Search

2570 Background: TIGIT is a promising emerging immunotherapeutic target. However, the specific sources of TIGIT expression within the tumor microenvironment are largely unknown. Here, we present an AI-powered spatial tumor-infiltrating lymphocyte (TIL) analyzer, Lunit SCOPE IO, to integrate image analysis from whole slide images with single-cell molecular profiling. Methods: We used The Cancer Genome Atlas (TCGA) RNA expression data across 23 cancer types (n=6,930). Lunit SCOPE IO was developed, trained, and validated based on >17k H&E whole-slide images, to segment cancer area (CA) and cancer-associated stroma (CS) and to detect tumor cells and TILs. The intra-tumoral TIL, stromal TIL, and tumor cell purity (TCP) in the CA+CS area were calculated. The public spatial transcriptomics (ST) dataset for breast cancer was downloaded from the 10X Visium web page. Lunit SCOPE IO was applied to the associated H&E WSIs to match distinct TIGIT expression to single cells identified in the WSIs. Results: TIGIT was highly expressed in TGCT (3.45±0.11; median±SEM), LUAD (3.07±0.05), and HNSC (2.89±0.06), and was highly enriched in samples with microsatellite instability-high or tumor mutational burden-high (≥ 10/Mb) compared to those without them (fold change = 1.30, p < 0.001). At a macroscopic, bulk-level in the TCGA dataset, TIGIT expression was positively correlated with intra-tumoral TIL density (R=0.37, p

It has been a common approach to apply Machine Learning (ML) techniques over sensory data for inferring human behavior, activities, emotions, and surrounding contexts. Especially, IMU (Inertial Measurement Unit) sensors are widely used to obtain dataset to train ML models for human activity recognition. A key challenge in building highly accurate ML models lies in collecting a wide variety of activity data from a large number of users. Such data collection is often a highly time-consuming and costly process.

Machine Learning (ML) models are widely used to infer human activities. However, collecting data to train ML models in realworld often requires significant time and effort. In this paper, we suggest a novel data collection framework to leverage pre-built VR applications and simulating tools. In particular, we applied the concept of virtual Inertial Measurement Unit (IMU) to capture activities of an avatar in simulation. Our initial results show that Random Forest (RF), Support Vector Machine (SVM), and Long Short Term Memory (LSTM) models built with the virtual sensor data can classify three activities (i.e., standing, running, walking) over a realworld dataset at the accuracy of 80.40% (87.83% precision and 80.12% recall), 67.52% accuracy (72.24% precision and 68.15% recall), and 77.67% accuracy (86.25% precision and 77.63% recall), respectively. The early results show the initial feasibility of simulation-driven machine learning without real-world data