Leland Hyman is the Lead Data Scientist at Sherlock Biosciences. He’s an experienced computer scientist and researcher with a background in machine learning and molecular diagnostics.
Sherlock Biosciences is a biotechnology company based in Cambridge, Massachusetts developing diagnostic tests using CRISPR. They aim to disrupt molecular diagnostics with higher, faster, inexpensive tests.
What initially attracted you to computer science?
I began programming at a really young age, but I used to be mainly fascinated about making video games with my friends. My interest grew in other computer science applications during college and graduate school, particularly with all the groundbreaking machine learning work happening within the early 2010s. The entire field gave the look of such an exciting recent frontier that would directly impact scientific research and our each day lives — I couldn’t help but be hooked by it.
You furthermore mght pursued a Ph.D. in Cellular and Molecular Biology, when did you first realize that the 2 fields would intersect?
I began doing this sort of intersectional work with computer science and biology early on in graduate school. My lab focused on solving protein engineering problems through collaborations between hardcore biochemists, computer scientists, and everybody in between. I quickly recognized that machine learning could provide precious insights into biological systems and make experimentation much easier. Conversely, I also gained an appreciation for the worth of biological intuition when constructing machine learning models. In my opinion, framing the issue accurately is the crucial element in machine learning. Because of this I consider collaborative efforts across different fields can have a profound impact.
Since 2022 you’ve been working at Sherlock Biosciences, could you share some details on what your role entails?
I currently lead the computational team at Sherlock Biosciences. Our group is accountable for designing the components that go into our diagnostic assays, interfacing with the experimentalists who test these designs within the wet lab, and constructing recent computational capabilities to enhance designs. Beyond coordinating these activities, I work on the machine learning portions of our codebase, experimenting with recent model architectures and recent ways to simulate the DNA and RNA physics involved in our assays.
Machine learning is on the core of Sherlock Biosciences, could you describe the variety of data and the quantity of knowledge that’s being collected, and the way ML then parses that data?
During assay development, we test dozens to tons of of candidate assays for every recent pathogen. While the overwhelming majority of those candidates won’t make it right into a industrial test, we see them as a possibility to learn from our mistakes. In these experiments, we’re measuring two key things: sensitivity and speed. Our models take the DNA and RNA sequences in each assay as input after which learn to predict the assay’s sensitivity and speed.
How does ML predict which molecular diagnostic components will perform with the best speed and accuracy?
Once we take into consideration how a human learns, there are two major strategies. On one hand, an individual could learn do a task through pure trial-and-error. They may repeat the duty, and after many failures, they’d eventually determine the principles of the duty on their very own. This strategy was pretty popular before the web. Nonetheless, we could provide this person with a teacher to inform them the principles of the duty instantly. The scholar with the teacher could learn much faster than with the trial-and-error approach, but only in the event that they have a superb teacher who fully understands the duty.
Our approach to training machine learning models is partway between these two strategies. While we don’t have an ideal “teacher” for our machine learning models, we will start them off with some knowledge in regards to the physics of DNA and RNA strands in our assays. This helps them learn to make higher predictions with less data. To do that, we run several biophysical simulations on our assay’s DNA and RNA sequences. We then feed the outcomes into the model and ask it to predict the speed and sensitivity of the assay. We repeat this process for all the experiments we’ve performed within the lab, and the model shows the difference between its predictions and what really happened. Through enough repetition, it will definitely learns how the DNA and RNA physics relate to the speed and sensitivity of every assay.
What are another ways in which AI algorithms are utilized by Sherlock Biosciences?
We have now used machine learning algorithms to unravel a wide selection of problems. A couple of examples that come to mind are related to market research and image evaluation. For market research, we were in a position to train models which study various kinds of customers, and the way many individuals may need an unmet need for disease testing. We have now also built models to investigate pictures of lateral flow strips (the variety of test commonly utilized in over-the-counter COVID tests), and routinely predict whether a positive band is present. While this looks as if a trivial task for a human, I can say first-hand that it’s an incredibly convenient alternative to manually annotating hundreds of images.
What are among the challenges behind constructing ML models that work hand in hand with leading edge bioscience technology corresponding to CRISPR?
Data availability is the fundamental challenge with applying machine learning models to any bioscience technology. CRISPR and DNA or RNA-based technologies face a particular challenge, mainly attributable to the significantly smaller structural datasets available for nucleic acids in comparison with proteins. Because of this we’ve seen huge protein ML advances in recent times (with AlphaFold2 and others), but DNA and RNA ML advances are still lagging behind.
What’s your vision for the long run of how AI will integrate with CRISPR, and bioscience?
We’re seeing a large AI boom within the protein engineering and drug discovery fields at once, and I expect this can proceed to speed up development within the pharmaceutical industry. I might like to see the identical occur with CRISPR and other DNA and RNA–based technologies in the approaching years. This may very well be incredibly impactful in diagnostics, human medicine, and artificial biology. We have now already seen the advantages of computational tools in our development of diagnostics and CRISPR technologies here at Sherlock, and I hope that this sort of work will encourage a “snowball” effect to push the sphere forward.