A new study published in the Journal of Investigative Dermatology reports an inexpensive, more accurate, and much less invasive way to diagnose cancerous melanoma cells in skin tumors, even when present at low levels. This is achieved with the help of machine learning used to detect malignant features in a melanocytic tumor using specific microRNA (miRNA) patterns.
Pixel-Shot | Shutterstock
Melanoma is a deadly form of skin cancer, though thankfully not the most common one. It arises from a cancerous change within the pigment-producing cells of the skin, or melanocytes. It is well known to be increased in fair-skinned people who are exposed to bright sunlight for long periods of time, due to ultraviolet-induced damage to the DNA of skin cells. It accounts for 1 in 100 skin cancers, but this small percentage is still responsible for most of the cancer-related deaths in this group. With advanced melanoma, less than one in five patients survive as long as five years, on average. Every year, over 10,000 people die of this cancer within the US alone. However, if caught early, this tumor is curable.
This type of cancer often looks innocent at its onset, with nothing more than a little discoloration or an increase in the size of a mole.
Discordance rates across dermatopathologists and care centers are high. Consequently, whether a patient is accurately diagnosed early can depend on where the patient lives.”
Researcher Robert L. Judson Torres
In other words, doctors disagree whether a given biopsy is malignant or not, which means the patient has a high chance of a missed diagnosis. This is one of the most common reasons underlying litigation alleging medical malpractice in the US at present.
This is an unfortunate situation for the patient, to say the least, and the ability to diagnose this tumor quickly and accurately remains a primary goal in skin medicine. This motivated the present attempt to look at miRNA.
What are miRNAs?
A miRNA is a short stretch of non-coding RNA, that is, RNA that does not encode a specific protein. RNA is a faithful copy of the DNA encoding a complete protein, which can be transferred out of the cell nucleus into the cytoplasm to actually produce proteins for the cell (‘translation’).
The miRNAs act to stop the production of protein by the RNA as and when indicated. This is typically by binding to the part of the RNA which is not involved in this protein encoding, resulting in a loss of stability of the RNA molecule and the cessation of translation. This function helps to fine-tune several major cell processes, including cell metabolism and growth, proliferation and differentiation.
The kinds of miRNAs present in a specific cell present a different pattern from those seen in other cells. As a result, the miRNA expression profile varies between tissues as well. This knowledge can help to tell which tissue a tumor originates from, making the miRNAs valuable biomarkers in cancers and many other disease conditions.
Unfortunately, a barrier to the use of miRNAs in the diagnosis of melanomas is the lack of agreement as to the exact criteria which differentiate benign and malignant melanocytic growths, that is, a benign mole as opposed to a malignant melanoma. One reason is that researchers so far have looked at the expression of miRNAs linked to specific genes. It is difficult to produce a standard pattern for melanoma diagnosis in this way because of the huge differences between tumors and the tissues in and around them.
There are over 500 miRNAs present at higher-than-usual concentrations in melanocytic lesions, as observed in several studies. These are seen in benign and cancerous tumors at various stages from early to advanced. Though many studies have proposed diagnostic sets of miRNAs, few have stood up to external validation so far.
The current research focused on a new strategy to evolve an objective and reproducible diagnostic score that would hold good no matter who did the testing or where it was done. This involved using machine learning to identify miRNAs.
The first step was to find all the differences in tumor characteristics and patient age groups that were most likely to bias the results when using miRNA-based tools. They then pruned the long list of melanocytic tumor-specific miRNAs to just six, to arrive at a set that could accurately distinguish benign from malignant lesions over a broad range of data sets as well as data collected in a variety of ways.
They then found eight expression ratios for miRNAs that could detect malignant melanocytic cells, but were not significantly affected either by the age of the patient or by the presence of a large amount of benign cells in the same tumor.
The current study included 82 biopsy specimens of moles and malignant melanomas, 41 of each type, taken from the medical records of the San Francisco Dermatopathology Section of the University of California. The scientists used the new method to detect malignant melanoma cells in the samples. They compared their predictions with the actual recorded outcomes. They achieved a sensitivity of 81% and a specificity of 88%. This means that 81% of tumors were detected, and out of the samples in which no tumor was predicted, there was indeed no tumor in 88% of cases.
They also observed that neither age nor other cells in the sample affected the results.
We found that by developing a classifier based on a ratio of diagnostically important miRNA we could provide a more robust biomarker that was less susceptible to changes in tumor cell content and platform allowing for a test that could be used on a greater variety of patient samples.”
Researcher Rodrigo Torres
The advantages of using miRNAs to distinguish benign and malignant melanocytic tumors include the fact that that they are easy to obtain from body fluids, are stable, inexpensive to measure and do not require very invasive techniques or a large amount of tissue. A very important benefit of this test is its ability to pick up just a few malignant melanoma cells scattered through a large volume of other benign cells.
The researchers look forward to having their results confirmed in full clinical studies to extend these benefits to all patients regardless of location.
In addition to improving the diagnostic accuracy for melanoma, this technique also has the potential to help catch melanomas earlier, when the tumor is entirely curable, which would significantly impact patient care.”
Researcher Maria L. Wei
Rodrigo Torres, Ursula E. Lang, Miroslav Hejna, Samuel J. Shelton, Nancy M. Joseph, A. Hunter Shain, Iwei Yeh, Maria L. Wei, Michael C. Oldham, Boris C. Bastian, and Robert L. Judson-Torres - MicroRNA Ratios Distinguish Melanomas from Nevi - https://www.jidonline.org/article/S0022-202X(19)31788-9/fulltext