Many of the “easy” targets are already served by multiple marketed drugs and, as a result, the molecules being designed today are becoming ever more complex.
This is often required to modulate targets that are considered difficult or, potentially, even undruggable. The increase in molecular complexity introduces greater production challenges.
It is not uncommon for synthetic routes to be 15–20 steps long or even more. The net result is lower overall yields, longer manufacturing times and increased costs. So, what’s being done to improve the situation?
Increasingly, the answer lies in applying in silico tools to the problem.
Back to the future
Retrosynthetic analysis was first developed by Harvard chemist and Nobel Prize winner E.J. Corey.
It’s a little like taking a complex LEGO model apart and then putting the pieces back together again to remake the model. The chemist considers the complicated molecule and imagines how it might be assembled from smaller fragments.
What chemical reactions could be used to make the bonds needed to put them back together again? And how could you break those fragments down into even smaller pieces?
This retrosynthetic process is repeated until all the pieces of the jigsaw — or model — are available. There are many ways that a large and complicated molecule might be broken down into fragments.
And therein lies the real challenge: which way is likely to be best? Which has the fewest steps? Which uses easy, safe and high-yielding reactions? Which minimises the number of purification steps? And, crucially, which is likely to be quickest and cheapest to run at scale?
Although the concept of retrosynthetic analysis has been routinely used for decades now, the application of computer techniques to the task has accelerated in recent times.
Corey realised early on that such an approach might help and more recent advances, such as computer-aided synthesis planning (CASP) tools, have expedited the process.
There are simply too many options to consider for process chemists to be certain they have identified the most likely synthetic route to make the molecule at commercial scale.
CASP tools can take information from large, curated reaction databases and use them to calculate possible routes back to simple starting materials. But limitations remain.
Significantly, whereas the tools may be able to search catalogues of research chemicals for potential starting materials, these information reservoirs provide no insight into what’s available in commercial quantities.
What the CASP tool might think looks like a great synthetic route — because it minimises steps and has a high yield — could well be completely impractical at scale.
The materials could be sourced, but if it’s not available from stock and there is a long lead time, the only options are to wait … or to add further steps to make it.
Even if an intermediate is readily available, it might be prohibitively expensive. With inventory issues continuing to burden the chemical supply chain, this problem is not trivial.
Data-driven approaches that include real-world commercial supply chain insights offer the potential to improve synthetic route planning, thereby creating feasible and cost-effective routes more quickly.
A path to better routes
To address these issues, Lonza Small Molecules has developed an AI-enabled Route Scouting service. This provides customers looking to make a new API with a rapid way to identify commercially viable synthetic routes.
It combines advanced computational tools for route design with data that give an insight into real-world supply chain availability. The aim is to find synthetic pathways that are optimised for commercial production at an early stage of the development process.
It is this integration of supply chain data, the technical knowledge and experience of experts that differentiates the Lonza service from commercially available CASP tools.
The unique advantage, in addition to their raw material sourcing analysis, is that they prioritise scalable synthetic routes — something often overlooked by AI tools. AI-based predictive chemoinformatics do a great job of proposing routes that may be synthetically possible.
What these methods cannot do, however, is triage them effectively from a commercial perspective.
By constraining the system to use Lonza’s vast databases containing the cost and availability of thousands of chemical building blocks, intermediates and basic chemicals, the automated system can produce routes that are more viable at a commercial scale.
With the input of skilled process chemists, it will pinpoint those that are most likely to be successful. Partly because of the vast amounts of accumulated data, Lonza’s decades of API manufacturing experience is invaluable.
It facilitates the ability to make informed predictions about the future prices of different raw materials based on historical data and trends.
All these data points are considered, along with the options proposed by the predictive retrosynthetic design tools, to add a commercial perspective to the route selection process.
The next step involves human input from Lonza process experts. Their knowledge allows them to rank potential reactions according to their synthetic viability and determine potential challenges.
Ultimately, they will identify a route that ticks all the boxes. The top-ranked route should balance commercial and synthetic viability.
Supply chain insights
Data about the commercial supply chain is key to determining the viability of the various routes. Pricing is only part of the story; it also covers the global supply situation for each required chemical, including who makes and sells them and total global production volumes.
Manufacturing decisions can then be made that are informed by real-world data. Combined with the sophisticated tools used to pinpoint potential synthetic pathways, optimised routes that are less likely to face complications during scale-up can be identified.
There are multiple further benefits to this approach. What might look like the best option because it has the fewest steps may prove problematic from a logistical perspective or require more costly building blocks.
Lonza’s AI-enabled Route Scouting service helps to create a right-first-time approach. This is equally valid for new chemical entities and for established APIs for which improved synthesis would be beneficial for commercial reasons.
The potential cost savings are another important factor. Drug developers and manufacturers will clearly feel the economic benefits of faster and more cost-effective syntheses; tools such as these can have significant financial benefits.
These can then be passed on to patients and health authorities by lowering drug product prices. It might also be a game-changer for small biotechs. These companies are reliant on hitting milestones to secure the funding they need to progress their investigational drugs through the different stages of development.
Anything that can help to progress the product through the pipeline as quickly as possible will clearly help to move those important milestone payments closer.
Retrosynthetic analysis has advanced dramatically since E.J. Corey’s days.
Modern CASP tools relying on big data have enhanced its power and now underpin the process of finding good potential synthetic routes.
By combining CASP technology with real-world market intelligence, Lonza can find the best ways to make individual molecules as quickly and cost-effectively as possible, accelerating their pathway into the clinic and to the patients who need them.
