Lunit Reveals Validation Results of AI Model Predicting Lung Cancer Mutations at American Cancer Society Conference

Medical AI company Lunit announced on the 10th that it will present the performance validation results of a new AI model predicting non-small cell lung cancer (NSCLC) mutations at the American Association for Cancer Research (AACR 2023).

AI biomarker platform 'Lunit Scope IO'.
[Photo by Lunit]

원본보기 아이콘

The research team developed a new AI model to predict the 'KRAS G12C' mutation, the most common mutation among KRAS gene variants accounting for 25% of NSCLC patients. The KRAS G12C mutation has recently attracted academic attention as drugs targeting it have been approved.

The predictive model used in the study applied Lunit's AI algorithm based on Lunit Scope to pathological images of NSCLC from the U.S. National Institutes of Health big data (TCGA), showing improved accuracy compared to previously reported KRAS mutation prediction studies.

The results showed that the Lunit Scope KRAS G12C mutation prediction model demonstrated high predictive power with an area under the curve (AUC) of 0.787, indicating the diagnostic accuracy of the AI algorithm. Validation on independent external data also showed a predictive power of 0.745. According to Lunit, this AI-based mutation prediction model can provide predictive results before performing molecular genetic diagnostic methods, helping medical staff decide on additional tests and treatment methods.

The second research abstract presented validated the performance of 'UIHC (Universal Immunohistochemistry Analyzer),' an AI-based image analysis tool that detects and quantifies various target proteins characteristically expressed in diverse cancer cells. The research team developed UIHC trained on PD-L1 and HER2 immunohistochemistry (IHC) pathological slide data from multiple cancer types, including lung, bladder, and breast cancers. This study confirmed that UIHC detects and quantifies new target protein expressions across various cancer types better than models trained on single cancer type IHC pathological slides.

The company explained that the UIHC model shows potential for useful application in clinical research of anticancer drugs targeting new antigen proteins expressed in tumor cells in the future.

Thirdly, Lunit will present research results using Lunit Scope as a biomarker for treating patients with advanced biliary tract cancer (BTC) at the conference. The research team classified pre-treatment cancer tissue slide images of advanced BTC patients into three immune phenotypes?immune active, immune excluded, and immune deficient?using Lunit Scope. The results showed that the immune active group had the best response to immune checkpoint inhibitor therapy. The researchers concluded that the immune phenotypes classified by Lunit Scope could serve as effective biomarkers for predicting treatment outcomes in advanced BTC patients.

Seobum Seok, CEO of Lunit, said, "We are honored to announce at this year's AACR conference that Lunit Scope can be applied not only as a biomarker predicting responses to existing immune checkpoint inhibitors but also for mutation prediction in NSCLC and quantification of target antigens." He added, "We plan to actively expand the scope of research so that Lunit Scope can be widely applied to provide personalized treatment methods for cancer patients."

This year, the American Association for Cancer Research (AACR) will be held from the 14th to the 19th in Orlando, Florida, USA. AACR is considered one of the world's top three cancer conferences alongside the American Society of Clinical Oncology (ASCO) and the European Society for Medical Oncology (ESMO).

© The Asia Business Daily(www.asiae.co.kr). All rights reserved.