On the hunt for cures
In November 2018, the U.S. Food and Drug Administration approved the drug Vitrakvi discovered by Array BioPharma and developed by Loxo Oncology. Vitrakvi is used for the treatment of patients with solid tumors that have a neurotrophic receptor tyrosine kinase (NTRK) gene fusion. It was the first small molecule cancer drug to be approved that treats people based on their genetic profile, rather than the location of their tumor. Vitrakvi is noteworthy because it originated from a new area of chemical space that had not been previously explored until recently.
Vitrakvi is a great example of how drug hunters are pushing the boundaries of the known chemical space by searching previously unexplored regions to find new drugs. Today, many researchers are hunting for cures through a more systematic exploration of the chemical universe. New technologies and approaches are making drug discovery far more proactive, saving time and resources, and allowing cures to get to market faster.
Exploring the vast expanse of chemical space
An analysis of CAS REGISTRYSM shows that the pace of innovation is accelerating as pharmaceutical companies race to discover new drugs in the uncharted regions of a dauntingly large chemical cosmos.
CAS REGISTRY, a curated collection of all known and studied chemical substances, has more than doubled in size over the last ten years and currently stands at about 150 million unique substances. However, it has been estimated that there are 1060 possible molecular combinations with drug-like characteristics in the chemical universe. Such a large number is difficult to wrap our minds around. By comparison there are an estimated 1024 stars in the known universe.
The reality is, only a tiny sliver of the chemical universe is known to us. The vast majority of space is yet to be explored and represents a tremendous opportunity for those seeking cures for our world's most challenging diseases. But the question is: how do we efficiently navigate this chemical space to find them?
The potential of AI to transform drug delivery
Business leaders in the pharmaceutical industry have recognized the transformative power of artificial intelligence (AI) to further their drug discovery endeavors. Leading companies have already started using AI applications to scour the vast reaches of chemical space, searching through hundreds of millions of virtual compounds to find potential candidates for development. In the future, it is expected that this will be a standard part of drug discovery processes.
However, the development and growth of AI is not without its challenges.
Clean data is key to improving AI success
Currently, data scientists spend as much as 80% of their time sourcing and cleaning up the data needed to feed their algorithms, rather than building the models themselves. One problem they face is finding clean, relevant data sets to train their AI and machine learning algorithms. There is plenty of raw data available, but it needs to be refined, translated, structured and indexed in order to be useful.
This highlights the importance of the non-artificial component in AI technology. The human element. AI systems operating in tandem with humans can accomplish more than either could on their own. Researchers use their cognitive abilities to provide clean, relevant data and feedback while the AI performs the repetitive, tactical functions far more quickly than a human is able. By reviewing vast amounts of data quickly, AI can identify patterns and anomalies and provide researchers with information they may never have uncovered otherwise.
CAS molecular descriptors accelerate research
CAS has developed a patent-pending approach that employs its proprietary molecular descriptors along with a variety of AI techniques to predict a molecule's biological activity. It does this by first reviewing all known molecules that interact with a target of interest. It then deconstructs those molecules to identify exactly which parts or fragments are shared across active molecules. Finally, it searches the chemical cosmos for virtual compounds containing these components to identify molecules that are most likely to have the desired biological activity.
This method greatly reduces the number of compounds that need to be synthesized and assayed in the laboratory to identify new drug candidates.
The CAS descriptors provide an abstract representation of certain structural features of a given molecule at a more granular level than other commonly used molecule fingerprints. In a series of validation tests, the accuracy of the CAS descriptors outperformed common machine learning algorithms utilizing the Morgan fingerprint by up to 45%.
Large pharmaceutical companies with their own AI platforms benefit from utilizing the CAS molecular descriptors to give them more accurate and detailed results. Smaller companies with limited AI capabilities can take advantage of the data, technology and expertise CAS offers to augment their discovery process. The goal is to increase research efficiency and return on R&D investment by pointing to specific regions within the chemical cosmos that show the most promise for their targets of interest.
For over 100 years, CAS scientists have been curating the most robust and detailed collection of chemical information available from patents, journals and other scientific publications around the world. This experience and dedication gives us a unique expertise and talent pool, including hundreds of scientists and technologists, perfectly suited to supporting success in this space. Contact us to discuss how partnering with CAS can provide the data, technology or expertise you need to move your discovery engine into high gear.