Precise Analysis of Pancreatic Cancer Fibroblast Subtypes Using Label Free Raman Spectroscopy and Artificial Intelligence

Research Team Led by Professors Jun Ki Kim and Eunsung Jun of Asan Medical Center Publishes Paper in Biomaterials Journal With a Five Year Impact Factor of 12.5

“This research will greatly contribute to establishing personalized treatment strategies by enabling patient specific evaluation of fibroblast characteristics in the future.”

▲ (From left) Professor Jun Ki Kim of the Department of Convergence Medicine at Asan Medical Center, University of Ulsan College of Medicine, together with Professor Eunsung Jun of the Division of Hepatobiliary and Pancreatic Surgery at Asan Medical Center, and researchers Minju Cho, a doctoral candidate, and Eun-Young Koh from the Department of Convergence Medicine at Asan Medical Center

Pancreatic cancer is known to be influenced not only by the tumor itself but also by the surrounding tumor microenvironment, particularly fibroblasts, which play a significant role in treatment response and recurrence. As it has become clear that these fibroblasts consist of multiple subtypes with distinct functions, identifying and modulating each subtype has emerged as a key challenge in pancreatic cancer treatment.

In response, a Korean research team has successfully achieved precise discrimination of fibroblast subtypes within the pancreatic cancer microenvironment using label-free Raman spectroscopy.

Professor Jun Ki Kim of the Department of Convergence Medicine at Asan Medical Center, University of Ulsan College of Medicine, together with Professor Eunsung Jun of the Division of Hepatobiliary and Pancreatic Surgery at Asan Medical Center, and researchers Minju Cho, a doctoral candidate, and Eun-Young Koh from the Department of Convergence Medicine at Asan Medical Center, quantitatively analyzed the molecular compositions of pancreatic cancer associated fibroblast subtypes.

The team studied stellate cells, which are normal cells constituting the pancreatic cancer microenvironment, as well as inflammatory fibroblasts and myofibroblasts differentiated from these cells. Using a Raman spectroscopic microscope and an artificial-intelligence-based classification algorithm, they precisely characterized each cell subtype without the use of fluorescent labeling or staining.

This study is expected to provide a crucial foundation for establishing personalized treatment strategies for pancreatic cancer patients, including the prediction of treatment responses based on fibroblast characteristics.

The research findings were published in the latest issue of the prestigious journal ‘Biomaterials Research’ (five-year impact factor 12.5), which covers the fields of chemistry, life sciences, and medicine.

Pancreatic cancer typically shows few symptoms in its early stages. By the time it is detected, the cancer has often already spread to surrounding organs, limiting the number of patients eligible for curative resection. Conventional diagnostic methods, which rely on imaging studies and blood tumor markers, focus primarily on tumor size or marker levels in the blood. As a result, they are limited in reflecting the complex changes within the tumor microenvironment and in precisely predicting treatment responses.

Patients with the same stage of pancreatic cancer can show different treatment responses and recurrence patterns, largely due to the tumor microenvironment. In particular, variations in the composition and function of fibroblast subtypes are believed to influence patient outcomes. This has driven growing demand for precision diagnostic technologies that assess not only tumor cells but also the surrounding stromal cells.

The research team first used spatial transcriptomic analysis to observe the distribution of cancer-associated fibroblast (CAF) subtypes in human pancreatic cancer tissues. They found that inflammatory fibroblasts (iCAFs), which secrete pro-inflammatory factors that promote tumor growth, and myofibroblasts (myCAFs), which produce fibrotic and stiff tissue that physically shields the tumor, occupy distinct locations within the tissue. Additionally, they confirmed that iCAFs exhibit prominent expression of genes related to inflammatory lipid metabolism, while myCAFs show elevated expression of genes associated with collagen production and the extracellular matrix.

Next, the team differentiated iCAFs and myCAFs from pancreatic stellate cells derived from the same individuals and extracted Raman spectra from each fibroblast subtype. Using machine learning techniques such as principal component analysis and partial least squares discriminant analysis, they trained an artificial-intelligence algorithm to analyze the chemical “fingerprints” distinguishing the two fibroblast subtypes.

As a result, clear differences between the two fibroblast subtypes were observed in Raman spectral regions reflecting collagen and protein, as well as lipid signals. The artificial intelligence classification algorithm was able to distinguish the two cell types with an accuracy of 99 percent.

Professor Jun Ki Kim of the Department of Convergence Medicine at Asan Medical Center, University of Ulsan College of Medicine, stated, “Previously, distinguishing fibroblast subtypes required invasive methods such as fluorescent antibody staining. The significance of this study lies in demonstrating that label-free Raman spectroscopy combined with artificial-intelligence analysis can read the chemical fingerprints within cells and classify fibroblast subtypes with very high accuracy. In the future, we plan to apply label-free Raman spectroscopy to tissue biopsies and cell cultures to develop a platform for precise diagnosis of the tumor microenvironment and the establishment of personalized treatment strategies.”

Professor Eunsung Jun of the Division of Hepatobiliary and Pancreatic Surgery at Asan Medical Center, University of Ulsan College of Medicine, said, “In pancreatic cancer patients, not only the tumor itself but also the surrounding stromal environment—particularly the aggressiveness of fibroblasts—greatly influences tumor progression, treatment response, and prognosis. The Raman- and AI-based fibroblast subtype analysis we have developed is expected to serve as an important indicator for evaluating patient-specific fibroblast characteristics before and after surgery, as well as for guiding the selection of targeted chemotherapy and immunotherapy combinations.”

This study was conducted with support from the National Research Foundation of Korea under the Ministry of Science and ICT, the Bio & Medical Technology Development Program, and the Asan Institute for Life Sciences.