Custom SYNTHIA Search Configuration Cuts Synthesis Route Planning to Hours, Case Study Shows
核心洞察
A case study from Merck KGaA shows that a custom SYNTHIA (搜索) retrosynthesis search configuration can prioritize reactions executable on the Synple (搜索) automated synthesizer.
Across ten representative molecules, the Synple (搜索) configuration returned shorter routes with more automatable steps than the General configuration, which favored low-cost starting materials.
For test compound C6, the automated Synple (搜索) route required one hour of hands-on time versus six hours for the manual Discovery route, with comparable yields of 46% and 49%.
Synthetic chemistry remains a demanding, time-consuming discipline in which route scouting and reaction optimization depend heavily on experienced chemists, creating opportunities for error and bias while limiting scalability. A new case study from Merck KGaA demonstrates that tailoring retrosynthesis search parameters to a specific automated synthesis platform can compress both route design and bench execution, potentially easing a persistent bottleneck in therapeutic discovery.
The work combines SYNTHIA (搜索) retrosynthesis software, which uses expert-coded rules to predict feasible routes from commercially available starting materials, with Synple (搜索)'s automated cartridge-based synthesis technology. Synple performs many reactions commonly used in discovery chemistry, allowing scientists to initiate and complete a reaction, work-up and product isolation with the press of a button. Interfacing the two tools was achieved by implementing a specific set of search parameters in SYNTHIA so that it identifies routes executable automatically on the Synple platform.
Comparing Three Search Configurations
The study compared two preset SYNTHIA (搜索) configurations — General and Discovery — with a third custom configuration, Synple (搜索), built on the Discovery setting to promote reactions that can be executed on the automated system. General balances step count against the cost of starting materials, while Discovery prioritizes shorter routes over inexpensive starting materials for chemists who need rapid access to small amounts of a target.
A test set of ten representative molecules drawn from known therapeutic candidates and a drug-like library was uploaded to SYNTHIA (搜索) as a single file using the Batch Retrosynthesis Module. Each retrosynthetic evaluation required between 9 and 16 minutes. For each configuration, the highest-ranked pathway was scored for total synthetic steps, including any necessary protection or deprotection, and for the number of steps compatible with Synple (搜索) execution.
As anticipated, the General configuration typically identified pathways with the highest number of steps using lower-cost starting materials, whereas Discovery and Synple (搜索) prioritized fewer-step routes. Detailed evaluation of the top pathways revealed two notable differences: Discovery frequently used more challenging or exotic reactions, while Synple returned steps mostly using routine transformations performable on the Synple system. Because the Synple configuration was engineered to promote protection/deprotections available as Synple cartridges, such as Boc (搜索), its routes were more likely to use commercially available protected starting materials and to include exact deprotection steps. Discovery routes more often included general protection/deprotection steps from unprotected starting materials, giving users flexibility to choose among compatible protecting groups.
Neurokinin Antagonist Example
The pattern is illustrated by C3, a neurokinin antagonist (搜索) candidate. The General configuration proposed a six-step route beginning with more cost-effective starting materials and including just one step automatable via Synple (搜索). The top Discovery and Synple routes each contained five total steps, but only one was potentially automatable in the Discovery route, while all five steps could be automated in the Synple route. The Synple route identified a Boc (搜索)-protected starting material and included Boc deprotection as an explicit step, whereas the Discovery route began from an unprotected starting material. The more common transformations proposed in the Synple route may also reduce the likelihood of failure and minimize the need for optimization.
Users are not locked into fully automated routes: even when searching with the Synple (搜索) configuration, they can choose from 50 possible routes containing both automatable and manual steps. The shortest route to compound C8, for example, involved two Synple steps and one manual step, illustrating how an optimal route can combine the strengths of the chemist and the machine.
Bench Validation and Yields
To assess the recommended routes, a test molecule, C6, was synthesized using both the Discovery route (manual synthesis) and the Synple (搜索) route (automated synthesis). The Discovery route required approximately six hours of hands-on working time, excluding reaction time during stirring, and its initial alkylation step required additional scouting for different conditions due to the reactivity of the alkyl halide. The Synple route required just one hour of working time to configure the two steps on the machine, followed by solvent evaporation after the automated reaction.
Both routes produced comparable overall yields, 49% versus 46%. However, the product from the Synple (搜索) route was sufficiently pure at 90%, eliminating the need for further purification. The authors concluded that C6 was accessed in just a few hours of working time with no drawbacks compared with the traditional method.
Broader Benchmarking Across Chemical Industries
A forthcoming large-scale evaluation will benchmark SYNTHIA (搜索) across pharmaceuticals, agrochemicals, and flavours and fragrances, using datasets representative of each sector and a library of 14 million commercially available chemicals, drawing on more than 120,000 expert-curated reaction steps. That assessment will examine success rates and the distinct challenges each industry presents, including the complexity of pharmaceutical molecules, the scalability requirements of agrochemicals, and the sensory characteristics of fragrance compounds.
Molecular complexity is a key factor in whether software delivers practical routes. Molecules with multiple chiral centers pose significant challenges because of the precise control required over stereochemistry, and synthetic accessibility — how difficult a molecule is to make from commercially available building blocks — also plays a crucial role. Metrics such as the synthetic accessibility score are being examined for their correlation with the likelihood of finding a feasible pathway. Shorter routes are particularly valued for cost efficiency and sustainability, with retrosynthetic software sometimes designing one-, two- or three-step syntheses from widely available building blocks.
According to information shared by Chemistry World, Angelo Lanzilotto, a chemist now contributing to Merck KGaA's digital chemistry division, and his team are actively developing and commercializing software solutions tailored to chemical industries, positioning retrosynthetic tools as integral components of the modern chemist's toolkit rather than mere aids. The case study authors note that because the Synple (搜索) automatable chemistry configuration was adopted in SYNTHIA (搜索) with relative ease, the application can be extended to prioritize any subset of reactions individual users prefer, positioning SYNTHIA as a tool for uniquely tailored, focused retrosynthetic evaluation that reduces manual effort in route planning, starting-material searches and bench execution.