CONVERSATION | Entelos’ Michael French explains what powers the systems biology engine

June 17, 2004 | The steady stream of platform announcements at New York's Penn Station complicated Michael French's long-distance call to describe systems biology to an office-bound journalist several hundred miles away. French, chief business officer at Entelos, was undeterred. Cutting through confusing background noise — much of it self-made — is an ongoing challenge for the systems biology crowd. Bio·IT World executive editor John Russell asked French to map out the routes Entelos and other systems biology hopefuls must travel to reach success.

Q: A lot of folks still scratch their heads at systems biology, wondering what it really is. What is it?
A: We think systems biology is an approach. It's not a deliverable. What you get out of systems biology is what's important to pharma. So internally, Entelos has a systems biology approach where we use proprietary in silico methodologies and technologies, computational capability, engineering, life science, mathematics, etc. What we deliver is a predictive biosimulation platform (PhysioLab).

We start by developing the technology platform for a specific disease, and determine what are the specific questions we want to address with this model. It's important that these are mechanistic models versus statistical models, so the knee bone [is] connected to the shin bone [is] connected to the presentations of human physiology. We might use some animal data because we know that the pathways are conserved between two species of animals, although the kinetics and biology may be significantly different between mice and men. We validate the [models] to human behavior and translate some understanding of human-animal systems into the context of a human physiological response.

Is that the right model?
I think the answer depends on what you expect of the deliverable. Everyone basically agrees that modeling and simulation is a cornerstone of the systems biology approach. Beyond that, what needs to be part [of your approach] depends on what you're doing. If you're doing target discovery and validation, maybe all you need are a few algorithms and lots of machines that do lots of measurements at different dynamics. Then play with [those data] to come up and say this is a good target. That's different than trying to do clinical trial or preclinical simulation.

There are two key things about integration, at least how we interpret it. Integration is taking just that information that's key to the decision process and interpreting that information, whether it's in vivo or in vitro or in animals or whatever, in the context of human physiological response. What's required to support that effort from a wet-lab capability [is] to be able to do certain assays or certain animal experiments to confirm [or not] the predicted responses.

How much wet-lab work does Entelos do?
We don't do any right now. All of the wet lab, all of the confirmation experiments, are done with partners. There was this article in Business Week last year where Bayer talked about how we did this novel gene target evaluation for them, and at the end of it we said, 'Because we have this representation of human physiology, you ought to be able to do this kind of experiment, in this cell system, and see this response if our hypothesis about what this novel gene does is correct.' Bayer ran those experiments and said, 'Yup, looks just like you said it would.' We find [doing] that a lot more effective right now than having [wet-lab capability]. But take a company like Beyond Genomics; they're all about wet lab.

What's the near-term environment for systems biology companies?
We don't see many companies doing what we do. There's a smattering of companies out there doing in silico types of things associated more with data mining and qualitative linking.

So you have a company like Ingenuity [Systems] that's using more of a data-mining technique to identify qualitative links to things. There are a ton of companies doing that, and they're going to have some moderate success. I don't think they're like us ...

Are Entelos' models quantitative? Are you building models that drill down to predict cellular mRNA concentrations?
Basically our models, since they're top-down, are more intercellular than intracellular, so we're looking at cell-to-cell communication. Yes, we're looking at quantitative changes, so we're saying that this is a kinetic between biology A and biology B, and this is the dose-response curve associated with this toxicity. Lots of people have qualitative pathway maps hanging on their walls. That's a long way from actually manipulating something and understanding what the quantitative effect is on an outcome.

But it's been a bumpy road for many companies tackling modeling. Look what happened to Physiome Sciences (which was acquired by and reconstituted as Predix Pharmaceuticals).
Physiome had some similar approaches to ours ... The challenge to a company like Physiome is that they put a lot of their eggs in the organ [modeling] basket. They did hearts and had some moderate success in [modeling] QT prolongation as a safety assessment. The problem is, [systems biology] is tough enough as it is. To be off and focused on organs instead of diseases is even more difficult. A pharma company doesn't have a heart group; it has a cardiovascular group. You're not bringing anything to the table in dealing with cardiovascular diseases. Being able to come to pharma and say, 'I can do diabetes because I can look at all the major physiological systems associated with glucose control' is different than going to them and saying, 'Look, I have an in silico pancreas.'

Gene Network Sciences has gone in the other direction — its cancer cell model is very granular.
I think they're doing good stuff too. I don't see any of these companies as competitors. I think my competitor is anybody who can come to pharma and say, 'I can do target validation the best, I can do preclinical candidate assessment, candidate prioritization, clinical trial optimization better than anybody.' That is my competitor, whether it's in silico or Exelixis doing zebrafish and Caenorhabditis elegans. I think [Gene Network Sciences] is doing really good work, and it's not counter to what we do. They're just looking at more of the intracellular cell signaling associated with oncology.

Part of our growth and evolution has been that we do this disease model better than anybody else. But you know what, we're just good modelers. We model biology really well, and we've learned a lot in doing four different diseases. Let's take that to the next level and develop understanding and models of cell signaling. This is work we're doing with Doug Lauffenburger at MIT. So we're moving from the top down, looking at some opportunities, to some bottom-up as it relates to drug development.

What's the next big change or growth for Entelos?
We're looking for opportunities to apply what we do to actual drug development. I don't think we want to go down to drug discovery. Our technology is exceedingly good for looking at combination therapies, low-dose combination therapies, the identification of biomarkers for diagnostics.

We originally had obesity and diabetes technology platforms. We integrated them into a metabolism platform, upon which we can conduct research in both diabetes and obesity now. I think as we move forward, we're going to look for the ability to integrate to do multiple disease areas.

So putting an RA (rheumatoid arthritis) PhysioLab together with a multiple sclerosis or irritable bowel or something like that makes sense. That now becomes an immunology platform we can focus on multiple diseases. If you want cardiovascular health, you'll go do a [test] panel and it'll measure your cholesterol and triglycerides and your HDL and LDL. Well, what's your immunology panel? White cells? There have got to be better ways of really understanding the immunological disease state of patients, particularly with some of the research now that suggests an inflammatory component to some of these metabolic disorders like arteriosclerosis.

The RA PhysioLab's success at optimizing rheumatoid arthritis targets for Organon produced milestone payments. How's that work going?
It wasn't quite an optimization of those targets. They gave us 25 targets, and we developed five. We then [asked], 'What are the possibilities of those targets applying a therapeutic effect if there was a compound that affected that target optimally?' We're trying to balance efficacy versus, let's say, inhibition. So 100-percent inhibition is not as good as 70-percent inhibition because the more you inhibit things, the more other things go awry. We want to modulate that target as least as we can and get the maximal effect.

We [then] provided a kind of worst-, best-, and medium-case examples of what efficacy might be. We also compared those against methotrexate. That allowed us to prioritize 1 through 30 [targets] against themselves, to draw a line where the methotrexate efficacy is, and determine how many targets are above and below the line. If you can say, here are our best-, worst-, and medium-case scenarios for target A, and all three are above the methotrexate line, that's probably going to be a winner for you.

Was the Organon collaboration on rheumatoid arthritis different from your Pfizer project on asthma?
It's actually similar to the Pfizer asthma project. It did have milestones associated with biomarker discovery and things like that, so it's not that different. One way it's different is it also includes royalties on potential products. The Pfizer relationship is also morphing. It's not in asthma any longer, I'll just say that much. We should be hopefully making some announcement here in the next quarter.