Lung cancer interception is critical, yet current risk models for ground glass nodules (GGNs), particularly partsolid nodules (PSNs), are limited. The development of robotic assisted bronchoscopy (RAB) systems enables more accurate and safer sampling of small nodules located throughout the periphery of the lung. These systems may also provide a valuable opportunity to obtain biopsies from persistent PSNs at higher risk of malignancy and guide clinical management. However, the diagnostic yield of PSN biopsies has not been extensively assessed in the setting of newer RAB systems and the clinical and imaging factors that affect the diagnostic yield of PSNs has not been quantitatively defined. In Aim 1, we will assess the diagnostic yield PSN biopsies obtained from RAB and determine the clinical and imaging factors affecting RAB diagnostic yield using a multi-institutional cohort. Other major challenges include how to handle non-diagnostic biopsies obtain from any type of bronchoscopic procedure. In Aim 2, we will develop integrated models to predict risk of malignancy in nondiagnostic and premalignant biopsies from PSNs. Our integrative models will use spatial transcriptomics, radiomics, and deep learning scores from existing models to distinguish true malignancy from false negatives and to predict likelihood of growth in the future. This innovative work will be the first to thoroughly assess RAB performance for PSNs and to integrate spatial transcriptomics with radiomics for PSN diagnosis and prognosis. This study has the potential to reduce false negative rates, identify high-risk PSNs, and establish a new paradigm for GGN management.