Machine Learning for RNA-Targeting Drug Design
Clinical Snapshot
PICO Framework
| P — Population | RNA molecules as therapeutic targets (including structured RNAs, riboswitches, viral RNAs, and disease-associated non-coding RNAs) in preclinical drug discovery contexts |
| I — Intervention | Machine learning (ML) approaches applied to RNA-targeting drug design tasks, including binding site identification, virtual screening, and lead optimisation |
| C — Comparator | Protein-targeted ML drug design models; conventional computational chemistry approaches; RNA-specific versus protein-adapted methodologies |
| O — Outcomes | Performance of ML models in predicting RNA–small molecule interactions, binding site identification accuracy, virtual screening enrichment, and benchmarked drug-RNA interaction prediction metrics |
Bottom Line
This review from researchers at Mines Paris PSL and Vanderbilt University addresses a genuine and important gap in drug discovery science: the inapplicability of protein-centric machine learning models to RNA-targeting drug design. RNA molecules differ fundamentally from proteins in structure and small-molecule interaction profiles, necessitating bespoke computational approaches. The authors comprehensively compare existing ML tools across binding site identification and virtual screening tasks, and provide a benchmark evaluating current models on drug-RNA interaction prediction. The review's principal contribution is methodological — it identifies open challenges, critiques the lack of standardised evaluation frameworks, and proposes guidelines to address this. However, as a narrative review without a systematic search strategy, formal risk of bias assessment, or GRADE-rated evidence, its conclusions must be interpreted cautiously. There are no clinical outcome data, and direct patient applicability is not established. For senior clinicians, this paper is not practice-changing but signals an important preclinical frontier. RNA-targeting small molecules could eventually expand the druggable genome significantly. Clinicians in oncology, infectious disease, and rare genetic disorders should monitor this space, as methodological advances reviewed here may underpin future therapeutic candidates.
Key Findings
P Value: Not reported
Effect Size: Not reported in abstract; benchmark results are described qualitatively as assessing 'the ability of current machine learning models to predict specific drug-RNA interactions'
Primary Outcome: Comparative performance of current ML models in predicting specific drug-RNA interactions, assessed via a purpose-built benchmark
Nnt Or Sensitivity: Not applicable — preclinical computational review; relevant metrics would include AUC-ROC, enrichment factor, and binding affinity prediction RMSE for virtual screening tasks, but these are not available from the abstract
Confidence Interval: Not reported — no pooled statistical analysis performed
Clinical Application
The tools reviewed are research-grade computational methods. Clinical translation requires progression through hit-to-lead optimisation, in vitro and in vivo validation, and regulatory approval — none of which are addressed in this review. Feasibility for clinical application remains distant and contingent on resolving the methodological challenges identified RNA-targeting therapeutics represent an emerging frontier relevant to Australian drug discovery research (e.g., WEHI, CSIRO, university-based medicinal chemistry groups). No currently PBS-listed or TGA-approved small-molecule RNA-targeting drugs are identified in this review's scope. RACGP guidelines do not yet address RNA-targeted small molecules. Australian researchers engaged in RNA biology and computational drug discovery (particularly in oncology and infectious disease) would find this review methodologically informative. The review's call for standardised benchmarks aligns with international regulatory science priorities that TGA would need to engage with as this drug class matures. Not directly applicable to patient populations at this stage. Relevant to medicinal chemists, computational drug designers, and RNA biologists working on preclinical development of small-molecule RNA-targeting therapeutics, including those targeting viral RNAs (e.g., HIV, SARS-CoV-2), oncogenic non-coding RNAs, and riboswitches
Abstract
Targeting RNA with small molecules offers significant therapeutic potential. Machine learning could substantially accelerate preclinical drug discovery, from hit identification to lead optimization. Yet a limitation emerges: drug design machine learning models, designed for proteins, are not readily applicable to RNAs because of fundamental differences between RNAs and proteins in both structural characteristics and interactions with small molecules. RNA-specific approaches have consequently emerged, primarily focusing on binding site identification and virtual screening. In this review, we comprehensively compare machine learning tools for RNA-targeting drug design according to the tasks they address, their methodology and their relevance in RNA-specific contexts. As open challenges will catalyze new method development, we emphasize the need for standardized, drug design-specific evaluation approaches. We provide clear guidelines to establish these standards and present a benchmark assessing the ability of current machine learning models to predict specific drug-RNA interactions.
References
- 1.Karroucha, W., Oliver, C., Stoven, V., & Mallet, V. (2026). Machine learning for RNA-targeting drug design. Journal of Chemical Information and Modeling. https://doi.org/10.1021/acs.jcim.6c01094
Related Research
Cell
Deep learning of functional perturbations from condensate morphology
25 July 2026
ACS infectious diseases
Machine Learning Approaches Using High-Throughput Profiling Data for Antibiotic Discovery
11 July 2026
Briefings in bioinformatics
EssTFNet: integration of adaptive time-frequency and DNA language models for interpretable human essential gene prediction
5 July 2026
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