Biovista’s Unusual Brand of In Silico Drug Discovery

By John Russell

May 14, 2009 | It’s interesting what crops up in company executive profiles. Aris Persidis is an accomplished bio-entrepreneur and the co-founder of Biovista along with his brother Andreas Persidis. Aris’s company profile includes the usual details—Cellzome co-founder, Ph.D. from Cambridge University in biochemistry, etc.—but the very last item notes he was “also U.K. varsity ballroom dancing champion.” 

It’s good to be a renaissance man. The fancy footwork may come in handy as Biovista seeks to build a business around repositioning and de-risking drugs for itself and others—something many biotechs are attempting but with limited success so far. It is always dangerous to carry a metaphor (dancing) too far, but Persidis tempts fate by comparing Biovista’s business model to the dating service e-Harmony.

“You’re familiar with these e-harmony services? They have their rate of success, but the way that they operate is very interesting,” he says. “Boys and girls characterize themselves in terms of many dimensions—what color do you like, what movies do you like, and all of these things. So imagine a systematic multi-dimensional dating service between drugs, diseases, and adverse events. This is what we have done and we have done so very, very effectively.

“In silico is where the dating happens, and it happens across a vast collection of these profiles. So we have 15,000 [drug/disease] profiles, each profile consists of at least 10,000 data points, per drug.”

No doubt it’s a stretch, and the dancing analogy will stop here, but one can almost see a judging panel of biopharmaceutical and investment executives, a la Dancing with the Stars, watching these drug repositioning companies strut their stuff; sometimes they like what they see (business model, technology, promising results) and provide more projects and money to keep the contestants going; other times, as in a Len Goodman punishing rebuke, they turn the spigot off. Repositioning is hot and they can’t all win.

Persidis is betting Biovista’s unusual take on in silico drug “discovery”, supported by external CRO-conducted validation, will make it a winner.

“We created what we call the critical outcomes search space technology,” explains Persidis. With just ten full-time and ten part-time staff split between Athens, Greece, and Charlottesville, Va., Biovista has built a database encompassing the mode of action, clinical outcome, adverse event activity, and patent status for 15000 “drugs”, none of which have failed for safety or efficacy, he says.  The staff, he says, is mostly comprised of “M.D.s, Ph.D.s, and Pharm D.s with a passion for coding.”

It probably doesn’t hurt that Aris’s brother, Andreas who is the CEO, has a Ph.D. in IT and brings expertise in databases, artificial intelligence, and machine learning. Aris is the company president. (Biovista management backgrounds: http://www.biovista.com/content.php?categ=4&pid=6)

In building a computational engine to work on this data, “We had to redefine what mode of action is. We don’t think of mode of action simply as the one or two targets or three targets a drug will hit because that is very poor as descriptor. For example, an EGFR inhibitor may do so effectively within the context of a diabetic patient but not effectively in the context of an osteoporotic patient simply because these other patient cohorts are subjected to drug and disease stress that completely modifies the pathway response to EGFR inhibition.”

“Everybody goes and does a basic simple pathway analysis around the target, right, what is the upstream and downstream pathway of EGFR or whatever target you have in mind. That is going to be pretty valid but it cannot distinguish between the clinical outcome with the context of an osterperotic or diabetic patient. It absolutely cannot. We enhance the [MOA] description. We ask what does that pathway look like in liver, in the heart, in the brain, in every major tissue and organ and cell line and cell type, and different classes of patients? We do this for every disease known to medicine. So you end up with a much richer description of what mode of action is because it becomes context dependent.”

The end result, says Persidis, is “we can match the mode of action of any drug or any target against the mode of action of over 8000 diseases known to medicine and the 12000 adverse events known to medicine, so basically medicine acknowledges approximately 20,000 critical outcomes, split between adverse events and efficacious outcomes.  So when you start a development effort you basically get 60,000 possible outcomes.”

While not detailing how the inference engine works, Persidis emphasizes the process is automated and produces unbiased predictions, which can then be tested in the lab. It is the automated reasoning which produces novel predictions, unlike most other similar companies—GeneGo, Ingenuity, Ariadne—whose results are limited to extracting known knowledge.

Searching the clinical outcome space can start with a drug or a disease. So, for example, given a drug and its mode of action, one might search additional uses, or even synergistic drugs that would permit lower dosing or enhance efficacy.

“You can begin with a disease and say I have a strategic interest in this therapeutic area. Are there drugs that I can find that would have efficacy and a better safety profile in this therapeutics area that I can use as mature chemistries. I can use knowledge, for example, of a generic or drug that is coming off patent or has 2 or 3 years left. If I can successfully show it can be repositioned and I can develop a new family of drugs around that chemical class, that’s an important asset,” says Persidis. “That for us, for now, is a sweet spot and yes we are looking for partners who will work with us on the clinical development of these assets.”

Biovista used the disease-centric approach to discover a possible multiple sclerosis drug, BVA 101, just announced in April. “We wanted to work in MS. In fact within the class of BVA 101 we found other drugs that have never been used like this that we believe have got extreme plausibility in MS simply because they are chemical analogs and variants and we can design chemistry things around those. So if you do this very rich type of repositioning you can actually look and expand your mature chemistry portfolio as well,” he says. This project took 90 days, according to Perisidis.

Targeting known safe drugs is good for all the reasons often cited, and Biovista is trying to do just that as fast as it can. Biovista now has five patents covering 12 drugs for new indications on multiple sclerosis, epilepsy and macular degeneration. Persidis says, “[by] this time next year we will have 20 of the most famous drugs out there under new use IP—we have 12 already—and the year after that it will be about 40 or 50. This is where we really think there is the potential to game-change the industry,” he says.

The Biovista journey seems to have been a long one, its web site indicates the company was founded in 1996. Persidis says it was more of a hobby in the early years, becoming serious three years ago. Grant funding from the EU has been important and late last year, Biovista raised a “very significant angel round that can see us through for the next two years even if nothing else happens.”

Last September Biovista announced a collaboration with Cempra Pharmaceuticals to systematically identify and profile adverse event associations for members of the macrolide drug class. The data generated will be used to better understand the unique benefit/risk characteristics of this important group of antimicrobial drugs. Biovista is talking with several pharma now and expects to land another collaboration by the end of the year.

Despite its small size, the company has plenty of capacity at present, he says. The company uses two server farms, one in the U.S. and one in Europe, and the database is continually updated. “Every year there’s about a million new papers in pub med and about a million or a million and a half new records at FDA AERS alone. Those alter and enhance the profiles.”

It will be interesting to check back in a year to see how the market judges Biovista and see which biopharma partners decide to take a spin on the dance floor.

SIDEBAR – Biovista’s Nutrition Portal
Drug repositioning isn’t all Biovista does. The company has nutrition guidance portal which is described on line as “We focus on helping you find the most credible information on the links between drugs, diseases, health, vitamins, supplements and diet. We are not a general purpose search engine: we focus on health, drugs, vitamins, supplements and diet ingredients.”

The company even launched an iPhone application—“BNi is the mobile version of the company's Biovista Nutrition web site where consumers receive similar information as well as having access to the underlying scientific publications. Biovista announced that in the near future it will be releasing a 'Pro' version of BNi.” Link to site: http://www.biovistanutrition.com/bnamazon/loadBN.do
--

This article first appeared in Bio-IT World’s Predictive Biomedicine newsletter. Click here for a free subscription.