One of the overlooked areas where artificial intelligence (AI) and machine learning (ML) are showing promise is in life sciences instrumentation. Clearly, AI/ML models hold enormous potential for increasing the speed, accuracy and reliability of drug development, disease research and more.
Although use cases for AI vary across the pharmaceutical industry, 59% of pharma and biotech companies cited drug discovery and development as a primary driver of adoption. AI can be extraordinarily useful for automating and optimizing tasks which have been historically manual in nature.
In the rapidly evolving field of structural biology, cryo-electron microscopy (cryo-EM) enables scientists to better understand the intricacies of life, providing atomic-level insights that ultimately help show how cells, proteins and viruses work. Companies and organizations like Thermo Fisher are embarking on an exciting journey that combines this leading-edge instrumentation with AI and ML so that users can now advance research in ways that were previously impossible.
Making cryo-EM more accessible
Since the “resolution revolution” began more than a decade ago, biologists have turned to cryo-EM to get 3D, near-atomic resolution structural images of macromolecules. Cryo-EM has enabled the structure determination of many classes of proteins which were previously intractable and is having a profound impact on the development of small molecule drugs and biotherapeutics, with numerous molecules designed with the help of cryo-EM currently in clinical trials. The figure above illustrates notable public disclosures for clinical stage assets aided by cryo-EM.
Cryo-EM has often required a certain level of expertise, which is gained through extensive training and past experimentation. Scientists need to know the quality of the images they’re acquiring, how to interpret the data and when to pivot. This presents a unique challenge, but also an opportunity for AI and ML.
Solving the data acquisition conundrum
Single particle analysis (SPA) is one of the most common cryo-EM techniques, but the SPA data acquisition workflow is still largely manual with a number of tedious tasks. In the SPA workflow, protein samples are deposited in a thin layer on grids and rapidly frozen. These grids are then placed in the microscope for imaging where users must manually select suitable areas on the grid for imaging. This process is time-consuming and often considered one of the main bottlenecks of SPA. Streamlining data acquisition in cryo-EM is a primary focus for much of the cryo-EM community — both academia and industry.
The New York Structural Biology Center’s Workshop on Smart Data Collection highlighted the need for automated data collection, especially for SPA cryo-EM, and many in the community are looking for ways to optimize with AI and ML. The incorporation of AI-based models within data acquisition software, for example, can automatically select the best areas on a grid for imaging, which dramatically improves ease of use and throughput.
At the end of the day, efficient and user-friendly data collection will continue to be fundamental for scientific discovery. AI and ML are are paving the way for significantly improved performance and throughput in microscopy, which will lead to a healthier society.
On the precipice of next-generation microscopy
The cryo-EM community, along with life science technology providers, shares a common goal to make cryo-EM more accessible, and many are fully embracing AI and ML to turn a once-distant dream into a tangible reality. There’s a renewed sense of excitement around the potential for accelerating science at a pace never seen before, and we know that AI is a piece of the puzzle.
Today’s rapid technological development is leading to increased automation and better data acquisition, which means that scientists can use cryo-EM to explore new frontiers in cellular and structural biology. While a frictionless workflow is still the ultimate goal, the industry moves a step closer every day. It’s a summit attempt backed by a dedicated community of scientists and innovators, and reaching this new height will allow many more researchers to leverage Cryo-EM in their research and accelerate the pace of their scientific discovery.
Edward Pryor is a senior director of product management and application development Thermo Fisher Scientific and holds a Ph.D. in Biochemistry and a MS in Computer Science. He has over 20 years of experience in protein structural biology and computational biology. In his current role, Edward and his team focus on improving the complete end-to-end cryo-EM workflow, which includes sample preparation, data acquisition, data processing, and data management.
Filed Under: machine learning and AI
