The integration of ML and DL has fundamentally transformed the field of drug discovery

The gap

Evidence profile

Sourced from the future work of the source papers, classified as general, spanning 3 journals. Those papers have been cited 2 times in total.

Supporting evidence — 3 representative gaps

• A biology-based quality-diversity algorithm for drug repurposing in Alzheimer’s disease using automated machine learning (2026) · BioData Mining · doi

  This study addresses a critical challenge in computational drug repurposing for Alzheimer’s Disease: the need for methods that can systematically discover mecha- nistically novel and interpretable therapeutic hypotheses. We presented a framework that achieves this by uniquely integrating biologically-informed GNN embeddings with a MAP-Elites quality-diversity search within the TPOT2 AutoML pipeline. Our primary contribution is the use of a biomedical knowledge graph (AlzKB) to guide the evolutionary search, defining novelty based on an embedding’s dis- tance to known AD-related entities. The identification of diverse candidates such as Exemestane and Felodipine from underrepresented regions of the feature space provides supporting evidence for the feasibility of this novelty-guided approach as a hypothesis-generation framework. By fusing graph-based biological priors with a diversity-aware evolutionary search, our work presents a methodological proof-of- concept for integrating biological knowledge into AutoML-driven drug repurposing. Future work will focus on enriching the biological data foundation by incorporating multi-omics datasets and exploring more advanced GNN architectures. We will also investigate more dynamic AutoML search paradigms using reinforcement learning and the integration of Large Language Models (LLMs) to automatically update the knowledge base. Ultimately, the most critical next step is the experimental valida- tion of our predicted drug candidates through in vitro and in vivo studies, and the application of this framework to other complex neurodegenerative diseases, such as Parkinson’s Disease, Amyotrophic Lateral Sclerosis (ALS), and Huntington’s Disease. Shao et al. BioData Mining (2026) 19:34 Page 16 of 17

  generalfuture work
  Keywords: search drug disease framework automl knowledge biological critical repurposing integrating diversity graph evolutionary novelty based
• Artificial intelligence in drug discovery from advanced molecular representation to pipeline applications (2026) · Frontiers in Bioinformatics · cited 2× · doi

  The integration of ML and DL has fundamentally transformed the field of drug discovery. By offering powerful tools for molecular representation, predictive modeling, and rational design, sophisticated DL techniques are now routinely achieving competitive or superior performance across all critical stages of the pharmaceutical pipeline. Successful integrations, such as the use of TransformerCPI to identify natural inhibitors (PP10 and PP24) for the pan-cancer marker CD133, underscore the power of AI in mechanistic elucidation and experimental validation. This AI-driven computational strategy is essential for mitigating the historical challenges of the R&D process, and enables the crucial early prediction and optimization of ADME/Tox profiles, which directly addresses the high failure substantial costs associated with drug development projects. rate and The continued advancement of AI in drug discovery relies on sustained innovation in methodology and integration. Future research efforts are expected to concentrate distinctly on three main areas: (1) Geometric Fidelity: Increasing the accuracy and speed of 3D molecular prediction and generation. This is vital for capturing the precise spatial and physical mechanisms of drug action; (2) Data Efficiency: Scaling up Few-Shot Learning and Meta-Learning approaches to tackle the inherent low-data problem (data sparsity) across many ADME/Tox biological targets; (3) Integration and XAI: Developing fully integrated in silico workflows that seamlessly link target validation, de novo design, and toxicity/PK prediction. This must incorporate advanced XAI techniques to ensure transparent,

  generalfuture work
  Keywords: drug integration prediction discovery molecular design techniques across validation adme learning fundamentally transformed field offering
• Artificial Intelligence in Pharmaceutical Innovation and Research: A Review (2026) · Zenodo (CERN European Organization for Nuclear Research) · doi

  and 13.1. Integration of AI with multi-omics and systems pharmacology In the near future, AI will increasingly be (genomics, integrated with multi-omics proteomics, metabolomics, transcriptomics, epigenomics) systems-pharmacology frameworks to build holistic, dynamic models of drug action and disease progression. By combining AI-powered pattern recognition with mechanistic pathway models, researchers can simulate how drugs perturb biological networks across multiple scales (from genes to organs), target selection and enabling more rational combination-therapy design. Such integrated platforms will help bridge the gap between in- outcomes, and silico improving the translation of preclinical findings into clinically meaningful effects. These systems-pharmacology-plus-AI workflows are especially promising for complex polygenic diseases (e.g., cancer, neurodegenerative and autoimmune disorders), where multiple targets and feedback loops operate simultaneously. As multi- omics datasets grow and standardized ontologies is improve, AI-driven systems-pharmacology predictions in-vivo expected to become a core component of modern supporting drug-development hypothesis-generation, mechanism-of-action elucidation, and biomarker-driven trial design.[25] pipelines, and 13.2. AI-driven personalized and precision medicine AI will play a central role in advancing personalized and precision medicine, tailoring drug therapy to individual patients based on their genetic background, comorbidities, environmental real-time physiological data. factors, Machine-learning models trained on large-scale genomic and electronic-health-record data can predict drug response, adverse-reaction risk, and optimal specific subpopulations, moving beyond “one-size-fits-all” approaches. In oncology, psychiatry, and rare diseases, such personalized models can guide the selection of therapies most likely to benefit a given patient while minimizing toxicity. Wearable-device-derived continuous- data, monitoring platforms, and AI-enabled decision support systems will allow adaptive treatment regimens that evolve as new patient-specific information becomes available. By embedding AI- based personalization tools into clinical-pharmacy regimens dosing for INTERNATIONAL JOURNAL OF PHARMACEUTICAL SCIENCES 3813 | P a g e Rahul kr. Rai, Int. J. of Pharm. Sci., 2026, Vol 4, Issue 4, 3798-3816 | Review workflows and electronic prescribing systems, healthcare providers can move toward safer, more effective, patient-centric therapy, reducing trial- and-error prescribing and improving long-term outcomes. [24] 14.3. Role of AI in global health and neglected diseases AI has significant potential to address challenges in global health and neglected tropical, infectious, and orphan diseases, where conventional R&D is often economically unattractive. AI-enabled virtual screening and de-novo drug design can rapidly identify hit compounds against poorly explored targets, while repurposing AI tools can uncover new uses for existing, low-cost, off-patent drugs relevant to resource-limited settings. These approaches can shorten discovery timelines and reduce costs, making it feasible to develop treatments for diseases that disproportionately affect low- and middle-income countries. In addition, AI-driven diagnostics, predictive- epidemiology models, supply-chain- optimization tools can support early-outbreak treatment-allocation strategies, and detection, more efficient drug-distribution networks in regions. As global-data-sharing underserved initiatives and open-source AI platforms expand, AI-assisted research can democratize access to cutting-edge tools, allowing local scientists and institutions in drug-discovery programs for diseases that have long received inadequate investment.[23] to participate and trends: quantum-machine federated 14.4. Emerging learning, explainable AI, and learning Several emerging AI trends are expected to reshape pharmaceutical research. Quantum- machine learning (QML) promises to accelerate such as molecular- optimization problems combinatorial- and search conformation

  generalfuture work
  Keywords: drug systems diseases models pharmacology driven learning tools multi omics therapy design platforms personalized machine

Questions about this gap