The trouble with biomarkers

4 minute read

Biomarkers can look very different even inside the same individual, and analysis might need a slightly broader view of cancer samples.

The concept of finding biomarkers – DNA, mRNA or protein signatures – that predict which patients with cancer will respond to different treatments lies at the heart of the precision medicine dream.

But what makes identifying biomarkers challenging is that, even inside the one individual, tumours and human tissues have very diverse molecular profiles.

If you take a biopsy from one part of a tumour, it might look molecularly very different from a biopsy taken from another section of the same tumour.

And, if you take a molecular reading from cancer at one site in the body, the biomarkers may be quite different from those taken from another site, even though the cancer samples came from the same patient.

All this noise means scientists need to look for very special types of biomarkers that fit into a “goldilocks zone” of being different enough from person to person such that patients can be classified into subgroups and offered different treatments, but stable across the tumour such that every patient is assigned to the  correct subgroup every time and no one is missed.

Dr Rosemary Balleine, working with the ProCan program at the Children’s Medical Research Institute and researchers at the Westmead Institute for Medical Research, presented on this topic on Saturday, 3 July at the Pathology Update 2021 hosted by the Royal College of Pathologists of Australasia.

“Now, something you really don’t want is where you can take a sample from a person and think something’s there and then you get another sample from the same person and you find that it’s not,” she said. “That’s a very unstable biomarker.

“So, what we’re really looking for is something that shows a consistent pattern of expression in all of the samples that you might take from one person,” said Dr Balleine.

“However, for it to be useful as a biomarker, it needs to show a degree of variation across the cohort – otherwise, you can’t tell people apart.”

Dr Balleine’s team call the characteristics of this special type of biomarker “intrinsic variable expression”.

Her team investigated the molecular profile of more than 400 samples, taken from 11 patients with ovarian cancer using mass spectrometry, and found more than 4,000 proteins. However, only around 5% fitted into the ‘goldilocks zone’ (i.e., had “intrinsic variable expression”).

To add to this complexity, the messenger RNA profile and the protein profile (the proteome) look very different from one another at the same point in time in the same sample. This is slightly counterintuitive because the mRNA is what makes the proteins, and you might expect them to be highly correlated.

The difference is explained by factors including the time lag between transcription and translation, and the mixing in of extracellular proteins, such as proteins in the blood that are less evident in the transcriptome, said Dr Balleine.

“Another one that’s important is that cells or tissues are living, breathing things, even though when we extract and look at them in the laboratory, they become static,” said Dr Balleine.

That means the mRNA and protein profiles change in relation to each other as signals are “flashing on and off”.

The solution to all this might lie in taking a slightly broader view of cancer samples, and looking at the microenvironment as well as what is going on inside the cancer cells, she said.

“This brings to light that a real opportunity exists for a combined benefit of proteomics and histopathology in assessing cancer samples, because there’s an increasing interest in trying to characterise the tumour microenvironment,” she said.

“The whole-tissue view is extremely important and informative. And it’s never been more important as some of the therapies that are now being described as useful for individual patients are actually targeting aspects of the tissue microenvironment.”


End of content

No more pages to load

Log In Register ×