You are here:
DNA methylation analysis allows for better differentiation of tumors
Researchers from Charité – Universitätsmedizin Berlin and the German Cancer Consortium (DKTK) have developed a new diagnostic tool that can help differentiate between liver tumors. In this section, they answer questions about their research findings.
What was the research question or scientific inquiry behind your study?
Often, with existing methods, it is impossible to differentiate between a bile duct tumor of the liver (“cholangiocarcinoma”) and a liver metastasis, especially of a pancreatic cancer. Even by using modern molecular methods such as mutational analysis this diagnosis problem cannot be solved. The outcome of the diagnosis is crucial, however, as a primary bile duct tumor may be resected, while a metastasis is treated with palliative chemotherapy. Hence, we decided to develop a tool to help distinguish between the two cancers.
How did you approach the topic?
We used DNA methylation, a specific chemical modification of the genetic material, as input for developing machine learning classifiers that can help differentiate between primary and metastatic tumors of the liver. If we perceive the DNA as three billion letters from a book, then DNA methylation should be understood as the punctuation marks. Because it was well known that changes in DNA letters are not helpful for distinguishing the two tumor entities, we decided to focus on the DNA punctuation and developed three different AI classifiers based on these data.
What did you discover?
We discovered that DNA methylation is extremely powerful in helping us differentiate between primary liver carcinomas and pancreas metastases to the liver. Based on DNA methylation data, we were able to recreate the anatomy of the liver, bile ducts, and pancreas at an epigenetic level, and found that cholangiocarcinoma can be subclassified into two main subtypes: large and small bile duct types. We also observed that the neural network tool is superior to other AI methods.
Was there anything that surprised you?
Interestingly, by using DNA methylation data from Asia, we were able to characterize Fluke positive cholangiocarcinoma.
What’s your takeaway?
By analyzing DNA methylation patterns using a neural network we can accurately differentiate between primary and metastatic liver tumors. Our method can be used to support clinical practice and thus improve the treatment of affected patients.
Dragomir MP, Calina TG et al. DNA methylation-based classifier differentiates intrahepatic pancreato-biliary tumours. EBioMedicine 2023 Jun 20. doi: 10.1016/j.ebiom.2023.104657