Solving Data Problems with Casual Human Biology: Q&A with Dr. Brent Richards

As technology improves, collecting clinical data has become much easier. However, this provides both benefits and adds new difficulties.

As researchers gather more data, they risk making decisions based on the wrong data or burying the good, actionable data. Last year, Pharmaceutical Executive reported on the use of new technologies are enabling a multiomic approach to drug development.

At the time, Dr. Jennifer Bath wrote that this approach allows for data from multiple data sets to be leveraged. New technologies are able to parse through the data and find connections across diverse sources and identify the actionable data from different points of view.

Pharmaceutical Executive recently spoke with Dr. Brent Richards, CEO of 5 Prime Sciences, about how pharma and biotech companies are continuing to handle the issue of having too much data. According to him, one of the ways to solve this issue is to focus on casual human biology, which allows for researchers to identify the important data trends earlier in the R&D process.

Pharmaceutical Executive: Why is drug development failure often a biology problem?
Dr. Brent Richards: Drug development is extraordinarily difficult, and execution always matters. But many programs fail because the underlying biological hypothesis was inapplicable to humans. A target may be associated with a disease, but not actually cause it. A gene may influence who develops disease, but not how the disease progresses once a patient has it. A biomarker may correlate with an outcome in animals, but not reflect the biology a therapy is trying to change in humans.

That distinction matters because by the time the problem becomes visible in Phase 2 or Phase 3, the program has already absorbed years of work and substantial capital. The question we should be asking much earlier is not simply whether a target is druggable, or whether there is interesting biology around it. It is whether the human evidence supports the specific therapeutic hypothesis.

That is the reason we built 5Prime around causal human biology. The goal is to bring the kinds of questions that often emerge too late in development, target causality, direction of modulation, progression biology, on-target safety, biomarkers, and patient selection, much earlier into the decision-making process.

PE: What makes human genetic evidence actionable inside pharma?
Richards: Human genetic evidence becomes actionable when it is tied to a decision. A genetic association is useful, but on its own it does not necessarily tell a development team what to do. The more valuable question is whether the evidence can help a team decide to advance a target, where and how to modulate the target, change the therapeutic direction, select a biomarker, enrich for a patient population, anticipate a safety issue, or stop a program before more capital is committed.

That requires more than data generation. It requires a framework for asking drug development questions of human biology. A statistical geneticist may need to evaluate the methods in detail, but a translational leader, clinical team, portfolio committee, or business development group needs to understand the implication for the program.

At 5Prime, this is the gap we are closing. We have built Centromere, our decision analytic platform, to translate complex human genetic and multi-omic evidence into decision-ready outputs. The scientific depth is still there, but the output has to be usable by the people making program and portfolio decisions.

PE: Does “more data” translate into better R&D decisions?
Richards: No, not by itself. The amount of human genetic and multi-omic data available today is extraordinary, and it is growing quickly. But more data does not automatically produce better decisions. In many cases, it simply creates more associations to interpret, each of which has its own sets of biases and thus may subtly contradict each other.

The harder task is knowing which questions to ask of the data. Is this the causal gene at the locus? Does changing the target alter disease risk or progression? Should we inhibit or activate the target? Are there predictable on-target safety signals? Which biomarkers reflect the mechanism and can be used as proximal readouts, and which patients are most likely to benefit?

Those are the questions that determine whether data are useful for drug development. Much of the work we have done at 5Prime has been focused on building the infrastructure and algorithms to answer these questions repeatedly and at scale, including by integrating public and proprietary datasets. The value is not data volume by itself. The value is being able to turn the right evidence into a clearer decision before the next major investment. Importantly, this must be done by weighing probabilities with imperfect information.

PE: How does causal biology impact the direction of a program?
Richards: Causal biology can change the direction of a program very substantially. It can show that a target should not be advanced, even when the surrounding biology looks plausible. It can identify programs that have a very high probability of getting to market.It can identify biomarkers that are more likely to reflect target engagement or disease modification. It can also point to the patient populations where the mechanism is most likely to matter.

One area I think is especially important is the difference between disease risk and disease progression. A target may influence whether someone develops a disease, but that does not necessarily mean it will change outcomes in patients who already have the disease. That distinction is highly relevant to clinical development, but it is still underappreciated.

This is why we have built tools like Trajecta, PRIMEx, and PRIMESuspect within Centromere. These AI-backed algorithms are designed to ask more development-relevant questions of human biology, not simply to produce more data and associations. We have seen how their output can influence whether a program advances at Board of Director level decision-making at large pharma, whether a program is selected to go to clinic, which patients are selected for proof of concept trials, and whether the evidence supports the next investment.

PE: How is the standard of evidence for target selection and early clinical strategy changing?
Richards: The standard is becoming higher and more integrated. It is no longer enough to say that a target is interesting, druggable, or associated with disease. Increasingly, teams need to know whether the target is causal in humans, how it should be modulated, whether it affects disease progression, what biomarkers should be used, what safety risks should be anticipated, and which patients should be included in early trials.

That is a very different standard from traditional target selection. It connects discovery, translation, early clinical strategy, and portfolio decision-making. It also means that human genetics and multi-omics should not be treated as separate evidence packages. They should be part of a decision framework, which is triangulated to make the best decisions.

That is where I think the field is moving. The opportunity is to use human causal biology earlier and more systematically, so that companies and their investors can increase the probability of success and avoid advancing programs where the biology does not support the investment. That is the purpose behind the platform we have built at 5Prime: to make that standard of evidence practical, repeatable, and usable for decision-making across the pipeline.