Using machine learning and multiomics to predict the pharmacokinetics of immunosuppressants and clinical outcomes in renal transplant patients

We conducted an LC-MS/MS-based untargeted plasma metabolomics study and determined the whole genome genotypes in patients with kidney transplants. We utilized various machine learning algorithms with the multiomics data to predict the pharmacokinetics of the immunosuppressants tacrolimus and mycophenolate mofetil and clinical outcomes, such as biopsy-proven rejection. In addition, we compared the performance of the machine learning approaches with the conventional population pharmacokinetics/pharmacodynamics method. The study demonstrated that combining machine learning and multiomics could be a powerful tool for optimizing immunosuppressant pharmacotherapy in renal transplant patients.