Demonstrating a reduction in the number of synthesis steps when using AI driven predictive tools

Bringing small-molecule drug candidates to market has become increasingly complex, so how can this costly, time-consuming process be accelerated?

Using a predictive tool to identify shorter and more effective routes is a state-of-the-art way to save time in R&D and manufacturing costs and for Lonza , a pharmaceutical company, using Reaxys Predictive Retrosynthesis resulted in 17% fewer synthesis steps on average – see case study.

To understand how Lonza made this determination, here is a brief overview of the research.

The goal: Lonza was working on developing a Route Scouting Service to help their customers design synthesis routes for active pharmaceutical ingredients (APIs), which involves using its intellectual property and in-house expertise to evaluate a variety of factors, both scientific and commercial, and recommend the optimal synthetic routes.

The challenges: In establishing the service, Lonza faced two key challenges:

1. Molecular complexity and the growing number of synthesis steps
2. Increasing costs of raw materials and building blocks

Both make the synthesis of APIs more complex, increase spending on raw materials, and expand the time it takes to test and manufacture new drug candidates.

Drug companies strive to rule out failed candidates quickly, to fast track successful products to market, and to keep costs as low as possible — all of which are difficult with a complicated manufacturing process.

The solution: To overcome the challenges in synthetic route design, Lonza selected Reaxys Predictive Retrosynthesis to run a series of experiments. The goal was to determine if the application of this computer-aided synthesis tool could result in time and cost savings for their customers compared with other tools. And it worked!

Why did Lonza select Reaxys Predictive Retrosynthesis for their computer-aided synthesis planning test and how did it work?

The tool offers:

• High-quality reaction data
• Comprehensive literature references informing the synthesis routes
• Robust library of building blocks (starting materials)
• User-friendly and intuitive interface
• Flexibility to add in-house data

Reaxys Predictive Retrosynthesis combines high-quality reaction data with AI to deliver scientifically robust predictions. It deconstructs a target molecule, breaking its bonds to arrive at commercially available building blocks. Within minutes it provides multiple options for published and predicted synthesis routes that chemists can quickly take to the lab. Chemists can then reverse the direction and apply the transformations forward to build the target molecule in the lab.

The experiments: Lonza ran a series of experiments on molecules with different molecular complexity, in terms of molecular weight and chiral centers, using six different preclinical and Phase 1 targets. In separate tests, they used Reaxys Predictive Retrosynthesis and other cheminformatics tools and then compared the synthesis routes that the tools predicted.

The results: Reaxys Predictive Retrosynthesis resulted in 17% fewer steps on average. In the most dramatic difference, Reaxys Predictive Retrosynthesis suggested 11 fewer steps for a molecule containing nine chiral centers. In addition, by integrating its proprietary building-block library with the extensive set available in Reaxys, Lonza observed a significant reduction in the cost of the building blocks and synthesis.

What was the impact for Lonza?

Integration of their proprietary building-block library with Reaxys Predictive Retrosynthesis enables the identification of efficient and commercially viable synthetic routes from the start. By combining advanced route design with market intelligence, Lonza avoids late-stage redesigns and delays, accelerating the development process and ensuring a robust manufacturing pathway. This approach helps deliver new therapies to patients faster and more efficiently.

To learn more details about the experiments and results, as well as the impact Reaxys Predictive Retrosynthesis has had for Lonza, read the full case study here.