Search

We introduce the concept of optimal drug design (ODD) as the use of an AI framework to optimize the exposure, safety, and efficacy of drugs. To exemplify the concept of ODD, we developed an artificial intelligence framework that integrates de novo molecular design, quantitative structure activity relationships, and pharmacokinetic-pharmacodynamic modeling. Specifically, our computational architecture has integrated a generative algorithm for small molecule design with a hybrid physiologically-based pharmacokinetic machine learning (PBPK-ML) model, which was applied to generate and optimize drug candidates for enhanced brain exposure. Publicly sourced data on the plasma and brain pharmacokinetics of 77 small molecule drugs in rats was used for model development. We have observed an approximate 30-fold and 120-fold increase on average in predicted brain exposure for AI generated molecules compared to known central nervous system drugs and randomly selected small organic molecules. We believe that with additional data and mechanistic modeling this in silico pipeline could facilitate the discovery of a new wave of optimally designed medicines for the treatment of CNS diseases.Graphical AbstractArtificial Intelligence Framework for the Optimization of Brain Pharmacokinetics.A genetic algorithm consisting of cross-breeding, mutating, scoring, and refining was used for de novo generation of a population of new molecular structures. SELFIE representations of molecules were used as input to a variational autoencoder for de novo generation/refinement of individual drug candidates. Molecular descriptors of individual drug candidates are generated and used as input into a trained neural network to generate drug-specific pharmacokinetic (PK) parameters. PK parameters are used as input into a physiologically-based pharmacokinetic (PBPK) model of the brain to predict brain PK of the drug candidate. Brain concentration-time profiles are integrated to obtain an area-under the curve (AUC), a metric of brain exposure, which is used to score and inform the design of new generations of molecules. Iterations of this framework generate novel drug candidates optimized for greater brain exposure. Created with BioRender.

There are several antibody therapeutics in preclinical and clinical development, industry-wide, for the treatment of central nervous system (CNS) disorders. Due to the limited permeability of antibodies across brain barriers, the quantitative understanding of antibody exposure in the CNS is important for the design of antibody drug characteristics and determining appropriate dosing regimens. We have developed a minimal physiologically-based pharmacokinetic (mPBPK) model of the brain for antibody therapeutics, which was reduced from an existing multi-species platform brain PBPK model. All non-brain compartments were combined into a single tissue compartment and cerebral spinal fluid (CSF) compartments were combined into a single CSF compartment. The mPBPK model contains 16 differential equations, compared to 100 in the original PBPK model, and improved simulation speed approximately 11-fold. Area under the curve ratios for minimal versus full PBPK models were close to 1 across species for both brain and plasma compartments, which indicates the reduced model simulations are similar to those of the original model. The minimal model retained detailed physiological processes of the brain while not significantly affecting model predictability, which supports the law of parsimony in the context of balancing model complexity with added predictive power. The minimal model has a variety of applications for supporting the preclinical development of antibody therapeutics and can be expanded to include target information for evaluating target engagement to inform clinical dose selection. Supplementary Information The online version contains supplementary material available at 10.1007/s10928-021-09776-7.

The development and application of quantitative systems pharmacology models in neuroscience have been modest relative to other fields, such as oncology and immunology, which may reflect the complexity of the brain. Technological and methodological advancements have enhanced the quantitative understanding of brain physiology and pathophysiology and the effects of pharmacological interventions. To maximize the knowledge gained from these novel data types, pharmacometrics modelers may need to expand their toolbox to include additional mathematical and statistical frameworks. A session was held at the 10th annual American Conference on Pharmacometrics (ACoP10) to highlight several recent advancements in quantitative and systems neuroscience. In this mini‐review, we provide a brief overview of technological and methodological advancements in the neuroscience therapeutic area that were discussed during the session and how these can be leveraged with quantitative systems pharmacology modeling to enhance our understanding of neurological diseases. Microphysiological systems using human induced pluripotent stem cells (IPSCs), digital biomarkers, and large‐scale imaging offer more clinically relevant experimental datasets, enhanced granularity, and a plethora of data to potentially improve the preclinical‐to‐clinical translation of therapeutics. Network neuroscience methodologies combined with quantitative systems models of neurodegenerative disease could help bridge the gap between cellular and molecular alterations and clinical end points through the integration of information on neural connectomics. Additional topics, such as the neuroimmune system, microbiome, single‐cell transcriptomic technologies, and digital device biomarkers, are discussed in brief.

Artificial intelligence, in particular machine learning (ML), has emerged as a key promising pillar to overcome the high failure rate in drug development. Here, we present a primer on the ML algorithms most commonly used in drug discovery and development. We also list possible data sources, describe good practices for ML model development and validation, and share a reproducible example. A companion article will summarize applications of ML in drug discovery, drug development, and postapproval phase., Laboratorio de Investigación y Desarrollo de Bioactivos

Quantitative systems pharmacology (QSP) is an emerging discipline that aims to discover how drugs modulate the dynamics of biological components in molecular and cellular networks and the impact of those perturbations on human pathophysiology. The integration of systems-based experimental and computational approaches is required to facilitate the advancement of this field. QSP models typically consist of a series of ordinary differential equations (ODE). However, this mathematical framework requires extensive knowledge of parameters pertaining to biological processes, which is often unavailable. An alternative framework that does not require knowledge of system-specific parameters, such as Boolean network modeling, could serve as an initial foundation prior to the development of an ODE-based model. Boolean network models have been shown to efficiently describe, in a qualitative manner, the complex behavior of signal transduction and gene/protein regulatory processes. In addition to providing a starting point prior to quantitative modeling, Boolean network models can also be utilized to discover novel therapeutic targets and combinatorial treatment strategies. Identifying drug targets using a network-based approach could supplement current drug discovery methodologies and help to fill the innovation gap across the pharmaceutical industry. In this review, we discuss the process of developing Boolean network models and the various analyses that can be performed to identify novel drug targets and combinatorial approaches. An example for each of these analyses is provided using a previously developed Boolean network of signaling pathways in multiple myeloma. Selected examples of Boolean network models of human (patho-)physiological systems are also reviewed in brief.

The overarching goal of modern drug development is to optimize therapeutic benefits while minimizing adverse effects. However, inadequate efficacy and safety concerns remain to be the major causes of drug attrition in clinical development. For the past 80 years, toxicity testing has consisted of evaluating the adverse effects of drugs in animals to predict human health risks. The U.S. Environmental Protection Agency recognized the need to develop innovative toxicity testing strategies and asked the National Research Council to develop a long-range vision and strategy for toxicity testing in the 21st century. The vision aims to reduce the use of animals and drug development costs through the integration of computational modeling and in vitro experimental methods that evaluates the perturbation of toxicity-related pathways. Towards this vision, collaborative quantitative systems pharmacology and toxicology modeling endeavors (QSP/QST) have been initiated amongst numerous organizations worldwide. In this article, we discuss how quantitative structure-activity relationship (QSAR), network-based, and pharmacokinetic/pharmacodynamic modeling approaches can be integrated into the framework of QST models. Additionally, we review the application of QST models to predict cardiotoxicity and hepatotoxicity of drugs throughout their development. Cell and organ specific QST models are likely to become an essential component of modern toxicity testing, and provides a solid foundation towards determining individualized therapeutic windows to improve patient safety.