PROJECT SUMMARY Histology is the current standard for diagnosis and predicting long-term disease outcomes in lupus nephritis (LN). However, diagnosis and prognosis are challenging due to significant inter-pathologist variance and multiple pitfalls in histopathology. We propose combining conventional histology with independent information from two complementary optical imaging modalities that provide additional morphological, biochemical and molecular context to LN, thus overcoming current diagnostic challenges. We will utilize milling with ultraviolet surface excitation (MUSE) to provide protein-specific histology and mid-infrared spectroscopic imaging (MIRSI) for label- free biochemical identification of small molecules and metabolites. Acquiring co-registered imaging data with high speed and good resolution from these imaging modalities is challenging, and we propose a new experimental platform for comprehensive biopsy imaging that addresses this challenge. We will identify new structural and molecular features across these modalities that are decisive for LN diagnosis. A deep learning architecture will be used to combine information from across all modalities, optimize feature selection and quantification. We present extensive preliminary data from kidneys of wildtype and LN murine models demonstrating the efficacy of our techniques. We will validate the efficacy of LN diagnostic metrics from murine models using archival human kidney biopsy samples. We also present data from human subjects with Class II LN (non-proliferative), Class IV LN (proliferative) and minimal change disease (control) and demonstrate statistically significant metrics derived from our imaging modalities that enable improved LN diagnosis.