Predicting Chemotherapy Response in Sarcoma using Molecular and Imaging Data Collected in a Digital Registry
One of the major challenges associated with clinical and translational research in patients with bone and soft tissue tumors has been the relative rarity of bone and soft-tissue tumors and a dearth of well-annotated clinical registries for these patients. We have developed a unique registry at Ohio State University which combines (i) a well-annotated, manually-curated dataset (REDCap) of key data elements known to be important in sarcoma outcomes with (ii) unstructured “big data” (labs, ICD-9/10 codes, medication history) derived from the patients’ electronic health record (EHR). In addition, we have the ability to incorporate next-generation sequencing, digital pathology images, and raw radiology images for selected patients enrolled on the registry. We hypothesize that data collected within this registry can be used to develop a predictive algorithm for response to systemic chemotherapy. We propose developing a predictive algorithm using standard clinical data alone (stage, histology, patient demographics) using a machine-learning computational algorithm designed by the PI. We will subsequently assess whether the addition of molecular data (exome sequencing) and histopathology imaging data (digital slide analysis) can improve predictive models.
To assess the potential impact of this model on patterns of clinical care, we will generate chemotherapy outcomes predictions for 3 common chemotherapy drugs used in the treatment of soft-tissue sarcoma. In a blinded fashion, we will determine if the predictive information provided by the algorithm is sufficient to alter treatment recommendations made by Sarcoma Medical Oncologists in a retrospective cohort of patients.
We believe that potential translational impact of this project is huge and provides a novel avenue for clinical translational research in Bone and Soft-Tissue Tumors.