Ian Ferria, CPO Artificial Intelligence, discusses how AI can revolutionize drug discovery for pharmaceutical companies.
Technological progress has given rise to a complex of startups exploring AI for drug discovery. Much of the research entails using AI to identify patterns in large volumes of data. Fruitful use cases have applied “narrow AI” to drug discovery, focusing on specific tasks such as analysis, understanding, and creating text and speech via natural-language processing, as well as artificial neural networks mimicking the way our brains sense the world.
While the AI visionaries of the 1950s envisioned machines that could sense, reason, and think just like people, that version of AI remains in the realm of science fiction. Nevertheless, the improvement of computer-processing power over the past several decades, the preponderance of large and easily accessible data sets, and the development of advanced algorithms have driven major improvements in machine learning and AI.
Drug discovery researchers have applied these methods to their work. Proponents claim AI robots test more compounds with better accuracy, reproducibility, and exhaustive, searchable record-keeping than their human counterparts.
For instance, on June 12, 2007, a robot called Adam identified the function of a yeast gene by searching public databases. Adam the robot drafted hypotheses about which genes code for key enzymes that catalyse reactions in the yeast Saccharomyces cerevisiae, and tested the predictions in a lab1.
When researchers at the UK universities of Aberystwyth and Cambridge independently tested Adam’s hypotheses regarding the functions of 19 genes, they found nine were novel and accurate. Only one proved wrong.
Later on, Adam’s more advanced robot colleague, Eve, which had been developed by the same team, discovered that triclosan, commonly used in toothpaste, could treat drug-resistant malaria parasites. Eve screened thousands of compounds to find those that halted or considerably slowed the growth of strains dependent on the malaria genes and which did not contain human genes.
Adam and Eve represent two examples of AI and machine learning used toward improved efficiency in drug discovery. Many others exist as well2.
For instance, Pfizer has deployed IBM Watson to leverage machine learning in the search for immuno-oncology drugs3. Sanofi has used artificial intelligence to uncover metabolic disease therapies4. Roche subsidiary Genentech has used AI to discover cancer treatments5.
AI and machine learning, particularly when combined with automation, might hail the coming of a quicker, cheaper, and more-effective drug discovery process.
AI platforms can look for drug targets and therapies with data from various sources, such as research papers, clinical trials, and patient records. Furthermore, an AI system can query such data and produce knowledge graphs of a medical condition. Drug-discovery AI should theoretically put this data into context and highlight the most notable information.
AI’s power of pattern recognition could transform the drug discovery process forevermore, leading to a fuller understanding of human biology and improved tools to address human disease. Over time, personalized medicine, whereby medical decisions are tailored to individual patients, could be enabled by advancements in AI.
The fields of biology and medicine are being transformed by AI. If these processes prove to create better drugs in a less expensive and expedient manner, AI would secure itself as the future of drug discovery.
Challenges of Drug Discovery With AI
The hope is AI for drug discovery can overcome specific challenges. As the NIH reports6, there are presently several drug discovery hurdles, including, but not limited to, unknown biological mechanisms and biomarkers of diseases; translational failures in animal models; a paucity of clinical phenotyping and patient stratification; no reliability in published data; and an inadequate collaboration among academia, industry, and government, as well as pipeline challenges.
AI faces significant data challenges in drug development, which entails data sets containing millions of compounds. The scale, growth, diversity, and uncertainty of the data could render traditional modeling tools unable to handle the data7.
In the pursuit of better drug discovery, Core Scientific is addressing pipeline and modeling challenges with the Cloud for Data Scientists, a platform designed by data scientists for data scientists, empowering them to focus on model development, testing and deployment, without worrying about infrastructure. With seamless integration into on-premises and public cloud infrastructure, and a cloud-like, easy-to-use UI, the Cloud for Data Scientists can get Data Scientists training their models in under 10 seconds.
What Is The Cloud For Data Scientists?
The cloud is much more than a service to rent a computer. It can make the lives of data scientists much easier. When they’ve built a data science algorithm, wrapped the code in a Docker container, and want to deploy it, the last thing they want is to host it at home on a single machine, which acts as a single point of failure, and lacks the scalability needed when demand increases.
In order to avoid such headaches, the data scientist has a couple of options. They can create a cluster architecture to turn their machines into one computing entity. That way, if one machine fails, no problem, as your model will move to a different part of the cluster. But, it is a painstaking exercise to set up such an architecture.
Alternatively, the data scientist can employ a cloud provider to instantly create a cluster to host their model, saving time and hassle. The data scientist can then focus on developing the best algorithms with state-of-the-art technologies and short set-up time, instead of IT and maintenance.
That’s where Core Scientific’s Cloud For Data Scientists comes in. This proprietary data science platform has been built specifically for AI and High Performance Computing using the best-in-class hardware components.
Our software, furthermore, has the ability to run on premise, so important data does not have to leave the facilities. Knowing your data is located within your in-house servers and IT infrastructure provides peace of mind. Core Scientific’s Plexus and Cloud For Data Scientists software provides an on-premise option which provides the greater protection that the pharmaceutical industry demands. For these companies, our software makes perfect sense.
If you’re ready to take your data science to the next level, look no further: take a test drive of the Cloud for Data Scientists.