Development of a Machine Learning Tool for Non-invasive Diagnosis of Liver Graft Pathology Using TruGraf

Given the significant investment of healthcare resources into transplantation, it is critical to identify recipients with graft pathologies such as Acute Cellular Rejection (ACR), NASH, cholestasis, etc. at an earlier stage to implement the appropriate intervention, rather than initiating empiric treatment that could be unsafe. This project will develop a practical Machine learning-based tool based on the results of the TruGraf assay alongside clinical and laboratory data for non-invasive diagnosis of graft pathology. TruGraf is a non-invasive test that measures differentially expressed genes in the blood of transplant recipients to identify patients who are likely to be adequately immunosuppressed and, in doing so, rule out graft damage. TruGraf measures the difference in gene expression for a precise panel of specific genes that have been empirically determined to discriminate between allografts that are truly healthy (Non-ACR), and those in transplant patients that have acute rejection on biopsy (AR). Nevertheless, the exact etiology of graft damage may be difficult to discern for the transplant clinician. The clinical characteristics and history of the liver transplant recipient as well as liver enzyme patterns can provide a pre-test probability of one diagnosis being more likely than the other (Acute cellular rejection, NASH, biliary or viral disease). The proposed tool will leverage our expertise in Machine Learning tools applied to clinical and molecular data (TruGraf assay results) to enable effective clinical implementation of the TruGraf assay.