TITLE: Combined Imaging and RNA Analyses to Develop Cervical Cancer Biomarkers ABSTRACT Despite significant advances in disease prevention and screening, cervical cancer continues to be an important worldwide public health problem. Treating cervical cancer patients with personalized strategies can potentially improve the chance of survival. Predicting early in treatment whether a tumor is likely to be responsive is one of the most challenging yet important tasks for stratifying cervical cancer patients and supporting personalized treatment strategies to improve cancer patient care. Various unimodal data, including ribonucleic acids (RNAs), radiologic and histologic imaging, and clinicopathologic data, have been employed for predicting cervical cancer treatment response and patient outcome. Each type of unimodal data analyzes tumor phenotypes from a different point of view and provides valuable while limited prognostic information. We and others have shown that RNAs are promising biomarkers and play critical regulatory roles in cervical cancer. Radiologic imaging biomarkers have shown promise in stratifying patients with favorable and unfavorable prognosis for multiple tumor sites. Their non-invasive characteristics also allow for convenient and longitudinal monitoring of tumor progression and heterogeneous response during the treatment course. Moreover, histologic images provide key information about microscopic structure of cells and tissues of organisms. Recent reports and our preliminary studies have shown that histologic imaging biomarkers, can aid in clinical decision-making by identifying metastases, subtyping and grading tumors, and predicting clinical outcomes. Clinicopathologic biomarkers show prognostic value through retrospective studies. Still, many cervical cancer patients have tumor recurrence despite favorable prognosis by these biomarkers individually. The major goal of this study is to develop a comprehensive and robust computational model for prediction of cervical cancer treatment response and outcomes. We will integrate our recently developed advanced learning- based techniques to build prognostic models using about 600 cervical patient cases collected from two institutions. The prognostic model will form a solid basis for individualized care of cervical cancer patients. Moreover, our work is expected to discover the correlations among multimodal data, leading to dynamic patient stratification to support adaptive treatment strategies in the future.