Location, location, location: this guiding principle for homeowners when purchasing real estate also matters when treating cancer. Thanks to advanced biomedical tools, we can now capture the fine details of cell activity—from genes expressed to chemicals secreted—across different parts of a tumour. Such location-based data can help clinicians better predict a cancer’s risk of recurring and adjust treatment plans accordingly.
That risk is particularly high in hepatocellular carcinoma (HCC), the most common form of liver cancer. “An estimated 70 percent of HCC cases see new tumours emerge after the initial ones are surgically removed,” said Joe Yeong, Group Leader at the A*STAR Institute of Molecular and Cell Biology (A*STAR IMCB).
To change that statistic, researchers are eyeing immune cells known as natural killer (NK) cells. Studies have shown that higher levels of NK cell infiltration and activity within HCC tumours are tied to improved survival rates. The spatial distribution and expression patterns of these cells also seem to affect recurrence—but it’s not clear how.
“Current prediction tools don’t capture the full biological complexity of the tumour microenvironment,” said Yeong. “They mainly stratify recurrence risk based on microscopic examinations of tumour tissue sections.”
Seeking better tools, Yeong and A*STAR IMCB colleagues teamed up with Nye-Thane Ngo, Tony Lim and colleagues from Singapore General Hospital and Duke-NUS Medical School, Singapore; Cheng Sun, Gengie Jia and colleagues from the University of Science and Technology of China; as well as other institutes in Singapore and China to perform spatial multi-omics analyses on tumours from 61 patients with HCC.
The researchers analysed over 75,000 spots from tumour tissue sections, looking at gene and protein expression levels as well as their spatial contexts. By comparing patients with recurrent versus non-recurrent HCC, they found that an increased presence of NK cells at a tumour’s invasive front was linked with a better prognosis.
Zooming in on a subset of eight patients, they then analysed the spatial distribution of 18,677 genes in NK cells that had infiltrated tumours. They found five genes (SPON2, ZFP36L2, ZFP36, VIM, HLA-DRB1) whose corresponding protein levels and expression patterns could most accurately predict recurrence risk.
Based on their data, the team developed the tumour immune microenvironment spatial (TIMES) scoring system. Powered by artificial intelligence (AI), TIMES analyses histopathological images of tumours to map spatial expression patterns of the five identified biomarker genes, then generate a personalised recurrence risk score.
“Compared to 118 clinical factors, including established predictors, TIMES showed significantly stronger associations with disease-free survival and recurrence, even when the patient cohort was first stratified into subgroups based on established risk factors,” said Yeong.
Yeong added that TIMES’ strength lies in its integration of spatial immune information; the predictive power of the five biomarkers emerged only when their spatial context was considered.
Moving forward, the team hopes to further leverage AI and build an accelerated pipeline from biomarker discovery to assay development, potentially enabling rapid and personalised treatment selection for patients.
The A*STAR-affiliated researchers contributing to this research are from the A*STAR Institute of Molecular and Cell Biology (A*STAR IMCB).
