Swedish scientists have found structural changes in sugar molecules called ‘glycans’ that occur in cancer cells can help identify specific types of cancers in early stages.
The team from the University of Gothenburg believes that – with the help of artificial intelligence – this research could eventually be used to develop a simple blood or saliva test for detecting cancer.
In 2020, about 19.3 million new cancer case were diagnosed worldwide ‘and that number is expected to reach 28.4 million by 2040’, noted medicalnewstoday.com. And ‘for most cancers the earlier it is detected and treated, the better the outcome.
‘For this reason, scientists are constantly exploring new ways to identify cancer quickly.’
Lead author Dr Danniel Bojar, whose study was published in the journal Cell Reports Methods, explained glycans are complex sugar molecules attached to proteins and fats in our bodies.
He said: ‘If those connections are changed, the function of the glycan changes. In cancer, several processes can change glycans.
‘Mutations in the tumour may change the proteins that build up these sugar chains, leading to altered glycans. Additionally, inflammation and various other systemic conditions that may accompany a tumour also have a known impact on which glycan structures are being produced.’
For this study, Dr Bojar’s team analysed tumour and healthy tissue data from 220 people with diagnosed cancers, focusing on gastric, skin, liver, prostate, colorectal and ovarian cancers.
Using a new method of studying the substructures of glycan using AI, the scientists were able to identify differences in the substructure of the glycan depending on the type of cancer.
Dr Bojar explained: ‘Glycans are structurally very complex, much more so than proteins or DNA, and are best understood with advanced analytical methods such as AI.
‘Further, the way glycans are currently measured – via mass spectrometry – typically leads to very heterogeneous data, including missing data points due to lack of sensitivity. This has really hindered the field from making robust assessments from this type of data in many cases.
‘Methods such as AI allow us to improve the data quality – which we have shown in the paper describing this method – and this allows us to identify these relevant substructures with high statistical significance.’
