By Diagnostics World Staff
November 19, 2024 | A team of researchers from Institute of Science Tokyo, Japan, are using nanowires with machine learning analysis to detect cancer-associated miRNAs in urine. Their research findings were published last month online in Analytical Chemistry (DOI: 10.1021/acs.analchem.4c02488).
Cancer cells use specific micro-ribonucleic acids (miRNAs), small noncoding RNAs, to regulate gene expression and promote tumor formation. While circulating miRNAs are viable biomarkers of early cancer disease, the identification of cancer-related miRNAs in blood and other body fluids remains a challenge.
Most microRNAs are encapsulated in extracellular vesicles (EVs) including exosomes, a unique subtypes of EVs. The team previously used zinc oxide nanowires to capture EVs based on surface charge and hydrogen bonding. They were successful, with the nanowires extracting massive numbers of miRNAs in urine.
Now the team has scaled their approach, comprehensively capturing EVs, including exosomes, in urine; extracting microRNAs from the captured EVs in situ; and differentiating between noncancer and lung cancer subjects through machine learning-based analysis.
The result could be a non-invasive cancer detection tool.
Initially, the scientists used ZnO nanowires to capture EVs in urine samples and incorporated microarray technology to identify specific gene sequences in EV-encapsulated miRNAs. An ultracentrifugation technique was used to compare and validate the efficiency of miRNA capture by nanowires. The results revealed that EVs containing miRNAs, including exosomes, were efficiently captured on nanowires. Moreover, the presence of 2,486 miRNA species was confirmed during the miRNA profiling analysis of 200 urine samples.
“The fact that urine has almost all human miRNA species led us to consider the following possible scenario: urinary miRNA species might be transferred from cancer tissues via blood circulation,” the authors write. The team hypothesized that most of the miRNAs in blood could be transferred to urine during the filtration process in kidneys, suggesting miRNA from cancers all over the body could be represented.
Subsequently, they employed a logistic regression classifier constructed using ML to identify lung cancer-associated urinary miRNA ensembles. The findings revealed one particular urinary miRNA ensemble, composed of 53 miRNA species, that could differentiate cancer and noncancer subjects with very high specificity and sensitivity.
The authors report distinguishing lung cancer and noncancer subjects with an AUROC of 0.997; even when the lung cancer was stage I, an AUROC of 0.987 was achieved. Furthermore, they identified miRNA ensembles to distinguish three classifications among brain tumor, lung cancer, and noncancer subjects with 86% sensitivity and 93% specificity.
The authors acknowledge that more samples and more trials are needed but say: “the present results are encouraging us to develop urine-based liquid biopsies for future medical applications and to develop urinary miRNA-based diagnoses for timely medical checkups of the presence of cancer presence.”