Mercury Computer Systems subsidiary SolMap brings compute muscle to
in silico drug design.
By Kevin Davies
Feb. 1, 2008 | In late 2005, Mercury Computer Systems conducted a successful partnership with the lab of Boston University biomedical engineering professor Sandor Vajda. It liked the results so much it decided to invest in to Vajda’s company, SolMap Pharmaceuticals.
“Mercury is in the business of solving very, very computationally intense — and intensive — problems,” says Mirza Cifric, general manager of SolMap. For two decades, Mercury was best known for providing 3D reconstruction engines for computed tomography applications in medical imaging, CT and MRI to companies such as Philips, GE, and Siemens.
In recent years, Mercury has sought to expand its life sciences business from what Cifric terms, “a hardware computer business into a more application-focused business.” That shift was prompted by the advent of more and more powerful Intel processors. Says Cifric: “The company had to grow vertically in terms of the value add. And that came by acquiring algorithms and bringing value to the application-level capabilities.”
Cifric, a native of Bosnia who moved to the United States in 1995, was tasked with developing Mercury’s business in preclinical discovery. Mercury already had some customers, notably Pfizer, for which it developed (originally for Agouron, which Pfizer acquired in 2000) a computational chemistry platform. Pfizer took the visualization software that Mercury had developed for Agouron and incorporated various algorithms and components into the computational chemistry framework, not unlike a Pipeline Pilot. “We learned lots about preclinical discovery and chemistry from that relationship and we were looking for ways to expand what we could contribute [to] pharmaceutical industry,” says Cifric.
One way to impact the drug discovery process was to apply its computational prowess to combinatorial problems in genomics, systems biology, and other fields. Cifric roamed university campuses from looking for intellectual property and companies with such capabilities. Vajda’s company proved attractive “because SolMap’s technology was heavily bottlenecked by computational limits,” says Cifric.
An expert in computational approaches to protein structure analysis, Vajda and his grad students had created technology around binding-site identification of protein targets. But even on an IBM Blue Gene supercomputer at Boston University, the simulations typically lasted 24–48 hours.
Mercury saw an opportunity: it was the first company to commercially deploy STI (Sony/Toshiba/IBM) Cell (cell broadband engine architecture) processor, the heart of Sony’s PlayStation 3. “With some engineering expertise, which is really the bread-and-butter of Mercury, we were able to take that 48-hour simulation down to literally minutes,” says Cifric. Today, those protein surface analysis calculations take 5 to 10 minutes for the whole simulation.
Despite the potential of SolMap’s technology, the question remains: how to make a successful business? Cifric explored options including providing hardware, software, and/or services to pharma companies. The reality however, he says, is that “unless you’re impacting the drug discovery pipeline in a very meaningful way, there really isn’t a significant value add to be captured in the pharmaceutical space.”
In order to turn SolMap (currently 15 staff) into a drug company capable of taking drugs from de novo design into the clinic, Cifric hired Frank Guarnieri, the founder of Locus Pharmaceuticals (See Locus Focus, Bio•IT World, December 2002). “Guarnieri has developed the small molecule design capability on this technology,” says Cifric. In addition, “we have married experimental capabilities, specifically structural biology with our computational tools. So we’re using NMR and X-ray crystallography, integrated with the computational analysis, to do fragment-based drug design.” Cifric says the addition of those two wet-lab capabilities has brought about “an absolute metamorphosis [compared to] what this company was.”
In addition to collaborations with a range of clients to enhance their drug discovery capabilities, SolMap is pushing three in-house programs. One program is in hypertension (a renin inhibitor program), another in COPD (chronic obstructive pulmonary disease), which Cifric hopes to speed into late preclinical stages. He notes that a unique structural insight in the hypertension program, initiated less than 18 months ago, is already in lead optimization, because of the platform’s capabilities.
The third is in the antibacterial space, developing an adjuvant to existing antibiotics that prevents drug resistance. Cifric says SolMap’s target is well validated. “If we’re able to inhibit this particular target, we will make all antibiotics that have been on the market for 20-plus years, which are completely unefficacious at this point, relevant once again… effectively resurrecting them!”
“The discontinuity between knowing everything about the binding site and designing a small molecule drug is bigger than Grand Canyon,” says Cifric, who claims that SolMap’s technology affords unique insight into protein structural analysis. “We have very unique insight into the bind and sub-binding pockets and the hot spots relative to the binding,” he says. For a completely de novo design, SolMap can pursue a fragment-based approach, exhaustively exploring the binding area. In other cases, SolMap might identify known binding sites as well as sub-pockets of those sites. The company then generates a rank order of hot residues most relevant to drug design. “It allows us to derive a known inhibitor to the additional sub-pockets or hot spots, which are often a unique finding of our platform. And then sometimes alleviate some of the problems that an inhibitor might have, such as potency or toxicity.”
SolMap’s fragment-based approach, which involves testing in silico fragment libraries, relies on a density function rather than binding specific fragments and “docking” them. The information obtained from the binding analysis produces a cloud of all possible chemistries that sit in the structural space. Then, Cifric says, using the simulation, a medicinal chemist can make fragment substitutions on a set of scaffolds, and make fragment substitutions on those scaffolds to obtain full coverage of the binding sites by generating a small focused medicinal chemistry library that is target specific yet extensible.
A former colleague of Guarnieri’s at Locus, Bruce Dorsey at Cephalon, was the first medicinal chemist to design an HIV protease inhibitor at Merck. “He didn’t go after potency, he went after extensibility and he was able to alleviate the problems further down in the biologically relevant tests,” says Cifric. “We’ve implemented that methodology in our in silico design, but then we rapidly go to high field NMR structure, which gives us the direct confirmation of the hypotheses.” SolMap typically synthesizes a library of 50 to 60 small molecules that are then examined using eNMR, selectively searching the target surface space.
Despite encouraging progress, SolMap has no intentions of launching clinical trials itself, says Cifric. The research collaboration model would deploy SolMap’s core capabilities effectively as a preferred partner in structure-based drug discovery programs. “It’s a lot more short-term oriented, but very cost effective and risk-sharing for the partners… Nowadays, if you don’t have intellectual property creation of your own, the valuation of your business will be dramatically impacted if you’re just a service business.”
This article appeared in Bio-IT World Magazine.
Subscriptions are free for qualifying individuals. Apply Today.