Ductal carcinoma in situ (DCIS) is a benign breast disease with the potential to develop into invasive breast cancer (IBC). DCIS is detected in around 25% of breast risk screens, yet it is not possible to determine from pathological evaluation which women will progress to IBC. This results in overtreatment by whole breast radiation, partial mastectomy, partial mastectomy or mastectomy to eliminate risk of progression. Contemporary literature reports that qualitatively, stroma collagen plays has a role in tumor sensing to limit progression and that collagen fiber width, length and curvature are predictive of progression. The molecular composition of stroma collagen represents a knowledge gap that could potentially predict progression and limit overtreatment, as well as lead to new molecular insights on stroma signaling related to breast cancer risk. Our preliminary data shows that peptides representing cell interactive collagen domains are regulated by DCIS lesion pathology that represents risk and are predictive of progression. From this data, our overall hypothesis is that stromal proteomic signatures derived from post-translational modification of collagen surrounding or within the microenvironment differentiate DCIS lesion architectures representing risk and are predictive of progression. This exploratory translational project will develop the concept that stroma proteomic signatures are predictive of outcome in breast cancer risk. Aim 1 will work to differentiate intrapatient and interpatient heterogeneity of DCIS lesion pathology from IBC in 90 lumpectomies (29% black women/71% white women) associated with DCIS by spatial stromal and cellular proteomics coupled to multiplexed immunohistochemistry. Biosignatures of DCIS risk (large lesion, high nuclear grade and comedo necrosis) will be evaluated compared to lower risk DCIS pathologies and IBC. In Aim 2, we will evaluate clinically characterized and cell defined tissue microarrays from the Resource for Archival Breast Tissue (RAHBT) consisting of tissue cores from primary DCIS diagnosis with median prospective follow-up of 11.4 years representing 35% black women/68% white women. Biosignatures from spatial stromal proteomics associated with nuclear grade and architectural g patterns of solid, cribriform, and comedo necrosis will be evaluated relative to outcome. Machine learning algorithms will be used to assess significance of collagen stromal regulation between pathologies, subtypes, and clinical characteristics including ancestry with an overall goal of using stroma signatures from the DCIS lesion to predict progression. A long-term goal is to change the clinical paradigm of patient management by overtreatment to using molecular markers of patient risk at the time of clinical diagnosis that guide treatment of high-risk patients and limit overtreatment for those with a low risk of progression.