Application of machine learning techniques to the analysis and prediction of drug pharmacokinetics.

In this review, we describe the current status and challenges in applying machine-learning techniques to the analysis and prediction of pharmacokinetic data. The theory of pharmacokinetics has been developed over decades on the basis of physiology and reaction kinetics. Mathematical models allow the reduction of pharmacokinetic data to parameter values, giving insight and understanding into ADME processes and predicting the outcome of different dosing scenarios. However, much information hidden in the data is lost through conceptual simplification with models. It is difficult to use mechanistic models alone to predict diverse pharmacokinetic time profiles, including inter-drug and inter-individual differences, in a cross-sectional manner. Machine learning is a prediction platform that can handle complex phenomena through data-driven analysis. As a resule, machine learning has been successfully adopted in various fields, including image recognition and language processing, and has been used for over two decades in pharmacokinetic research, primarily in the area of quantitative structure-activity relationships for pharmacokinetic parameters. Machine-learning models are generally known to provide better predictive performance than conventional linear models. Owing to the recent success in deep learning, models with new structures are being consistently proposed. These models include transfer learning and generative adversarial networks, which contribute to the effective use of a limited amount of data by diverting existing similar models or generating pseudo-data. How to make such newly emerging machine learning technologies applicable to meet challenges in the pharmacokinetics/pharmacodynamics field is now the key issue.

Cytosolic Ca2+-dependent Ca2+ release activity primarily determines the ER Ca2+ level in cells expressing the CPVT-linked mutant RYR2.

Type 2 ryanodine receptor (RYR2) is a cardiac Ca2+ release channel in the ER. Mutations in RYR2 are linked to catecholaminergic polymorphic ventricular tachycardia (CPVT). CPVT is associated with enhanced spontaneous Ca2+ release, which tends to occur when [Ca2+]ER reaches a threshold. Mutations lower the threshold [Ca2+]ER by increasing luminal Ca2+ sensitivity or enhancing cytosolic [Ca2+] ([Ca2+]cyt)-dependent activity. Here, to establish the mechanism relating the change in [Ca2+]cyt-dependent activity of RYR2 and the threshold [Ca2+]ER, we carried out cell-based experiments and in silico simulations. We expressed WT and CPVT-linked mutant RYR2s in HEK293 cells and measured [Ca2+]cyt and [Ca2+]ER using fluorescent Ca2+ indicators. CPVT RYR2 cells showed higher oscillation frequency and lower threshold [Ca2+]ER than WT cells. The [Ca2+]cyt-dependent activity at resting [Ca2+]cyt, Arest, was greater in CPVT mutants than in WT, and we found an inverse correlation between threshold [Ca2+]ER and Arest. In addition, lowering RYR2 expression increased the threshold [Ca2+]ER and a product of Arest, and the relative expression level for each mutant correlated with threshold [Ca2+]ER, suggesting that the threshold [Ca2+]ER depends on the net Ca2+ release rate via RYR2. Modeling reproduced Ca2+ oscillations with [Ca2+]cyt and [Ca2+]ER changes in WT and CPVT cells. Interestingly, the [Ca2+]cyt-dependent activity of specific mutations correlated with the age of disease onset in patients carrying them. Our data suggest that the reduction in threshold [Ca2+]ER for spontaneous Ca2+ release by CPVT mutation is explained by enhanced [Ca2+]cyt-dependent activity without requiring modulation of the [Ca2+]ER sensitivity of RYR2.

Prediction of HIV drug resistance based on the 3D protein structure: Proposal of molecular field mapping.

A method for predicting HIV drug resistance by using genotypes would greatly assist in selecting appropriate combinations of antiviral drugs. Models reported previously have had two major problems: lack of information on the 3D protein structure and processing of incomplete sequencing data in the modeling procedure. We propose obtaining the 3D structural information of viral proteins by using homology modeling and molecular field mapping, instead of just their primary amino acid sequences. The molecular field potential parameters reflect the physicochemical characteristics associated with the 3D structure of the proteins. We also introduce the Bayesian conditional mutual information theory to estimate the probabilities of occurrence of all possible protein candidates from an incomplete sequencing sample. This approach allows for the effective use of uncertain information for the modeling process. We applied these data analysis techniques to the HIV-1 protease inhibitor dataset and developed drug resistance prediction models with reasonable performance.

A model-based comparative meta-analysis of the efficacy of dolutegravir-based and efavirenz-based regimens in HIV-infected patients.

Currently, combinations of typical types of antiretroviral agents have been adopted as chemotherapy for human immunodeficiency virus (HIV) infection, comprising two nucleoside analogue reverse transcriptase inhibitors plus one of a non-nucleoside reverse transcriptase inhibitor, an integrase strand-transfer inhibitor, and a protease inhibitor. Although several meta-analyses have been conducted to determine first-line combination antiretroviral therapy, this has yet to be confirmed due to the technical limitation associated. In the present study, we applied a model-based meta-analysis (MBMA) approach, because it allows integration of information from clinical trials with varying dosing, duration, and sampling time points, resulting in enlargement of available data sources. We performed a bibliographic search to identify clinical trials involving dolutegravir (DTG)-based and efavirenz (EFV)-based regimens in HIV-infected, antiretroviral therapy-naïve adults, and then identified 30 independent trial data. The time course of drug effect was described by a consecutive first-order kinetic model and analyzed using the nonlinear mixed effect modeling approach. The developed model suggests that the DTG-based regimen provides a faster-acting and more sustainable drug effect than the EFV-based regimen. Moreover, the drug effect tends to appear more slowly and decay faster in severe patients having higher viral load or smaller baseline CD4 count.