A new study led by Dr Wil Gardner, in collaboration with the Olivia Newton-John Cancer Research Institute (ONJCRI), has used machine learning to reveal a ‘spatial-molecular fingerprint’ in breast cancer tissue, with the potential to help researchers better understand disease progression and prognosis.
“Modern imaging technologies can generate images that capture extremely detailed molecular information from tissues and materials,” he explains.
“These techniques are used in many fields including cancer biology, materials science, plant science and advanced manufacturing.”
Despite widespread use of these techniques, extracting meaningful information from such large and complex data is challenging.
To address this, Dr Gardner worked with the team in Professor Robin Anderson’s Metastasis Research Laboratory at ONJCRI to develop a new machine learning approach called stability selection-augmented multiple instance learning (SSAMIL).
“We applied this approach to mass spectrometry imaging data that we acquired from a collection of triple-negative breast cancer tissue,” he says. “We showed that SSAMIL could accurately discriminate tumours with increased and decreased expression of BMP4, a protein that is linked to metastasis. It also identified specific spatial regions and molecular signatures that distinguished the tumour subtypes.”
Dr Gardner says this discovery is significant for cancer research because spatial variation in molecular composition can reveal how tumours grow, reshape their microenvironment and metastasise.
His next step will be to extend this approach to human tissue biopsies, working collaboratively with ONJCRI postdoctoral researcher Dr Kellie Mouchemore.
“Our aim is to predict treatment response and patient prognosis directly from spatial-molecular tissue data, while also further understanding the hidden features driving these outcomes.”
