AI-Powered Quantitative Systems Pharmacology for AD Drug Repurposing

点击翻译按钮获取中文摘要

Abstract Alzheimer's disease (AD) poses a triple threat to public health, as its prevalence is on the rise, its costs are immense, and there is no effective therapy. However, drug development attempts for the treatment of AD have met with minimal success. The failure is largely attributable to a reductionist concept of "one drug, one gene, one disease." As AD is a multigenic heterogeneous illness, a new therapeutic strategy is urgently required to concurrently target the numerous pathogenic processes involved for the genesis and progression of AD in each individual patient. Many translational bioinformatics strategies for AD drug repurposing have been developed in recent years. Existing target-based, phenotype-based, network-based, and patient-based drug repurposing strategies are unable to fully address the challenges of AD drug repurposing due to the lack of thoroughly validated drug targets, potent lead compounds, and high-throughput phenotype readouts that can characterize the molecular complexity of AD. Over the past decade, we have built an artificial intelligence- based quantitative systems pharmacology (AI-QSP) platform that attempts to predict and characterize genome-wide chemical-protein interactions and functional activities, as well as correlate molecular interactions with phenotypic responses. Our AI-QSP platform integrates diverse omics data synergistically and incorporates machine learning, biophysics, and systems biology methodologies. The AI-QSP platform has been effectively applied to drug repurposing including AD, polypharmacology, side effect prediction, and precision medicine. Established our proof-of-concept studies, we propose to develop and thoroughly evaluate a unique computational methodology that combines target-based and mechanism-driven phenotypic chemical screening for AD individualized drug repurposing. Using a novel domain adaptation strategy, we will expand our context- independent phenotypic compound screening methodologies to AD patient-specific, cell type-specific, transcriptome-based drug repurposing. In addition, we will analyze the ADME features of repurposed pharmaceuticals in the human brain utilizing cutting-edge physiologically based pharmacokinetics (PBPK) techniques. We will improve state-of-the-art drug-gene-disease network models for Alzheimer's disease drug repurposing by incorporating understudied dark proteins that are abundant in the target list suggested by AD omics studies and their inhibitory or activatory effects, and by applying graph mining techniques for drug-gene- disease link predictions. Using cell-based disease models and RNA-seq studies, we will combine complementary phenotype-based and target-based techniques to rank drug candidates and confirm their efficacy and toxicity on AD treatment. In conclusion, the successful completion of this project could provide the scientific community with a novel translational bioinformatics resource for identifying potential therapeutics for effective personalized AD treatments and advancing drug repurposing to a new phase of lead optimizations and clinical trials.