Cytopia's Linux investment speeds molecular candidate modelling

A month after its installation, a major computerised modelling system, based on the Linux operating system, is delivering substantial cost and time savings for biotech venture Cytopia.

Majority owned by Medica Holdings, Cytopia has emerged from Melbourne's St Vincent's Hospital with a singular focus on exploring potentially disruptive treatments for inflammatory conditions such as inflammations (hay fever, food allergies and rheumatoid arthritis) as well as Interleukin-6 dependent afflictions like prostate cancer. Cytopia holds worldwide rights to the JAK kinase enzymes that regulate IL-6, and has been extensively modelling candidate molecules to assess their interaction with the enzymes.

Cytopia has always understood the potential of virtual modelling software to speed its drug discovery process. It initially installed off-the-shelf packages such as Accelrys to begin down this path, but over time sourced software development expertise to customise that package to its needs.

Combined with a number of freely accessible open source applications, Cytopia has over three years built Chemaphore, a highly-tailored solution that fits its requirements. The system allows Cytopia to map the structure of candidate base molecules in three dimensions, then append thousands of tail molecules and score their interaction with the target. Potentially promising molecules were identified, then passed to Cytopia's Medicinal Chemistry Group to be manufactured or purchased.

In the past, Cytopia had spread its modelling system across 10 dual-processor servers, which would each handle part of the computational workload. But this approach was proving hard to scale, says Cytopia head of computational biology Herbert Treutlein, because it was difficult to maintain consistency across the various systems.

"The problem we're looking at is quite complex and embarrassingly parallel [meaning that 100 processors can analyse 100 molecules at the same time," he explains. "Our previous approach had worked, but it was harder to extend if you're planning [for growth] ahead. After 20 processors, it gets a bit complicated from an administrative point of view; you need to have an extra person [to keep it running]."

Earlier this year, Cytopia began investigating alternatives and quickly discovered the potential of IBM's eServer xSeries. Unlike proprietary and therefore expensive high-end systems running Unix, the xSeries is based on low-cost, industry standard Intel Xeon processors (Cytopia installed 64 processors, but the system will support 128 or more processors in the future). It runs the Linux operating system -- a free, easily customisable environment that's challenged Microsoft Windows by offering high-end computing capabilities on commodity servers at a fraction of the price.

Because of its scalability and broad support, Linux is a linchpin of IBM's two-year-old push to establish itself within the biotechnology sector, where the platform's low cost appeals to innovative but penny-pinching tiny start-ups that need to get to market as quickly as possible.

Cytopia ported its application to Linux, a process Treutlein says "wasn't really that hard", and immediately began to see speed gains in the discovery process. The system allows Cytopia to screen a candidate molecule every 10 seconds, or more than 8000 molecules in a given 24-hour period. That's more than four times the speed of the old cluster, and the new speed has already delivered considerable cost savings for the company by slashing the number of molecules it must source for lab testing.

Screening a typical group of 100,000 or more molecules might have previously turned up 1000 or more candidate molecules, each of which needed to be sourced -- at a cost of $5 to $50 each -- for wet-lab testing. With the computerised modelling allowing matching that's been demonstrated at nearly 100 per cent accurate, the number of candidate molecules might typically be reduced to just a couple of hundred.

That means big savings for Cytopia, which needs to keep its costs under control as it readies for anticipated human trials next year and a possible public listing some time after that. As it continues to grow its computing investment, Cytopia may eventually look to sell access to its system as a service for other cash-strapped biotech aspirants.

"On the computer, we can make and test many more molecules than we can make in reality," says Treutlein. "This helps us optimise our libraries of molecules, and it's a big step for us."