Using mathematical modeling to derisk screening and lead optimization
A drug discovery program at its inception can feel like such a gamble. The biology is often still being understood, and it’s often unclear whether the chemistry will cooperate. When running a discovery project, we’re often faced with the conundrum of “so many unknowns, so little time (and money)”. When money is tight, reaching the next value inflection point can make the difference between life and death for a project, or a small biotech.
If you’re in the business of discovering new medicines, you should know that there are ways to:
· Hone in on the “best” candidate molecules to test on the way to a new product
· Increase the likelihood of discovering a safe, effective compound; a.k.a. a drug candidate
· Save far more money on candidate identification and testing than you spend on a mathematical model to guide you
The ability to more confidently and quickly proceed from an interesting compound to an actual drug candidate is critical, especially for smaller pharmaceutical and biotech companies.
How drug candidates are usually identified
To understand the promise of mathematical modeling, it’s important to look first at how drug discovery has traditionally been handled.
Let’s say you aim to cure cancer, or at least develop a better treatment for lymphoma, for example, than anything else on the market. Your modality of choice is small molecules, and you’ve identified a compound with compelling attributes, say it binds your target of interest and kills cells in tissue culture. As a result, you now have some investor backing to explore it further. (Congratulations!).
The next step toward your blockbuster drug is to identify a derivative that displays excellent cancer-fighting capabilities at a dose humans can tolerate. (Bleach, for example, will kill cancer cells in a tissue culture plate, but no one has proposed using it as an anticancer drug. Yet.) Generally, drug developers are looking for the sweet spot where three essential qualities overlap:
· Binding
The compound must bind to the targeted protein to deliver its tumor-suppressing effect. This is what makes it a drug candidate in the first place!
· Metabolic liability
The compound needs to stick around in the human body long enough to have an impact.
· Solubility
The compound must be soluble enough to create an injectable or ingestible medicine and to work in the watery environment that is the human body.
The traditional method of finding a molecule with the secret sauce to become a medicine someday is to conduct a screening cascade—something that, in layman’s terms, closely resembles trial and error. In a hypothetical case, researchers might:
· First identify molecules with exceptionally strong binding properties with the targeted protein.
· Iterate to discover strong-binding molecules that are also adequately soluble.
· Address the metabolic criteria in the hopes that one of the strong-binding, soluble compounds will hang around in the body long enough to serve as a medicine.
Easy, right? Unfortunately, it’s not.
Like playing the lottery
Such linear sequential iteration tends to wind up in dead ends. Returning to our hypothetical case, it is highly likely that none of the strong-binding, soluble compounds will meet the metabolic criteria. That’s just the way chemistry works sometimes.
Presented with this disappointing result, researchers have no choice but to go back to the drawing board with metabolic liability in mind. Find a molecule that meets this criterion, however, and you’re likely to find that it’s not very soluble.
This is a very simplistic example, but it shows how researchers wind up chasing their tails, tweaking molecules to meet one criterion only to miss on another. And we haven’t even talked about dose yet!
Using this approach, drug development looks disturbingly like playing the lottery—with about the same chance of hitting the jackpot. That’s not great news for a young biotech company with limited dollars and a need for a positive result (a good candidate drug) to garner more funding. And it’s especially bad news for the millions of patients waiting for new treatments and cures for a variety of devastating diseases.
There must be a better way
As a hands-on researcher in the mouse room, I found myself reminded of this phrase from Winnie the Pooh (the children’s book series) hundreds of times. As my team and I tested what felt like one random compound after the other, we would make progress on solubility with one molecule, only to find that the pharmacokinetics (PK) of the more soluble molecule was now worse. The next compound in the series would have better PK, but with a 10-fold reduction in binding potency. It often felt like we were making glacially slow progress towards a real candidate, and I would keep saying to myself, as we did in vivo work sacrificing mice by the dozen, there must be a better way. Couldn’t we (and shouldn’t we) be more strategic about where and how we look for compounds that can reach the clinic, I asked myself?
FractalTx emerged as an answer to that question. Over the past decade, the team at FractalTx has demonstrated—and you can check out the published proof on our website—that mathematical modeling can accelerate and de-risk the drug discovery process. Models can transform what is, for some teams, a lottery into a far more focused search for promising chemical matter.
To be clear, not all models can do that. Models that are overly focused on describing the biology of the target in depth can quickly vanish down a rabbit hole, never to contribute to the drug discovery process. These Systems Biology models have their place (if you can find the right journal for them!) but they are complex, expensive to commission, and tend to include parameters that are not measurable experimentally. This means they cannot, in general, help with the drug-candidate identification problem we’re talking about here.
Fortunately, simpler and less costly modeling does offer the power we’re seeking. The key is to frame the modeling exercise in terms of the pharmacologyparameters being used in the screening cascade. This is as much an art as science, but a parsimonious model that is framed in terms of directly measurableparameters can be incredibly useful during the drug discovery process. Such a model can can provide estimates of the developability potential for each iteration of the lead optimization process, and can be directly integrated into the project team’s dashboard for measuring progress. Quantitative systems pharmacology (QSP) models- done right- can get you through lead optimization more quickly and make a “win” easier to achieve.
Five key benefits of QSP models
In our experience, QSP models provide valuable insights throughout the drug development process. They can be used to:
1. Select screening thresholds and/or design a screening cascade in light of the parameters expected to have the biggest impact on the molecule’s performance.
2. Target the most promising chemical matter to expand on, defining the most impactful parameters to focus on during lead optimization.
3. Project the clinical performance of top candidates to ensure the lead optimization process is working and the effort is actually improving the clinical potential.
4. Conduct in silico experiments to dry-run in vivo studies for efficiency and cost-savings when compared with “going right to mice” – this can be surprisingly useful in making sure that the dose ranges tested in the in vivo study are the most informative.
5. Aid clinical trial design, including informing the first-in-human doses to be tested.
That’s a lot of benefit from a single model, and pharmaceutical and biotech firms find that they more than recoup their investment because the model empowers them to implement more efficient, cost-effective studies. Most importantly, by looking for new medicines in the best places, companies have a greater chance of doing what they set out to do, finding that blockbuster drug. In our experience, when a model is framed with the goal of driving the project forward, it very quickly begins to pay for itself- a model like this costs as much as a couple of pharmacology experiments, and can eliminate a lot of dead ends!
Identifying drug candidates this way is a heck of an improvement over buying lottery tickets, even if it does involve math. But don’t worry, we take care of the math for you.
Want to talk about what it can do for your drug development projects? Have questions that this article didn’t answer? Feel free to reach out! As you can probably tell, I’m evangelical on this topic, because I strongly believe that strategic use of mathematical modeling can transform drug discovery, increase returns for biotech firms, and bring more treatments and cures to patients. If you go over to our webpage: https://fractaltx.com, you will find white papers that provide more detail on each of the five benefits described above. Alternatively, you can just ping me on LinkedIn!