"AI Predicts Prognosis by Analyzing Tumor Morphology"... UNIST Develops Technology to Simulate Cancer Growth Environment

Development of 3D-Printed Artificial Tumor Tissue Mimicking High Stiffness and Hypoxic Environment by UNIST and Asan Medical Center Seoul
AI Analysis of Tissue Morphology Achieves 99% Accuracy in Predicting Specific Gene Expression, Published in Advanced Science

A 3D-printed artificial tumor tissue that can cultivate cancer cells derived from actual cancer patients in an environment that closely mimics in vivo conditions has been developed.

Additionally, artificial intelligence (AI) technology has been introduced that can predict patient prognosis simply by analyzing photos of the growth of this artificial tumor tissue.

The research teams led by Professors Taeun Park and Hyunwook Kang of the Department of Biomedical Engineering at UNIST, and Professor Seungjae Myung of Asan Medical Center Seoul, have developed an artificial cancer tissue called 'Eba-PDO' that replicates the high stiffness and hypoxic environment of real tumor tissue. When the morphology of this artificial cancer tissue is analyzed using AI, it can predict the expression of key prognostic marker genes for colorectal cancer with an accuracy of 99%.

Research team (from left) Professor Taeun Park of UNIST, Professor Hyunwook Kang, Professor Seungjae Myung of Asan Medical Center Seoul, Dr. Jonghyuk Han of UNIST (currently affiliated with Emory University School of Medicine), Dr. Hyejin Jung of UNIST (currently affiliated with IBS). Provided by UNIST

Cancer cells proliferate rapidly, resulting in high density, which makes the tissue stiffer than normal tissue, and they grow in an environment with insufficient oxygen. Although conventional artificial cancer tissues are made from cells taken from actual patients, they have not been able to fully replicate these conditions, leading to distorted patterns of cancer cell growth and drug response.

The research team developed a new artificial cancer tissue by mixing cancer organoids, which are three-dimensionally cultured from cancer cells taken from patients, with bioink and printing them in a bead-like arrangement. The bioink is formulated by combining gelatin and extracellular matrix components, enabling it to accurately recreate the stiff and hypoxic environment in which cancer grows.

The artificial cancer tissue grown using this method maintained a consistent shape for the same patient, but the size and morphology varied among different patients.

Differences between conventional cancer organoids and cancer organoid models 3D bioprinted in bead form.

Based on these characteristics, the team developed an AI that can predict the expression of the CEACAM5 gene from microscope images alone. CEACAM5 is a protein commonly found in solid tumors, including colorectal cancer, and is known to increase the likelihood of metastasis and resistance to anticancer drugs.

When this protein is overexpressed in the artificial cancer tissue, the cell-to-cell adhesion weakens, resulting in a less dense and more unbalanced tissue structure. The AI was trained to learn these morphological changes and predict gene expression levels accordingly.

Furthermore, the artificial cancer tissue achieved a 90% similarity in gene expression to actual patient-derived tumor tissue, which is a 29% improvement over the previous level of 70%. It also accurately replicated differences in responsiveness to the anticancer drug 5-fluorouracil (5-FU) among patients.

This research was led by Dr. Hyejin Jung and Dr. Jonghyuk Han of UNIST as co-first authors.

Patient-specific artificial tumor tissue bioprinting and AI for patient prognosis prediction based on it.

The research team stated, "By replicating and analyzing the growth of real cancer cells outside the body, this approach is expected to enable more precise, patient-specific treatments," adding, "In the future, we plan to expand this model into an even more sophisticated artificial cancer model by integrating immune cells and vascular structures."

The research was supported by the Korean ARPA-H project of the Ministry of Health and Welfare, the Alchemist Project of the Ministry of Trade, Industry and Energy, the Advanced Bio Technology and Human Resource Exchange Support Project of the Bio and Medical Technology Development Program, and the Glocal University Project (University of Ulsan) COMPaaS joint research of the Ministry of Education. The results were published online in the international journal Advanced Science on March 28.

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