Latest research! New artificial intelligence model can predict the effect of brain glioma treatment

Gliomas are tumors that originate from brain glial cells and are the most common primary intracranial tumors in adults. The annual incidence of gliomas in China is 5 to 8 cases per 100,000 people, and the 5-year mortality rate is second only to pancreatic cancer and lung cancer among systemic tumors. However, the underlying mechanism of glioma is still unclear. The two risk factors currently identified are: exposure to high-dose ionizing radiation and high-penetrance genetic mutations associated with rare syndromes.

The standard treatment for glioma is mainly surgical resection, combined with radiotherapy and the use of the chemotherapy drug temozolomide (TMZ) for comprehensive treatment. Surgery can relieve clinical symptoms and prolong survival, but it can only extend the patient's life by about three months because almost all patients will face the problem of glioma recurrence. The medical community has not yet clarified why this standard treatment causes glioma to worsen.

Image source: unsplash.com Photographer: Robina Weermeijer

To solve this mystery, a research team led by Professor Ji-Guang Wang, Associate Professor of Life Sciences and Professor of Life Sciences at the Hong Kong University of Science and Technology's Division of Life Sciences and Department of Chemical and Biological Engineering, comprehensively analyzed tumor molecular samples and clinical data of 544 glioma patients, including 182 East Asian patients, to identify genomic and transcriptomic predictors of the evolution of different types of gliomas. The relevant results were published in Science Translational Medicine.

Through big data analysis, the research team found some early predictors related to TMZ resistance and rapid deterioration of glioma. For example, when patients were first diagnosed, they were found to have an increased number of certain genes, or a certain gene was over-stimulated, or a certain gene was missing. Their chances of rapid deterioration of the tumor in the later stage were higher. The discovery of these early predictors of glioma recurrence will help develop precise treatment plans for this malignant tumor and benefit patients.

The team further found that East Asians have significantly different gene mutations in brain tumors than Caucasians. For example, East Asian patients are less likely to have chromosome 7 amplification and chromosome 10 deletion in brain tumors, but are more likely to have MYC replication gene amplification. There is a high-risk factor for glioma, rs55705857 (G), which is more commonly found in Caucasian patients, but rarely appears in East Asian populations. Professor Wang Jiguang said that these research results confirm the importance of developing personalized treatment plans for cancer patients.

Image source: National Cancer Institute

In order to better evaluate the progress and results of patients' treatment, the team also developed an artificial intelligence model called CELLO2 to evaluate the patient's condition after the initial diagnosis. The trained model can accurately predict whether recurrent tumors will worsen under TMZ chemotherapy and identify high-risk patients. This model is placed on a public interactive website (CELLO2) and is open to the public. The website also provides patients and doctors with a long-term tracking database of brain gliomas.

Professor Jiang Tao, co-first author of the paper, Professor at Beijing Tiantan Hospital, Capital Medical University and Director of Beijing Institute of Neurosurgery, said: "CELLO2 is the first effective tool that can predict whether the grade of recurrent tumors is elevated or drug-resistant based on the molecular characteristics of the primary tumor, providing an important reference for clinical management of patients and estimating patient prognosis."

In the future, the team led by Professor Wang Jiguang will continue to integrate more patient data, further optimize the machine learning model, and plan to develop an artificial intelligence platform that integrates medical imaging and multi-omics data to promote the development of precision neuro-oncology.

Image source: unsplash.com Photographer: Markus Spiske

Planning and production

Author: Zeng Xinyue, popular science creator

Reviewer: Tang Qin, Director and Researcher of the Science Popularization Department of the Chinese Medical Association

Editor: Qi Yuan