High-grade serous ovarian carcinoma (HGSOC) is the most common aggressive and lethal histologic subtype of ovarian cancer. The role of minimally invasive surgery in the management of HGSOC has expanded such that laparoscopic assessment of disease burden is now a widely accepted approach. Previously, we reported comprehensive multi-platform omics analyses, including an integrated analysis and immune assessment of primary and metastatic tumors from highly clinically annotated HGSOC samples. In this proposal, we will apply artificial intelligence (AI) in this context of morphologic and mutli-platform omics differences observed in HGSOC. We believe that deep learning (DL) aided software can provide accurate prediction of patient outcomes at the time of laparoscopic assessment for advanced HGSOC by incorporating still frame surgical images, multi-platform omics data, along with clinical variables such as routine laboratory tests. We hypothesize that multimodal AI can learn features from morphologic, clinical, and omics data obtained before and during laparoscopic assessment of HGSOC and that, these features could then be applied to the development of an assay to predict patient outcomes. To achieve these aims, advanced DL techniques will be employed to develop multimodal DL models utilizing surgical images, clinical laboratory data, and multi-omics data. We expect the appropriate features from morphological, laboratory and multi-platform omics data can be analyzed in combination as complementary modalities that could lead to development of an assay that could aid in the prediction of the outcomes for advanced HGSOC at the time of laparoscopic assessment of disease. After training the models for the two Specific Aims, we will explore the learned features to identify any correlation between the morphologic, clinical, and multi-platform omics patterns that may lead to the discovery of new hybrid biomarkers with prognostic and/or predictive values for patient outcomes. The term hybrid biomarkers refer to biomarkers that can help identify patient outcomes across multiple modalities that cannot be identified when using single modality. To the best of our knowledge, we are the first team to propose investigating such hybrid biomarkers to interrogate the clinical outcomes at the time of laparoscopic assessment for HGSOC. This is only possible due to availability of comprehensive clinical data from the Ovarian Cancer Moon Shot program at MD Anderson Cancer Center on large and diverse population of ovarian cancer patients.
Shayan Shams, Ph.D. is an Assistant Professor at the Department of Applied Data Science at San Jose State University. He is also an adjunct Assistant Professor at the School of Biomedical Informatics (SBMI), and the Institute for Stroke and Cerebrovascular Diseases at the University of Texas Health Science Center (UTHealth Houston).
As a computer scientist and enthusiastic data scientist, Dr. Shams has extensive experience and a strong track record in developing artificial intelligence algorithms for challenging problems in medicine. Dr. Shams has developed novel artificial intelligence based methodologies to identify patterns and biomarkers from biomedical data such as magnetic resonance imaging, X-rays, computer tomography, PPG and EEG signal, and Electronic Health Records (EHR). Dr. Shams has authored more than 34 peer-reviewed articles in prestigious journals including Nature, Journal of Biomedical Informatics, Journal of clinical periodontology. Dr. Shams has received competitive grants from the National Institute of Health, Cancer Prevention Research Institute of Texas (CPRIT) and private foundations.