler fragments step by step until those fragments are simple, stable and commercially available.

Formally, retrosynthesis (or retrosynthetic analysis) is the systematic process of breaking down a complex structure into a sequence of simpler precursors that can realistically be made or bought. A typical question is:

“Which bond would I ideally form last – and what must the two building blocks on either side of that bond look like?”

Once you have those two precursors, you ask the same question again for each of them, continuing until you reach known chemistry and accessible materials.

Ethyl acetate (C₄H₈O₂), a common ingredient in nail polish remover, can be broken down retrosynthetically into ethanol (C₂H₅OH) and acetic acid (CH₃COOH). The key step is:C₄H₈O₂ (ethyl acetate) ← C₂H₅OH (ethanol) + CH₃COOH (acetic acid)

Both ethanol and acetic acid are simple, commercially available chemicals, so this marks the end of the retrosynthetic path. In the lab, the forward synthesis mixes these two under acidic conditions to form ethyl acetate and water – a reaction called esterification.

Doing retrosynthesis well demands experience – a feel for which disconnections tend to work in real flasks, not just on paper. It requires deep knowledge of the literature: which reactions, catalysts and conditions have worked before, and under what constraints. It also involves judgement, balancing elegance against practicality, cost, safety and scalability. In practice, many of these decisions are made under time pressure with incomplete information. That is how you end up with 14-step routes to molecules that, in hindsight, might be made in 6 steps with a different strategy.

Retrosynthesis is becoming more challenging because modern drug molecules are more complex – like PROTACs (which recruit proteins for degradation), molecular glues (which stabilize protein interactions), covalent inhibitors, and multifunctional linkers. At the same time, projects must move faster and cheaper, with smaller teams, while considering sustainability factors such as waste generation (E-factor), atom economy, solvent use, and energy consumption. These pressures make trial-and-error forward synthesis costly, so structured retrosynthetic planning is essential.

A good retrosynthetic plan looks beyond just the number of steps. It considers whether starting materials are commercially available and scalable, ensures the correct 3D arrangement of atoms (stereochemistry), minimizes extra steps like protecting groups that add complexity and waste, and evaluates whether reactions can be safely and efficiently scaled from milligrams to kilograms. Balancing all these factors is complex, which is why retrosynthetic planning has traditionally been a specialized skill.

In recent years, software has begun to support this kind of backwards thinking. These tools encode large numbers of reaction rules – general patterns of how bonds can be formed or broken – coded based on millions of examples from publications and patents. They can very quickly explore thousands of possible ways to cut a target molecule into simpler pieces.

One example of such a tool is SYNTHIA® Retrosynthesis Software from our Life Science business. A user can input a target structure, for instance as a SMILES string, an InChI code, a molfile or a hand‑drawn structure. The software then analyzes the molecule and proposes alternative synthetic routes. These can be ranked and filtered by criteria such as step count, similarity to published syntheses, cost of starting materials, or protecting group requirements, and many more. 

For each proposed route, the program can also provide suggested reaction conditions for each step, identifiers such as CAS numbers, sustainability-related information, and catalog links for proposed building blocks.

SYNTHIA^® Retrosynthesis Software does not remove the need for experimental work, but it shortens the path from concept on the screen to a realistic plan for lab implementation.

The software is a powerful tool, but users must recognize its limits. It does not prove that a route will work but instead suggests plausible options. It does not replace chemical judgement but simply changes how that judgement is applied. Tools like SYNTHIA® are amplifiers of human-powered retrosynthesis. 

The tool helps teams surface non‑obvious disconnections, or the process of identifying which chemical bonds need to be broken within the molecule to move forward. This isn’t always obvious, and it allows chemists to explore “what if” scenarios like avoiding specific reagents or solvents in synthesis. When experienced chemists partner, challenge, refine and reject the suggestions, the best results come.