The right tool for the job

Analysing what might seem to be meaningless masses of numbers can lead to important discoveries, such as a powerful new anti-cancer drug. But you need to use the right statistical tools, according to Dr David Mitchell of CSIRO Bioinformatics.

Mitchell's fellow researchers have developed a new statistical toolkit especially for the kind of data that biotech and pharmaceutical companies deal with every day and they have taken it to BIO2003 in Washington DC to get the word out.

"There are biotech companies making critical decisions based on outmoded statistical tools. It's a bit like using a shoe to hammer in a nail.

"Unless you use the right tools on your data, you could either make an expensive mistake or miss important information. A pharmaceutical company could waste millions of dollars, for example, pursuing a drug target that has no hope of making it through animal testing just because they missed key clues in earlier data analysis."

The kind of data being generated by emerging technologies like microarrays is unique. For each sample, thousands of data points are gathered. Compare this to the kind of data biotechnologists are more familiar with: gene or protein sequence data, which requires very basic statistical analysis.

"You can't use the same tools for both", says Mitchell. "Array data needs a whole new approach".

In response, CSIRO Bioinformatics has spent two years developing statistical tools specialised for microarray data and data like it.

The company has benchmarked the tools on published data and found it can, for example, pinpoint fewer genes than other technologies when trying to find which genes predict cancer subtypes. This could make cheap diagnostic kits a reality.

Other potential applications include toxicogenomics, identifying compounds likely to fail toxicology tests, and pharamacogenomics, identifying sequences in patients' genes that predict the effectiveness of particular drugs.

CSIRO Bioinformatics is pursuing business opportunities with pharmaceutical companies in the US and Europe and has received a favourable response. Several companies are considering using the technology for diagnostics.

"It's not only the increase in sensitivity that makes these methods so important", Mitchell says, "but our ability to extract meaningful information from what was once essentially regarded as 'noise'."

Item provided courtesy of The CSIRO