Time is money; this holds particularly true for drug research and development. The global pharmaceutical industry spends about $100 billion a year on R&D. Researchers are, therefore, always looking for faster ways to synthesize molecules, so they can cut time from initial discovery and patenting to the launch of a new drug. ChemPlanner is a new cheminformatics software tool targeted at organic chemists for predicting synthetic routes to target molecules. The tool, developed by Wiley and launched in September 2015, is currently sold as Software as a Service (SaaS) solution and hosted on the publishing company’s servers, with a local installation version coming this year. Dr. Michael Reubold speaks with Dr. David Flanagan, Director, Lab Solutions at Wiley, about how ChemPlanner can assist pharma researchers and what differentiates the product from other reaction prediction tools.

CHEManager: Dr. Flanagan, what are the key features of ChemPlanner and who can benefit from it?

D. Flanagan: ChemPlanner is a reaction prediction tool based on machine learning for organic chemists. A chemist can draw their target molecule and ChemPlanner will predict the shortest, fastest, cheapest route to the target, even through predicted, never-before-reported reactions, based on what it has learned about organic chemistry.

ChemPlanner has three key features. First, it boosts your productivity. ChemPlanner reduces literature researching drudgery and cuts down on planning time so that you can synthesize more molecules and complete more projects faster. Second, it boosts your creativity. ChemPlanner can suggest routes you wouldn't have necessarily considered and unlocks ideas for new routes. Finally, ChemPlanner is easy to use. We've spent a lot of time working with chemists to design the interface and the user experience so that ChemPlanner fits well into researchers’ workflows.

How does ChemPlanner work, what data is it based on?

D. Flanagan: ChemPlanner works by combining industry-leading cheminformatics software with high-quality reaction data. ChemPlanner analyzes a database of reactions to isolate individual reaction cores that it can then cluster into rules. It is these rules, automatically extracted and clustered by algorithms into reaction classes that a chemist might recognize, for example as a particular name reaction. When ChemPlanner analyzes millions of reactions, it can derive rules that basically cover all of organic chemistry. ChemPlanner currently uses the ChemInform Reaction Database (CIRX) to generate its rules knowledgebase, and in the future customers will be able to derive rules from their own reactions as well. While not necessary, this will allow you to use a customized version of ChemPlanner with focused coverage in areas of chemical space especially relevant to your particular targets.

What makes ChemPlanner particularly interesting for pharma companies?

D. Flanagan: ChemPlanner is useful for anybody who synthesizes organic molecules, including industries like pharmaceuticals, fine chemicals, CROs, agrochemicals, flavors and fragrances, and biotech. But, it is particularly important for the pharmaceutical industry to be able to make more molecules more efficiently. If you consider that it takes 10 to 12 years for a drug to move from prediscovery to FDA approval, any product that could potentially decrease that time would be very valuable. Not only would patients have access to life-improving drugs earlier, but also the companies would have their molecule in the market while under patent protection longer.

Can you roughly estimate the cost saving potential ChemPlanner is offering for pharma companies?

D. Flanagan: It is hard at this point to estimate the value or cost savings that a product like ChemPlanner could deliver, but we can consider the value that pharmaceutical companies place on shortening the time from discovery to FDA approval. For example, the FDA will give you what's called a Priority Review Voucher if you develop a new drug for neglected or pediatric diseases. This voucher will let you fast track your FDA application for a new drug, taking four months off the process. These vouchers are tradable, with the last voucher sold for $350 million. From this, you can calculate that taking one month off a 12 year process could be worth up to $87 million for the right pharmaceutical company. We think that ChemPlanner has the potential for this kind of time savings.

What differentiates ChemPlanner from other tools like Reaxys or SciFinder?

D. Flanagan: Today, chemists tend to use A&I services like Elsevier’s Reaxys or SciFinder from CAS to plan synthetic routes. However, these products are used for straight database lookups, and leave most of the mental work planning the synthesis for the chemist to do. ChemPlanner does two key things standard reaction database services don't offer. First, ChemPlanner will analyze all of the possibilities and predict the shortest, fastest, cheapest, most efficient overall route to a target molecule, instead of making the chemist piece together individual reactions and hope that they find the overall best route. It is hard for a person to keep track of all of the variables involved in planning a route, such as number of steps, reliability of the reaction, cost and availability of starting materials, and interference of different functional groups, in their head. But for a computer, finding the globally optimal combination of these variables from millions of possible solutions is relatively easy. In addition, since ChemPlanner in effect “knows” all of organic chemistry, it can predict reactions that haven't been reported in the literature and captured in databases.

Here’s one way to think about the power of predicted reactions. There are about 100 million (10⁸) molecules registered in the CAS database. But, the number of possible small molecules is calculated to be somewhere between 10⁵⁵ and 10⁶⁵. That means that only a vanishingly small number – 10^-47 percent – of possible molecules have been reported in the literature and databased. If we take the number of molecules captured in the CAS database as a proxy for the number of reactions captured in databases, we can also assume that a vanishingly small number of possible reactions have been captured in databases. This is why we think ChemPlanner’s rule-based learning approach that can cover significantly more chemical reaction space is preferable to a simple reaction database approach.

What kind of feedback have you got from researchers so far?

D. Flanagan: ChemPlanner 1.0 launched in September 2015, and we’ve already had a lot of positive feedback from customers who are using it for their own synthetic targets. They like the user interface, they like the quality of her predictions that our cheminformatics algorithms produce, and they like that we have a solid platform and a roadmap for further improvements and developments for the product. Our customers include some of the world’s top 10 pharma companies, and we are looking forward to more chemists at more companies adopting ChemPlanner as a tool that they can use every day to improve their productivity and be more creative. We’ve shipped two updates since then incorporating customer feedback and suggestions.

A global top-ten pharmaceutical company has purchased the first ChemPlanner license. What are your expectations for the adoption of ChemPlanner on the market?

D. Flanagan: This specific customer has purchased a group license and will use it for process chemistry, to reduce the cost of producing a molecule for clinical trials, as well as for discovery chemistry. The company has been an early adopter of computer-aided synthesis design technology, and was one of the first customers of ARChem, which was the predecessor to ChemPlanner. They particularly like ChemPlanner’s user interface, the quality of the predictions that our cheminformatics algorithms produce, and that we have a solid platform and a roadmap for further improvements and developments for the product. We think there is scale for sales to grow, both horizontally as we increase the number of users and sites at a company using ChemPlanner, as well as vertically as we monetize new features and capabilities on our roadmap.