A major goal for cancer research is predicting which cancer cells will metastasise, and why, researchers said.
"The approach we've taken is a reverse approach from what is conventionally done," said Cynthia Reinhart-King, associate professor of biomedical engineering at Cornell university and senior author of the research.
"Instead of looking at what molecules are being expressed by the tumour, we're looking for the phenotype - that is, the behaviour - of individual cells first. Then we can determine what molecules are causing that behaviour," Reinhart-King said.
Searching for biomarkers of metastasis has focused on screening for certain molecules or genes expressed by large numbers of migrating cancer cells.
The problem is that it's easy to miss subtle differences in the tiny subpopulations of cells that are the most aggressive.
Taking, for example, 100 tumours and seeking out molecular biomarkers for metastasis, one particular molecule might be identified as being 'upregulated' in those tumours, Reinhart-King said.
But it's not the whole tumour expressing that particular molecule - some cells express the biomarker and some do not.
The researchers decided to first sort cells with the most aggressive behaviour, and analyse only them for molecular changes.
Their innovation is a microfluidic device that contains side channels to wash out the less aggressive cells, while herding the more aggressive ones into a separate channel.
The researchers used the device to screen for cells with migratory responses to Epidermal Growth Factor, for which the receptor is known to be present in most human cancers and is tightly linked to poor prognosis.
"The thing we're most excited about, in addition to the physical device, is the conceptual framework we're using by trying to shift gears and screen for cells that are causing the worst parts of the disease," Reinhart-King said.