Finding the Primary Site of Metastatic 'Mucous Tumors' Using RNA Deep Learning

Professor Kim Gidong's Team at Bundang Seoul National University Hospital
Accuracy 85.7%... Expected to Establish Optimal Treatment Strategies

[Asia Economy Reporter Lee Gwan-ju] Domestic researchers have devised a method to identify the primary site of metastatic 'mucinous tumors.' Generally, mucinous tumors, which consist of sticky mucus rather than hard tumor masses, are difficult to trace back to their origin. Finding the primary site is expected to enable more effective treatment.

The team led by Professor Kim Ki-dong from the Department of Obstetrics and Gynecology at Bundang Seoul National University Hospital announced on the 18th that they succeeded in developing an algorithm to identify the primary site of mucinous tumors using RNA (ribonucleic acid) deep learning.

The research team focused on the fact that RNA expression patterns differ depending on the organ from which cancer cells originate and conducted a study to determine whether 'transcriptome analysis' could be used as an accurate diagnostic method. The transcriptome refers to the total sum of all RNA molecules present within a single cell.

Professor Kim Ki-dong, Department of Obstetrics and Gynecology, Bundang Seoul National University Hospital.

Based on transcriptome data from 1,960 cancer specimens, the research team trained machine learning models on RNA expression patterns that vary according to seven primary cancers: cervical cancer, endometrial cancer (uterine corpus cancer), ovarian cancer, uterine sarcoma, pancreatic cancer, gastric cancer, and colorectal cancer. As a result, they confirmed that the accuracy of identifying the primary site of mucinous tumors reached approximately 85.7%, which is twice that of existing methods.

This study is significant as it is the world's first to confirm that RNA analysis can be a breakthrough in the field of mucinous tumors, where it has been difficult to identify the primary site and establish optimal treatment strategies. Professor Kim said, "If the origin of cancer cells can be accurately identified, treatment strategies that improve patient prognosis can be established," adding, "We plan to develop this into a clinically applicable diagnostic method through follow-up studies."

The results of this study were recently published in the international academic journal Cancer Informatics, published by SAGE.

This content was produced with the assistance of AI translation services.