The Use of Deep Learning in RNA Therapeutic Development
Clinical Snapshot
PICO Framework
| P — Population | Not applicable in the traditional clinical sense — the review addresses RNA therapeutic development pipelines, encompassing mRNA therapeutics, siRNA, CRISPR-based gene editing tools, and RNA delivery systems across preclinical and early clinical contexts |
| I — Intervention | Deep learning model architectures (including convolutional neural networks, recurrent neural networks, transformers, graph neural networks, and generative models) applied to RNA structure prediction, CRISPR activity optimisation, and RNA delivery system design |
| C — Comparator | Conventional computational methods, experimental high-throughput screening, and physics-based modelling approaches for RNA therapeutic optimisation |
| O — Outcomes | Model performance in RNA secondary and tertiary structure prediction accuracy, CRISPR guide RNA activity and specificity prediction, lipid nanoparticle and delivery system optimisation, and generalisation of deep learning models across RNA therapeutic design tasks |
Bottom Line
This narrative review from a leading MIT-based RNA therapeutics group provides a technically comprehensive overview of how deep learning is being applied to accelerate RNA therapeutic development across three domains: RNA structure prediction, CRISPR guide RNA optimisation, and RNA delivery system design. The review is well-scoped and authored by recognised experts, making it a valuable orientation for researchers entering this field. However, clinicians should note that this is a narrative — not systematic — review, with no formal search strategy, no risk of bias assessment, and no quantitative synthesis. The evidence base described is predominantly preclinical and computational, with no patient-level clinical outcomes synthesised. Deep learning tools in this space show genuine promise for navigating the vast RNA design space, but significant gaps remain between in silico optimisation and clinical translation. The authors honestly acknowledge data scarcity, model generalisation challenges, and computational barriers. For Australian clinicians, this review contextualises the research pipeline underpinning next-generation RNA therapeutics that may eventually reach TGA evaluation and PBS consideration, but no immediate changes to clinical practice are warranted based on this work alone.
Key Findings
P Value: Not applicable at the review level; no meta-analytic statistics performed
Effect Size: No pooled effect size reported; individual model performance metrics from primary studies are discussed qualitatively (e.g., improvements in structure prediction accuracy, CRISPR on-target activity scores, and delivery efficiency predictions relative to baseline computational or experimental methods)
Primary Outcome: Narrative synthesis of deep learning model performance across RNA secondary and tertiary structure prediction (e.g., applications of AlphaFold-derived and RNA-specific architectures), CRISPR guide RNA activity and off-target prediction, and lipid nanoparticle / RNA delivery system optimisation
Nnt Or Sensitivity: Not applicable — this is a preclinical/computational review. No NNT, sensitivity, specificity, or hazard ratio data are synthesised. Model performance is discussed in terms of prediction accuracy, area under the curve (AUC), and correlation with experimental outcomes where reported in primary studies
Confidence Interval: Not reported at the review level; individual study confidence intervals not systematically extracted or synthesised
Clinical Application
Deep learning tools for RNA therapeutic development are currently in the research and early translational phase. Clinical implementation requires substantial further validation, including prospective experimental confirmation of computationally optimised candidates, regulatory-grade safety and efficacy data, and scalable manufacturing. The review identifies data scarcity and computational resource requirements as current barriers to broader adoption. Australia has an active RNA therapeutics research ecosystem, including the MRFF-funded RNA Therapeutics Initiative and collaborations through the Doherty Institute and various university research centres. The TGA has established regulatory pathways for RNA-based medicines following COVID-19 mRNA vaccine approvals (Comirnaty, Spikevax). PBS listing of RNA therapeutics remains limited to specific approved products; broader RNA therapeutic classes (e.g., siRNA for rare diseases such as inclisiran for hypercholesterolaemia, patisiran for hATTR amyloidosis) are subject to ongoing PBAC evaluation. RACGP guidelines do not yet address deep learning-assisted RNA therapeutic design, as this remains a research-stage technology. Australian clinicians and researchers should monitor TGA guidance on novel RNA therapeutic modalities as the pipeline matures. This review is primarily relevant to researchers and translational scientists working in RNA therapeutic development, including mRNA vaccine design, siRNA therapeutics, CRISPR-based gene editing, and lipid nanoparticle delivery system optimisation. It is not directly applicable to patient care at this stage. Clinicians in oncology, rare genetic disease, infectious disease, and gene therapy fields may find the translational roadmap informative for understanding the pipeline of next-generation RNA therapeutics.
Abstract
Ribonucleic acid (RNA)-based therapeutics have emerged as promising methods of disease treatment due to their ability to target the human genome and influence protein production, their versatility, and their relative lack of toxicity compared to other gene therapies. However, the RNA therapeutic design space is extremely large, encompassing multiple variables, including codon identities, secondary structure, and design of specific regions. RNA therapeutic optimization is difficult due to the impracticality of exploring such a vast design space experimentally. To address this limitation, deep learning methods have been employed to optimize RNA therapeutic development. In this review, we examine the application of deep learning models across three key aspects of RNA therapeutic development (RNA structure prediction, CRISPR activity, and RNA delivery), highlighting major contributions in these fields and analyzing how deep learning model architectures could affect model performance. We then discuss challenges associated with using deep learning for RNA therapeutics, such as computational and data limitations. Finally, we offer perspectives on areas for future exploration, such as emerging model architectures and methods of integration with more advanced high-throughput screening techniques. Ultimately, this review provides an overview of how deep learning is used in RNA therapeutic development and how it can evolve in the future.
References
- 1.Subramanian, D. A., Yao, S. L., Chan, A., Witten, J., Reker, D., Anderson, D. G., Langer, R., & Traverso, G. (2026). The use of deep learning in RNA therapeutic development. ACS Nano. https://doi.org/10.1021/acsnano.6c02090
This content is for educational purposes for healthcare professionals only and does not constitute clinical advice. Clinical decisions should be based on individual patient assessment, current guidelines, and appropriate specialist consultation. Editorial Standards · Privacy Policy · Terms of Service