Computational chemistry can predict the fragments worth making.
By Kevin Davies
May 18, 2010 | ‘Computational chemistry got a bad rap early on because it wasn’t necessarily as predictive as it might have been,” says David Pompliano, CEO of BioLeap. “We’re on the cusp of something very different.”
In computational chemistry, there’s screening and there’s docking and then there’s chemical potential annealing, a form of fragment-based drug discovery that is BioLeap’s specialty. By computationally designing compounds and determining their rank order binding affinity, BioLeap makes many fewer molecules en route to a drug candidate. “Typically, in normal lead optimization chemistry in pharma, you try several things and you have to make all of them. We can tell [clients] beforehand which ones are worth making. We don’t screen or dock. We create molecules de novo from a fragment map.”
BioLeap was founded in 2004 by John Kulp (the CTO) and Gerry Evans. Frank Guanieri and Kulp co-founded Locus (see, “Locus Focus,” Bio•IT World, Dec 2002). Kulp’s technology has found many applications in agriculture (herbicides), oil (catalysts), and pharma for early drug discovery. Pompliano took the helm of Bio-Leap in February 2009. His background is in big pharma , with positions at Merck, DuPont and GlaxoSmithKline (GSK) on his resume.
While BioLeap is keeping a couple of infectious disease programs in house, it’s preferred method is to collaborate with partners, who will typically do the wet chemistry work. Publicly announced partners include an early discovery program for GSK (signed in September 2009) for targets for which traditional methods have failed. Another deal is Lycera Corporation in the field of depression.
BioLeap uses computational Monte Carlo methods to generate maps of where small chemical fragments bind to a target protein and with what relative affinity. Pompliano explains, “This is done in a thermodynamically principled way so data are relatable to a free energy metric. So we can predict relative binding affinity of all those little fragments, then when we design compounds we rank order those compounds.”
The chemical potential annealing approach, which was initially developed by Frank Guanieri at Sarnoff, allows BioLeap to quantitatively account for entropy. (There is also good accounting for bound water molecules, which have been the bane of many computational methods). “We’ve found a reasonably quick method of doing the calculations,” says Pompliano. “Importantly, we can have an impact in the time frame of a drug discovery program.”
Many programs using docking or virtual screening strategies do not generally take into account entropy. Pompliano offers a quick refresher in basic thermodynamics: free energy, the metric associated with binding affinity, is composed of enthalpy and entropy. “If you miss the entropy component, you’re potentially missing binding affinity,” he says.
This method “is more predictive to tell you the next best molecule to make. We can rank order compounds with enough accuracy to make a decision whether to make a compound or not.”
Pompliano isn’t claiming he can calculate precise binding affinity. But, he says, “we can tell you these compounds have this relative affinity, so if you’ve got 100 you want to make, we can say don’t make the bottom 80—and in a timeframe that’s reasonable.”
The calculations for a large number of fragments binding to a given target takes 1-3 weeks on some 40 nodes on standard issue computers that BioLeap is currently leasing. The database creation is the first part of the process, with tools to interpret those data, design, and evaluate compounds on the screen. “It’s really a chemists’ tool, allowing BioLeap’s chemists to design new compounds and know in advance what the relative binding affinity will be.”
In traditional docking or screening processes, the target is matched against a set of pre-conceived compound structures to identify a conformation of ligand and protein that fit together. The BioLeap process begins with a target protein structure, which is a pre-requisite. Pompliano describes the process from there:
“We put [the target] in a box, and put in one fragment at a time. Rather than come to some free energy minimum, we impose a free energy and see how the system responds. When we impose high free energy, nothing special happens. The molecules are all there. But as you lower the imposed free energy in the system, you essentially start to find spots on the target protein that have an interaction with the fragment. They stick, if you will, while the others essentially ‘boil off.’ Now you’re left with a target with this particular fragment stuck at various places around the protein. We can record all that data—opposition, relative binding affinity, etc.”
BioLeap next keeps lowering the imposed energy and recording the results until there’s just one fragment bound—the highest affinity. With a record of the position and relative binding affinity of that fragment, the process is replicated with a group of other fragments, perhaps 100 in total. The result is a database with all the fragmentary building blocks around the protein and their relative binding affinities. BioLeap’s proprietary software generates a fragment map.
At this point, BioLeap has the capability to manipulate the most promising fragments. For example, it can select a fragment that binds to the active site or an allosteric site. That fragment might have six modifiable chemical positions. BioLeap chemists can then click on one of those positions, which Pompliano says, “returns a list of all other fragments within the chemical bonding distance and the chemical angle for proper synthetic chemistry to make connections.” This gives the chemist choices in selecting and modifying fragments based on affinity or SAR considerations, and build up a picture of what the final compound is going to look like. “It’s very cool!”
The fragments are relatively small, up to 80-100 molecular weight. Pompliano stresses the advantages of computational methods versus experimental, which overcomes issues such as polarity, trying to assemble a non-polar molecule with polar fragments, or experimental restrictions on using soluble, polar fragments. “We can throw in any moiety and test it out, which increases chemical diversity.”
After the computational results, Bio-Leap designs a small number of compounds, perhaps 20 high-affinity molecules in all, 4-5 compounds each with 4-5 different scaffolds. A partner synthesizes the compounds and tests them in various biochemical assays to determine their relative affinity. “That’s sent back to us to do the next round of design,” says Pompliano. “It gives us a handle on how the model is working. We repeat the cycle 2-3 times.”
With some funds in the bank, Pompliano says his three near-term priorities are business development; further development and refinement of the technology platform; and exploring newer models for drug discovery and fee for service. •
This article also appeared in the May-June 2010 issue of Bio-IT World Magazine.
Subscriptions are free for qualifying individuals. Apply today.