Multi-omics and artificial intelligence for precision drug discovery and potential clinical applications
TL;DR
Integrating multi-omics data with artificial intelligence accelerates precision drug discovery by shifting from single-target approaches to network-based models, significantly reducing development timelines and improving target identification.
Problem / question
Traditional drug discovery faces a 90 percent clinical trial failure rate and 2.6 billion USD average cost because reductionist one-drug-one-target models fail to capture the complex, multi-layered molecular interactions driving diseases like cancer and neurodegeneration.
Methods
The authors conducted a systematic literature review across PubMed, Web of Science, and Scopus to analyze the integration of multi-omics technologies, such as single-cell RNA sequencing, MERFISH spatial transcriptomics, and mass spectrometry, with AI architectures, including Graph Neural Networks, Transformers, and Generative Adversarial Networks, across the drug development pipeline.
Key findings
AI and multi-omics integration has yielded concrete successes, such as the generative model GENTRL designing a DDR1 kinase inhibitor in just 21 days, and AI identifying the TNIK inhibitor INS018_055 for pulmonary fibrosis in 18 months. AI-designed SOMAmer therapeutics reduced phase II development time by 60 percent. Furthermore, AI models achieved 89 percent accuracy in predicting drug-induced liver injury and successfully repurposed existing drugs, such as identifying the drug baricitinib for Alzheimer's disease and the antidepressant vortioxetine for glioblastoma.
Why it matters
This convergence replaces slow, linear trial-and-error screening with rapid, data-driven computational design, enabling the targeting of complex disease networks and paving the way for personalized therapies using patient-specific digital twins.
Limitations
Major hurdles include the lack of standardized data across public repositories, the static nature of current omics measurements that miss dynamic disease changes, the black-box opacity of deep learning models, and the risk of algorithmic biases amplifying healthcare inequities.
Takeaway
Combining multi-omics profiling with AI fundamentally transforms drug discovery, enabling the rapid design of novel compounds and the repurposing of existing drugs by mapping and targeting the entire molecular network of a disease.