New bio-IT firm aims to sort out the rank and file

It's one thing to design a system that can allow a library of 100,000 compounds to be screened against a protein structure in a day to assess each compound's structural compatibility with the target site. It's another to offer a way to score or rank the compounds according to the best-predicted fit.

That's the stated capability of Compuscreen, a small bio-IT company that has developed from research at the Walter and Eliza Hall Institute and the now-defunct Biomolecular Research Institute.

The company was formed to exploit a powerful in silico drug screening system developed by WEHI postgraduate student Kim Branson and his supervisor Dr Brian Smith.

The reason for performing in silico screens of compound libraries is of course to save time and money by winnowing down the field to a set of compounds most likely to contain drug lead candidates. With the commercially available libraries containing upward of 50,000 compounds, high-throughput screening can take weeks, if not months.

"It's significantly cheaper to do a virtual screen than to do a high-throughput screen," says Smith. "The idea is that you then don't have to buy as many compounds to test in assays."

Branson and Smith's screening system is based on DOCK, a program developed at the University of California in San Francisco that examines the ways that two molecules, such as a protein and a small molecule, might fit together. But around DOCK, the researchers have developed some innovative tools to maximise the chances of identifying small molecules with high binding affinity to the target site.

"The big problem in molecular screening is finding out which molecules will dock or bind tightly," says Branson. "The challenge is to evaluate how sticky something is, its binding affinity."

Increasingly, Branson says, consensus methods are used, which combine the data from a variety of different algorithms generating a consensus score. But the problem with this method is that combining the data from different methods of analysis often misses good candidates.

So Branson and Smith have developed a package of software components, built on the backbone of DOCK, that they call Elucidator. Combined, the system is faster and more accurate than DOCK. Essentially, for each compound, a variety of orientations in the active site are considered, and the best solution is picked, taking into account whether structural water is required for ligand binding.

Finally, the compounds are given a score and ranked in order. This allows a subset of the chemical library to be selected for further testing in assays, with a high likelihood of finding the most active molecules in the subset.

"The point is to significantly reduce the number of compounds that need to be tested. The accuracy of predicting the binding is also quite good," says Branson.

Powerful processors

Running the system is a cluster of powerful processors. Branson and Smith have built two clusters so far, and both are in use. The initial set-up, nicknamed Caduceus, consists of a cluster of 64 AMD Athlon Linux processors whimsically arranged as a double helix. The second uses 32 dual processors, and is about 70 per cent faster than Caduceus. And for bigger problems, a global grid of computing power is accessible.

The researchers have already tested out Elucidator in a variety of ways, by comparing results of their system with other systems such as DOCK in chemical libraries seeded with known ligands to various proteins, and by comparing results from high throughput screening assays to virtual results. For example, using neuraminidase, at least one of the known ligands was ranked in the top 10 compounds when seeded into two commercially available libraries.

The challenge now for the fledgling company is to work out how to best commercialise the product. "The question is, how is the company going to evolve?" says Smith. "It's obvious we can't exist as a fee-for-service company alone."

The scientists have been in discussion with a number of VCs to look at the options available to them. But Australian VC companies tend to shy away from tool-based companies, preferring to fund companies focused on the leads obtained from application of their technology to develop drugs, Smith says.

One option is to remain purely in the informatics field, marketing Compuscreen's technology to pharma and biotech companies overseas. Smith and Branson see Elucidator as having applications in a variety of settings, as one tool of many employed by pharmaceutical companies, as an efficient method to initially screen libraries for money-conscious biotechnology companies, and as a tool for public sector research.

But another possibility for Compuscreen is to use the technology internally to develop lead candidates for target proteins, an option the team is leaning towards.

"We'll be looking at opportunities overseas for tool-based companies, but we're heading away from being just a tool-based company. We have our own internal targets that we're working on," Smith says.

They are involved in a number of collaborations with researchers in the public sector, including several projects with WEHI and the CRC for Cellular Growth Factors, and with former colleagues now at the CSIRO. Using Elucidator small molecule inhibitors have been identified for inhibition of the Interleukin-6 Receptor. Follow-up structural work confirms the binding mode and location of these inhibitors.

In addition, a project with the Victorian Infectious Diseases Reference Laboratory (VIDRL) is using a model of the SARS protease to develop potential inhibitors for the deadly virus. Another project, in collaboration with Assoc Prof Rick Thompson at St Vincent's Institute of Medical Research in Melbourne, has identified a number of compounds demonstrating activity against the MMP-2 protease, which plays a role in tumour metastasis.