Drug Discovery
Predicting small molecule–RNA interactions without RNA tertiary structures
Small molecules can bind RNAs to regulate their fate and functions, providing promising opportunities for treating human diseases. However, current tools for predicting small molecule-RNA interactions (SRIs) require prior knowledge of RNA tertiary structures. Here we present SMRTnet, a deep learning method that uses multimodal data fusion to integrate two large language models with convolutional and graph attention networks to predict SRIs on the basis of RNA secondary structure. SMRTnet achieves high performance across multiple experimental benchmarks, substantially outperforming existing tools. SMRTnet predictions for ten disease-associated RNA targets identified 40 hits of RNA-targeting small molecules with nanomolar-to-micromolar dissociation constants. Focusing on the MYC internal ribosome entry site, SMRTnet-predicted small molecules showed binding scores correlated closely with observed validation rates. One predicted small molecule downregulated MYC expression, inhibited proliferation and promoted apoptosis in three cancer cell lines. Thus, by eliminating the need for RNA tertiary structures, SMRTnet expands the scope of feasible RNA targets and accelerates the discovery of RNA-targeting therapeutics.
Unlocking RNA Therapeutics: Key Innovations and Impact
SMRTnet redefines the landscape of RNA drug discovery, offering unprecedented capabilities in predicting small molecule-RNA interactions. Explore the metrics that highlight its groundbreaking impact.
0.844+
Max auROC
92.6%
Top 1 Ranking
40
Validated Hits
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
SMRTnet utilizes a sophisticated deep learning architecture to predict small molecule-RNA interactions without relying on complex RNA tertiary structures. It integrates two large language models, convolutional neural networks, and graph attention networks to process both RNA secondary structure and small molecule chemical information.
Key components include an RNA encoder (RNASwan-seq + CNNs with ResNets), a small molecule encoder (MoLFormer + GAT), and a multimodal data fusion (MDF) module that progressively integrates pairwise binding information through co-attention and self-attention.
SMRTnet demonstrates robust performance across various experimental benchmarks, consistently outperforming existing tools like RNAmigos2 and traditional docking methods. Its auROC values range from 0.830-0.844 on the SMRTnet dataset, significantly higher than competitors.
In decoy evaluation tasks, SMRTnet achieves a mean ranking of 92.6% for true binders within the top position, far surpassing docking tools (27.3-46.6%) and other deep learning methods (16.0-23.8%). This highlights its superior ability to identify true binders from structurally similar small molecules.
Applied to ten disease-associated RNA targets, SMRTnet successfully identified 40 RNA-targeting small molecules with nanomolar-to-micromolar dissociation constants. This demonstrates its potential for accelerating the discovery of novel RNA therapeutics.
A specific case study on the MYC internal ribosome entry site (IRES) showed that SMRTnet-predicted binding scores correlated closely with observed validation rates. One compound, Irinotecan hydrochloride trihydrate (IHT), not only bound to MYC IRES but also significantly downregulated MYC expression, inhibited proliferation, and promoted apoptosis in cancer cell lines, showcasing direct biological activity.
SMRTnet's Predictive Workflow
Model Inputs (RNA Seq & Struct, SMILES)
→
Deep Neural Networks (Encoders & MDF)
→
Model Outputs (Binding Score)
→
Evaluations & Applications (Screening & Validation)
| Feature / Method | SMRTnet | RNAmigos2 | AutoDock Vina |
|---|---|---|---|
| RNA Tertiary Structure Required | No | Yes | Yes |
| Average auROC (PDB) | 0.830-0.844 | 0.567-0.596 | N/A |
| Mean Ranking (Decoy Eval) | 92.6% Top 1 | 16.0-23.8% Top 1 | 27.3-46.6% Top 1 |
| RNA Secondary Structure Utilized | Yes | No (3D required) | No (3D required) |
| Multimodal Data Fusion | Yes | No | No |
Breakthrough in Predictive Accuracy
0.844 Peak auROC on PDB Dataset, demonstrating superior predictive power without 3D RNA structures.MYC IRES Targeting with Irinotecan
SMRTnet successfully identified Irinotecan hydrochloride trihydrate (IHT) as a binder for the MYC IRES. This discovery is significant because MYC is a major oncogenic transcription factor, previously considered 'undruggable'. IHT's binding was confirmed experimentally, showing direct biological impact in cancer cell lines.
- MYC Expression Downregulation: IHT reduced MYC mRNA levels by ~56.9% and protein levels by ~71.6% in HeLa cells.
- Anti-Proliferative Effect: Decreased cell proliferation by 19.6-48.4% in HeLa, Jurkat, and Raji cancer cells.
- Apoptosis Induction: Increased cell apoptosis by 56.6-124.2% in these cell lines.
- Target Specificity: Luciferase reporter assays confirmed IHT's specific binding to the MYC IRES, not a fully base-paired control.
Advanced ROI Calculator
Estimate the potential financial and productivity gains your enterprise could achieve by implementing AI solutions based on SMRTnet's principles.
Estimated Annual Savings
$0
Hours Reclaimed Annually
0
Your AI Implementation Roadmap
A structured approach to integrate SMRTnet's capabilities into your drug discovery pipeline and beyond.
Data Integration & Expansion
Integrate multi-omics data (e.g., gene expression, transcriptomics) to predict broader biological effects beyond mere binding, bridging target engagement with functional impact.
High-Throughput Platform Development
Prioritize experimental high-throughput screening platforms for large-scale profiling of small molecule-RNA interactions, generating critical training data for advanced AI models.
AI-Driven Drug Discovery Refinement
Further refine AI methods to concurrently predict binding, downstream biological effects, and potential therapeutic outcomes, accelerating the development of RNA-targeting therapeutics.
Clinical Translation & Validation
Translate promising RNA-targeting small molecules into clinical candidates through rigorous validation, aiming to address previously 'undruggable' disease targets.
Ready to Transform Your Drug Discovery?
Leverage SMRTnet's innovative AI to accelerate the identification of RNA-targeting therapeutics and unlock new possibilities in treating complex diseases. Schedule a consultation to discuss how our solutions can integrate with your research.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat