The next frontier of peptide therapeutics: from targeted conjugates (PDC/ODC) to AI-driven safety-by-design

The pharmaceutical landscape is witnessing a profound “Peptide Renaissance,” driven primarily by the clinical success of GLP-1 receptor agonists and reinforced by engineering advanced drug delivery platforms. This review provides a comprehensive analysis of current trends in peptide-based drug development, using a target-localization framework that distinguishes molecular recognition from pharmacological action. We delineate the functional divergence between extracellular, cell-surface, and intracellular targeting, highlighting how peptides bridge the gap between small molecules and monoclonal antibodies. Central to this discussion is a comparative evaluation of next-generation conjugation technologies—Peptide-Drug Conjugates (PDCs) and Oligonucleotide-Drug Conjugates (ODCs)—against the backdrop of conventional Antibody-Drug Conjugates (ADCs). Rather than a simple performance comparison, this analysis focuses on how the distinct structural architectures of these modalities dictate their pharmacological profiles. We examine the inherent trade-offs between the long systemic half-life of bulky, antibody-based scaffolds and the construct-dependent tissue distribution and immunogenicity of peptide and oligonucleotide delivery systems, thereby elucidating their respective strengths and limitations in targeting complex disease environments. PDCs can use peptide-mediated recognition or uptake to deliver therapeutic payloads, whereas ODCs deliver oligonucleotides that act intracellularly through mechanisms such as RNA interference or splice modulation. Furthermore, we explore the transformative role of generative Artificial Intelligence and structural algorithms, such as AlphaFold 3, in shifting the development paradigm toward a “Safety-by-Design” philosophy. By integrating predictive toxicology (e.g., pepADMET) into the nascent design phase, researchers may prioritize candidates for experimental testing. Importantly, the pharmacoeconomic case for this approach remains provisional rather than established. A retrospective analysis of AI-derived candidates reported Phase I success rates of approximately 80%–90% and a Phase II success rate of approximately 40%, comparable to historical Phase II averages. The limited sample and historical comparison do not establish a causal effect of AI or a peptide-specific economic benefit. We therefore present AI-guided safety-by-design as a proposed strategy for candidate prioritization, and set out explicitly the comparative evidence that would be required before claims of reduced development cost, shortened timelines, or improved affordability can be regarded as demonstrated.