Artificial intelligence in drug discovery: A comprehensive review with a case study on hyperuricemia, gout arthritis, and hyperuricemic nephropathy

Junwei Su, Cheng Xin, Ao Shang, Shan Wu, Zhenzhen Xie, Ruogu Xiong, Xiaoyu Xu, Cheng Zhang, Guang Chen, Yau-Tuen Chan, Guoyi Tang, Ning Wang, Yong Xu, Yibin FengarXiv 2025 · DOI · arXiv

Abstract

Artificial intelligence and machine learning (AI/ML) are transforming the drug discovery landscape by integrating traditionally isolated stages into a cohesive, data-driven pipeline. This review examines the end-to-end application of AI—from target identification to clinical trial optimization—complemented by a case study on hyperuricemia and gout. By bridging theoretical methodologies with real-world applications, the authors demonstrate how hybrid AI models accelerate the transition from "bench to bedside."

Introduction: Addressing the "$2 Billion Gamble"

The traditional drug discovery process is notoriously inefficient, often requiring 10–15 years and over $2 billion to bring a single therapeutic to market. With clinical attrition rates exceeding 90%, the primary challenge lies in the fragmented nature of the development pipeline. A "hit" molecule identified in early screening often fails later due to unforeseen toxicity or poor pharmacokinetics. This paper advocates for a holistic AI-integrated approach to mitigate these late-stage failures by maintaining data continuity across all development phases.

Methodology: The AI-Driven Pipeline

The authors categorize AI/ML interventions into three primary phases, emphasizing the shift from manual heuristics to automated, high-dimensional analysis.

Comprehensive AI/ML pipeline for drug discovery, spanning target identification to clinical trial optimization.
  1. Target Identification: Moving beyond manual literature reviews, researchers now employ Natural Language Processing (NLP) models like BioBERT and Network Medicine approaches. Graph Neural Networks (GNNs) are particularly effective here, as they model biological systems as complex, interconnected networks of proteins and genes.
  2. Hit Screening & Design: The "generative revolution" has introduced Variational Autoencoders (VAEs), Generative Adversarial Networks (GANs), and Diffusion Models. These allow for de novo design—creating entirely new molecules tailored to specific protein pockets—rather than merely searching existing chemical libraries.
  3. Lead Optimization: This phase focuses on predicting Absorption, Distribution, Metabolism, Excretion (ADME), and toxicity. Hybrid models combine classical physics-based pharmacokinetic modeling with machine learning algorithms like XGBoost to enhance both accuracy and interpretability.

Key Insights: Data Representation and Hybridization

A fundamental aspect of AI's success in this field is how biological and chemical information is represented. Biological data is inherently relational, making graph-based structures superior for modeling how diseases or drugs perturb cellular networks.

Common data representations in AI-driven drug discovery: SMILES strings, amino acid sequences, and graph structures.

Furthermore, the "Digital Twin" concept is emerging in clinical trials. By creating virtual patient models, researchers can simulate drug responses and adverse effects before human administration. The paper emphasizes that the most robust models are "hybrid," combining data-driven ML with mechanistic biological principles to ensure predictions remain physiologically grounded.

Results and Case Study: Gout and Hyperuricemia

The practical utility of these methods is demonstrated through several milestones:

  • Accelerated Discovery: Insilico Medicine utilized AI to identify a novel fibrosis target and design a clinical-stage candidate in a fraction of the usual time.
  • Drug Repurposing: AI-generated knowledge graphs identified Baricitinib as a potential COVID-19 treatment, later validated in trials.
  • Gout Case Study: AI-assisted QSAR (Quantitative Structure-Activity Relationship) modeling identified 15 novel molecules as superior Xanthine Oxidase (XOD) inhibitors. These candidates offer the potential for more effective gout treatments with fewer side effects than current standards like Allopurinol.

Limitations and Future Directions

Despite significant progress, three major bottlenecks remain:

  1. Data Quality: Models are limited by small or biased datasets, particularly for less-studied protein families.
  2. Explainability: Regulatory approval requires "Explainable AI" (XAI) to understand the logic behind model predictions.
  3. Biological Complexity: The gap between in silico prediction and in vivo reality remains significant.

Future advancements are expected to stem from Quantum Machine Learning and Federated Learning, which allows for collaborative model training without compromising sensitive primary data.

Summary updated

Abstract

This paper systematically reviews recent advances in artificial intelligence (AI), with a particular focus on machine learning (ML), across the entire drug discovery pipeline. Due to the inherent complexity, escalating costs, prolonged timelines, and high failure rates of traditional drug discovery methods, there is a critical need to comprehensively understand how AI/ML can be effectively integrated throughout the full process. Currently available literature reviews often narrowly focus on specific phases or methodologies, neglecting the dependence between key stages such as target identification, hit screening, and lead optimization. To bridge this gap, our review provides a detailed and holistic analysis of AI/ML applications across these core phases, highlighting significant methodological advances and their impacts at each stage. We further illustrate the practical impact of these techniques through an in-depth case study focused on hyperuricemia, gout arthritis, and hyperuricemic nephropathy, highlighting real-world successes in molecular target identification and therapeutic candidate discovery. Additionally, we discuss significant challenges facing AI/ML in drug discovery and outline promising future research directions. Ultimately, this review serves as an essential orientation for researchers aiming to leverage AI/ML to overcome existing bottlenecks and accelerate drug discovery.
BibTeX
@misc{su2025artificial,
  author = {Junwei Su and Cheng Xin and Ao Shang and Shan Wu and Zhenzhen Xie and Ruogu Xiong and Xiaoyu Xu and Cheng Zhang and Guang Chen and Yau-Tuen Chan and Guoyi Tang and Ning Wang and Yong Xu and Yibin Feng},
  title = {{Artificial intelligence in drug discovery: A comprehensive review with a case study on hyperuricemia, gout arthritis, and hyperuricemic nephropathy}},
  year = {2025},
  eprint = {2507.03407},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2507.03407},
}
Figure