Artificial Intelligence-Personalized Medicine (AI-PM)

The current drug development roadmaps are prohibitively long, about 10-15 years, and cost billions of dollars. The success rate of development from in vitro assays to regulatory approval is much less than 5 percent. In the clinical applications, the average response rate of chemotherapy is one out from four patients.

The two fundamental challenges that impede drug development and clinical applications are: 1) Conventional approaches cannot pinpoint the best in vitro drug composition and dose ratios that simultaneously mediate the highest efficacy and lowest toxicity from a large search space. 2) Once preliminary doses are determined using in vitro tests, the translation of these doses to in vivo applications is performed by matching the PK or by weight/surface area scaling or by determining the maximum tolerated dose (MTD), which implicitly preclude optimization. These are the two primary roadblocks.

We took a mechanism independent approach to bypass the dynamically varying and insurmountable maze created by the drug-dose space. Specifically, our team used the artificial intelligence (AI) based feedback control and search algorithm approaches to identify the most potent drug-dose combination that inhibits herpesvirus infection. It was surprising that the feedback system control (FSC) only needed 10-20 search iterations to pinpoint the optimal combination from thousands or millions of possible drug-dose combinations. Hence, FSC has enabled us to optimize the drug-dose combination with a mechanism independent approach. However, FSC is a serial search method. With an AI based neural networks approach, a more efficient parallel search platform, Phenotypic Response Surface (PRS), has been developed. Only a few tests can guide toward the optimized drug-dose combination for a specific patient.

We have discovered that the drug-dose inputs are correlated with the phenotypic outputs with a Phenotypic Response Surface (PRS). With a few calibration tests to determine the coefficients of the quadratic algebraic equation governing PRS, PRS dictates the composition and the ratio of a globally optimized drug combination from a very large search space in in vitro tests. In vitro drug compositions and ratios will not translate into successful preclinical or clinical validation. Since only a very small number of tests are needed, we can re-optimize drug-dose ratios in animal and clinical studies. Re-optimization has already been successfully demonstrated to converge upon key drugs/combinations of interest. Based on the PRS platform, Phenotypic Personalized Medicine (PPM) can realize unprecedented levels of adaptability to identify the optimized drug combination for a specific patient, even if changes to the regimen and dose/drug optimization are needed in a continuous basis. PRS is an indication agnostic and mechanism free platform technology, which has been successfully demonstrated in about 30 diseases for children and adults with 0 misses.

Control of Turbulent Shear Flow Microfluidics