.png)
This is part four of our four part series on Octant's multiplexing capabilities. To read more on multiplexing for drug discovery, check out our posts on Cellular Intelligence, broad target scans, and reporter development.
Target-based drug discovery, which focuses on drugging specific proteins linked to a disease, has been revolutionized by monumental advances in understanding protein structures and protein-drug interactions. This progress is in part thanks to technologies like x-ray crystallography, cryo-electron microscopy, and, more recently, projects like AlphaFold. However, we still lack a comprehensive understanding of how drug interactions with different regions of a protein affect protein function and disease. To gain this knowledge, what’s needed is another dimension of data that layers insights about function on top of protein structures. Armed with this data, we can design more effective drugs and better understand the effect drugs would have on different patients.
At Octant, we’ve pioneered a technology that enables a more dynamic view of proteins. Deep mutational scanning (DMS) is a cutting-edge technique that links protein structure to function by systematically probing every position of the protein. Using DMS like a microscope that delves into the behavioral intricacies of thousands of protein variants at a time, we can supercharge each stage of drug discovery, ultimately leading to precision treatments for specific patient populations.
To conduct a DMS, we comprehensively assess the function of a protein by building uniquely barcoded human cell lines for every possible single amino acid variant of the protein, probing the cell lines together under carefully controlled conditions. This enables us to simultaneously test the activity of all of these variants with very high resolution. Most similar technologies provide a binary readout on whether or not a given mutation increases or decreases protein function, but high-resolution DMS measures the degree to which different mutations affect protein function. It’s this ability to make precise comparative measurements between different variants and conditions that generates valuable insights. Because of our multiplexing capabilities and computational infrastructure, we are able to measure effects on protein function at resolutions not yet seen in the field.
DMS drives valuable insights across different stages of drug discovery, from target discovery to the clinic. Below are examples of unique insights provided by DMS:
We recently collaborated with Bristol Myers Squibb to deeply characterize TYK2, an immune-relevant target. While TYK2 has a known main “catalytic” binding site, drugging that site produces side effects. We used DMS to systematically scan TYK2 to explore whether we could identify druggable secondary sites on the protein (sometimes called “allosteric” sites). Using DMS we were able to identify multiple such sites, offering insight into therapeutically relevant alternative regions for drugging the protein.
To do this work, we conducted DMS experiments on two cellular functions: IFN-signaling (a pathway that plays a critical part in the human immune response) and TYK2 protein expression (the abundance of the protein). By identifying variants that impact IFN-signaling but not protein expression, we can understand which parts of the protein, when perturbed, affect IFN-signaling. By clustering these positions, we were able to elucidate the known catalytic site and a druggable allosteric site, in addition to two previously uncharacterized sites.
Mutations in MC4R are the leading cause of inherited obesity. The gene plays a role in human appetite and energy expenditure, but some activities of the protein can also activate toxic side effects like high blood pressure. We used DMS to better understand which positions on the MC4R protein signal to downstream pathways that drive the desired effects, versus those that drive toxicity, in order to design drugs that are “biased” against activation of side effects.
One way to do this is to map how existing molecules with known effects on the target are causing different outcomes through the target protein’s different signaling pathways. This information can be used to improve new drug candidates to produce more desired activities. Shown below is a 3D view of the MC4R protein bound by a drug, with protein positions colored by whether one or both molecules are acting through that position on the protein. What we see is a nice “map” of protein-drug interactions that define the site of action for this therapeutic target. Comparing maps like these for different peptides and small molecules helped us build molecules that more precisely interacted with that site of action to confer favorable therapeutic outcomes.
While drugs that target GLP-1R and GIPR – like Ozempic and Mounjaro – were originally developed to treat diabetes, their weight-loss effects quickly led to expansion into the obesity market (as Wegovy and Zepbound, respectively) and became instant commercial blockbusters. These therapies have revolutionized diabetes care and weight management, and emerging evidence supports a therapeutic benefit against cardiac and kidney disease. On the heels of these success stories, pharmaceutical companies are looking for insights to guide the development of next-generation therapies.
By running Deep Mutational Scans across multiple compounds, we gained a sense for which parts of these proteins are universally important for activity, and, more compellingly, we identified drug-specific variant effects that provide insights into vectors for drug optimization. Using this data, we can better understand the constraints for making a drug that targets both GLP-1R and GIPR.
Octant’s lead drug program is for rhodopsin-associated autosomal dominant Retinitis Pigmentosa (RHO-adRP), a disease that causes progressive deterioration of vision and eventual blindness. There are over 200 mutations that cause RHO-adRP, the majority of which result in the misfolding of the rhodopsin protein. One of the major challenges in creating a drug for RHO-adRP is the hundreds of mutations spread across the patient population, each of which causes the rhodopsin protein to misfold into a different, difficult-to-predict shape. Because these mutations lead to different structural outcomes, a compound that corrects rhodopsin for one mutation might not necessarily correct misfolding for other mutations. This is where DMS comes into play.
With our DMS capabilities, we first measured the functional effect of every single RHO mutation to understand which of them lead to mistrafficking (a functional consequence of misfolding). Then, we screened our lead compounds against each mutation to identify which variants are rescued. All together, this gives us the ability to predict with precision Octant’s ability to treat patients across the entire spectrum of RHO-adRP mutations: our lead compounds are able to treat more than 80% of patients with a misfolding variant.
The DMS experiments we run generate massive amounts of high-resolution data that are ideal for applying AI to drug discovery. Collectively, the experiments described above generated ~100M data points. This precise and repeatable functional data is useful for AI/ML approaches to discover promising new islands of chemical space, generate compound libraries, and optimize drug efficacy. It’s also uniquely suited to improve protein-function foundation models, unlocking exponentially better ways to engineer more precise medicines.
It’s the early days for DMS technology, and the future is bright. Mapping a protein’s actual function onto its structure is a new way to think about drug discovery and will eventually impact all stages of drug discovery and development, similar to the advent of structure-based discovery.
If you’re curious to learn even more, you can check out our award-winning poster and our presentation at the 2024 Mutational Scanning Symposium. These detail how we used DMS to understand the biology and druggability of MC4R.
.jpg)
.jpg)