Recent & Upcoming Talks | Ki Wook Lee

Poster Presentation in 2026 AACR (#3594)

Mon, 20 Apr 2026 09:00:00 +0000

🏆 Achievement & Recognition I am deeply honored to announce that I have been selected as a recipient of the AACR-Margaret Foti Foundation Scholar-in-Training Award for the AACR Annual Meeting 2026. This prestigious award recognizes highly meritorious abstracts submitted by early-career scientists.

Poster Presentation Details

Abstract Title: Expression-based immune-phenotyping ML model predict ICI response and long-term clinic benefit in lung adenocarcinoma
Abstract Number: #3594
Session: Integrative Computational Approaches 1
Date & Time: Monday, April 20, 2026 | 9:00 AM – 12:00 PM
Location: Poster Section 5 - Board Number 4, San Diego Convention Center

Research Overview In this study, I present a machine learning-based framework that leverages transcriptomic immune-phenotyping to predict responses to Immune Checkpoint Inhibitors (ICI). Our model not only predicts immediate response but also identifies patients likely to experience long-term clinical benefits, providing a robust tool for precision oncology in lung adenocarcinoma.

Acknowledgements This research and my participation in the AACR Annual Meeting 2026 are generously supported by:

AACR-Margaret Foti Foundation (Scholar-in-Training Award)
National Research Foundation of Korea (NRF)
Hyundai Motor Chung Mong-Koo Foundation (On-dream Future Industrial Talent Scholarship)

I look forward to engaging with fellow researchers in San Diego. Please feel free to stop by my poster to discuss bioinformatics, AI in oncology, or potential collaborations! ✈️🇺🇸

Selected Oral Presentation (Young Scientist Session)

Thu, 05 Feb 2026 09:00:00 +0000

🏆 Recognition

Young Scientist Presentation Award, The Korean Genome Organization (KOGO), Feb 2026.
Selected as an Oral Presenter for the Young Scientist Session.

**Presentation Title** > **Transcriptome-driven Immune Phenotyping AI Framework Predict ICI Response and Long-term Benefit in Lung Adenocarcinoma**

**Highlights** As a selected presenter, I presented my research at the KOGO Winter Symposium.

Objective: To improve the accuracy of predicting clinical responses to Immune Checkpoint Inhibitors (ICI).
Methodology: Integration of RNA-seq data preprocessing optimization with advanced deep learning models.
Key Finding: Established a robust immune phenotyping framework that significantly correlates with long-term therapeutic benefits in lung adenocarcinoma cohorts.

Oral Presentation in UKC 2025, Atlanta

Sat, 09 Aug 2025 08:30:00 +0000

Title
Transcriptomic AI-based classification of immune phenotypes for predicting immunotherapy response in lung cancer

Main
With full support from the Hyundai Motor Chung Mong-Koo Foundation, I will attend to UKC 2025 as a presenter at the Medical and Pharmaceutical Science symposium.
Additionally, to gain the opportunity to engage with senior experts and receive valuable advice for establishing myself as an independent researcher, I will participate in the SCIENTISTS AND ENGINEERS EARLY CAREER DEVELOPMENT (SEED) 2025 program.

Poster Presentation in 2025 AACR (#7263)

Wed, 30 Apr 2025 09:00:00 +0000

Background
Tumor immune microenvironment composition significantly influences immunotherapy outcomes in lung cancer, with immune-inflamed tumors demonstrating higher response rates compared to immune-non-inflamed tumors. This study aimed to develop a machine learning framework to classify immune phenotypes in lung cancer using TCGA data and validate its utility in predicting immunotherapy response with real-world patient samples. Existing approaches, such as the Lunit SCOPE technique, have been employed for immune phenotype characterization but face limitations in handling tumor heterogeneity and adapting to diverse datasets, underscoring the need for more robust tools.

Methods
A baseline machine learning model was constructed using six tree-based classifiers: Random Forest, Extremely Randomized Trees, AdaBoost, Gradient Boosting, XGBoost, and LightGBM. Feature importance scoring identified critical immune-related gene expression profiles, including markers from interferon signaling, T-cell activation, and cytokine regulation pathways. Model optimization incorporated 15 classifiers, leveraging advanced hyperparameter tuning and ensemble learning strategies to improve accuracy, Matthew’s correlation coefficient (MCC), and area under the ROC curve (AUROC). Classification training and validation were performed on 447 TCGA lung cancer samples, divided into 80% training and 20% validation subsets, stratified into immune-inflamed and immune-non-inflamed phenotypes based on bulk RNA-seq gene expression datasets.
The final model demonstrated robust adaptability to tumor heterogeneity and was independently tested on 87 patient samples labeled through the Lunit SCOPE framework, stratified into immunotherapy responders and non-responders. Correlation analysis assessed the relationship between immune phenotypes and therapeutic outcomes. The optimized model exceeded the performance of Lunit SCOPE, achieving an AUROC >0.90 in classifying immune-inflamed versus immune-non inflamed samples in the TCGA dataset. Testing on real-world patient samples further confirmed its ability to accurately predict immunotherapy response, supporting the hypothesis that immune-inflamed tumors derive greater benefit from treatment. This machine learning framework demonstrates superior accuracy and adaptability compared to existing techniques, offering significant potential to improve patient stratification, reduce ineffective treatments, and guide precision immunotherapy in lung cancer.

Conclusion
Future directions include integrating multi-omics datasets to further refine predictive capabilities and assess its generalizability across other cancer types.

Poster Presentation in 2025 AACR (#2427)

Mon, 28 Apr 2025 09:00:00 +0000

Background
Current research in medicine and biology predominantly focuses on applying artificial intelligence (AI) to analyze patient-specific imaging data, such as CT and MRI scans. However, the application of AI to next-generation sequencing (NGS) data, particularly transcriptomics, remains underutilized. RNA sequencing data is prone to biological and technical variability during generation, and the choice of error-correction and preprocessing methods can significantly influence downstream analyses. These challenges underscore the need for clear and systematic guidelines to optimize transcriptome data for AI-based analyses. In this study, we aim to identify optimal preprocessing strategies for developing AI models leveraging transcriptomic data.

Methods
We utilized RNA sequencing datasets from 25 independent cohorts, comprising over 5,800 patients with conditions such as lung adenocarcinoma, colorectal cancer, diabetes, and other diseases. A total of 18 combinatorial preprocessing pipelines were systematically assessed, encompassing 6 normalization methods (Raw, CPMTMM, RLE, UQ, RPKM, TPM) and 3 scaling methods (None, MinMax, Z-score). Over 20,000 transcripts were processed using each combination of normalization and scaling techniques. We then developed disease diagnosis models using 13 machine learning and deep learning algorithms. The performance of these models was evaluated to identify the most effective preprocessing strategies for transcriptomic data.

Conclusion
We identified algorithm-specific optimal preprocessing strategies to develop robust AI models utilizing transcriptomic data. Performance variations driven by normalization and scaling methods were observed for each algorithm, with these differences being particularly pronounced in datasets characterized by inherently lower model performance (difficult tasks). Our findings underscore the critical role of preprocessing in shaping model outcomes and provide a foundation for the development of tailored AI frameworks for disease diagnosis. This study offers a systematic approach to optimize transcriptomics-based predictive modeling, advancing the integration of AI into transcriptomic data analysis for clinical applications.