International Journal of Molecular Sciences · Nov 17, 2026
Machine Learning-Guided Design of Peptide-Targeted CRISPR Vectors for Tissue-Specific Non-Coding RNA Knockdown
By Abdullah SR, Smail SW, Salihi A
CRISPR-based genome/epigenome editing and RNA-targeted therapeutics offer unprecedented precision for correcting disease-causing mutations, yet clinical translation remains constrained by two unresolved bottlenecks: suboptimal guide RNA (gRNA) design limiting editing efficiency and specificity, and the absence of safe, tissue-selective delivery platforms. While machine learning (ML)-driven gRNA design and peptide-mediated delivery have each advanced substantially in isolation, to our knowledge, no existing framework jointly optimizes predictive guide-design models and delivery-vector design within a single pipeline a disconnect that likely underlies persistent efficacy–safety trade-offs in clinical translation. This review addresses that gap by critically synthesizing ML-driven guide design and peptide-engineered delivery within a single mechanistic and computational framework. We first outline the structural basis of peptide-derived vectors, emphasizing how charge, hydrophobicity, secondary structure, and receptor-targeting functionalization govern organ- and cell-type specificity. We then critically appraise ML methods for predicting on-target activity, off-target propensity, and specificity across CRISPR platforms, alongside computational approaches for non-coding RNA target identification, focusing on training-data limitations, feature representation, and cross-platform generalizability that constrain predictive reliability. We synthesize design principles for tissue-specific delivery, including nanocarrier integration, and highlight preclinical cases where ML-guided peptide engineering enabled targeted delivery of mRNA, antisense oligonucleotides, and CRISPR ribonucleoproteins. Finally, we assess persistent barriers data sparsity, limited model transferability, off-target genotoxicity, immunogenicity, and delivery constraints and propose emerging directions: multi-omics integration, AI-enabled de novo peptide design, and translational frameworks. This review establishes a convergent paradigm in which joint ML-guided optimization of guide and delivery design enhances the specificity, efficiency, and safety of next-generation precision genetic medicines.