Complete Loss of the Tumor Suppressor Gene RB1 Identified as a Poor Prognostic Factor in Liver Cancer

Highly Accurate AI Diagnostic Model Developed, with Combination Therapy Using Cell Division and PARP Inhibitors Proposed

Study Published in a Top 0.2% Global Journal with an Impact Factor of 81.2

▲ (From left) Professor Ju Hyun Shim of the Division of Gastroenterology and Professor Chang Ohk Sung of the Department of Pathology at Asan Medical Center, Professor Joong Sup Shim of the University of Macau, Professor Sang Hyun Park of Pohang University of Science and Technology (POSTECH), and Professor Jihyun An of Hanyang University Guri Hospital

A new treatment strategy has been proposed for patients with refractory liver cancer, a difficult to treat disease with a low survival rate, using artificial intelligence for rapid pathological diagnosis and personalized targeted therapy.

A joint research team led by Professor Ju Hyun Shim of the Division of Gastroenterology and Professor Chang Ohk Sung of the Department of Pathology at Asan Medical Center, Professor Joong Sup Shim of the University of Macau, Professor Sang Hyun Park of Pohang University of Science and Technology (POSTECH), and Professor Jihyun An of Hanyang University Guri Hospital identified complete loss of the tumor suppressor gene RB1 as a new biomarker for liver cancer. The researchers also developed an AI based diagnostic model targeting this biomarker and found that combination therapy with a cell division inhibitor and a PARP inhibitor produced a strong anticancer effect.

The findings were published in the latest issue of Signal Transduction and Targeted Therapy, a leading global journal in translational medicine with an impact factor of 81.2, which ranks in the top 0.2% of journals according to the Journal Citation Reports (JCR).

Current treatment options for advanced liver cancer include combination therapy with atezolizumab and bevacizumab. However, these treatments have limitations, as only some patients respond and resistance eventually develops. In addition, only a small proportion of patients have genetic alterations that respond to approved targeted therapies, highlighting the urgent need to identify biomarkers that can enable personalized treatment.

Previous studies of patients with liver cancer found mutations in the RB1 gene, which acts as a brake to prevent excessive cell proliferation that can lead to cancer. However, these studies did not clearly distinguish between monoallelic loss, in which one copy of the RB1 gene is damaged, and biallelic loss, in which both copies are lost. Monoallelic and biallelic loss can differ substantially in their effects on cancer progression and treatment response. As a result, the biological and clinical significance of RB1 loss itself remained unclear.

The research team assembled a study cohort of 561 patients with liver cancer, comprising 206 patients from Asan Medical Center and 355 patients from The Cancer Genome Atlas (TCGA), a large publicly available cancer research database maintained by the U.S. National Cancer Institute. The researchers conducted multiomics profiling, integrating genomic and transcriptomic data through whole exome sequencing and RNA sequencing.

To further enhance the objectivity and reliability of the findings, the team validated the results in an additional cohort of 450 patients with diverse disease stages and treatment backgrounds, using genomic, single cell, and spatial transcriptomic data.

The analysis showed that patients with complete RB1 loss (RB1 Bi), in which both copies of the RB1 gene were deleted or inactivated, accounted for approximately 14.6% of all liver cancer cases. This patient group showed poorer tumor differentiation and more rapid tumor progression. Compared with other patients, they had a 3.32 fold higher risk of death and a 3.15 fold higher risk of recurrence. These findings confirmed that complete RB1 loss is an independent poor prognostic factor in liver cancer.

The research team developed a deep learning based pathology AI model (FR MIL) to rapidly identify this high-risk group in clinical practice without the need for costly and complex genomic analyses. The model predicts complete RB1 loss using only routinely stained histopathology slide images. It demonstrated high accuracy in an external validation cohort, achieving F1 scores ranging from 84.39% to 91.58%.

The researchers further cultured liver cancer cell lines with defective RB1 genes (Huh7, PLC/PRF/5, and HepG2) and treated them with 876 different drugs, including epigenetic agents, kinase inhibitors, and highly selective inhibitors. They found that RB1 deficient liver cancer cells were selectively killed by certain drugs that block cell division or DNA repair, including PARP inhibitors. While normal cells remained viable when exposed to these drugs, cancer cells with defective RB1 genes underwent synthetic lethality when the additional drug induced a second defect, ultimately causing the cancer cells to die.

In animal studies using cancer cells and mice, the research team applied a combination therapy pairing a cell division inhibitor with a PARP inhibitor. The treatment produced a powerful anticancer synergy, maximizing tumor suppression without significant systemic side effects such as liver toxicity.

This study is significant in that it represents a new milestone in multidisciplinary translational research, covering the entire process from identifying novel therapeutic targets through omics analysis and developing AI based pathology diagnostics to establishing a personalized combination targeted therapy strategy for patients who have shown limited responses to existing standard anticancer treatments. The approach could offer a new treatment pathway for high-risk liver cancer patients with extremely poor prognoses.

Professor Ju Hyun Shim of the Division of Gastroenterology at Asan Medical Center, University of Ulsan College of Medicine, said, “It is encouraging that this study has opened up a new therapeutic avenue tailored to the genetic characteristics of patients with refractory liver cancer, who currently have limited treatment options and are prone to developing resistance. If the AI diagnostic model and combination treatment strategy developed in this study are introduced into clinical practice, they are expected to make a significant contribution to substantially improving patient survival and advancing precision personalized medicine.”