Use OligoAI Online

OligoAI: Accurately Model and Optimize ASO Efficacy

OligoAI is a cutting-edge deep learning model capable of jointly modeling RNA target context, ASO sequence, sugar and backbone chemistries, and dosage to predict in vitro antisense oligonucleotide (ASO) efficacy.

How OligoAI Works

Antisense oligonucleotides (ASOs) are synthetic, single-stranded nucleic acids engineered to modulate gene expression. Cleavage and reduction of target RNA are achieved by recruiting RNase H to the DNA-RNA heteroduplex. Traditionally, discovering effective ASOs has required large-scale, costly experimental screening.

To overcome this, OligoAI leverages the ASO Atlas—a comprehensive database containing 188,521 RNase H-mediated ASO sequences targeting 334 unique genes along with quantitative knockdown efficacy data mined from published patents.

OligoAI features a transformer-based multi-stage encoding process:

• RNA Foundation Model: Independent encoding of ASO sequences and their target pre-mRNA contexts ($\pm50$ flanking nucleotides) uses RiNALMo-giga, a 650-million parameter bidirectional transformer trained on 36 million non-coding RNA sequences. This captures local secondary structure accessibility and sequence context effects.

• Chemical Tracking: Position-specific sugar modifications (2'-MOE, cEt, DNA) and backbone linkages (phosphorothioate, PS; phosphodiester, PO) are captured through learned embeddings.

• Cross-Modal Integration: These representations pass through a bottleneck network to capture complex sequence-chemistry interactions and combine with dosage-scaled transfection method embeddings via a multi-layer perceptron (MLP) to generate final percent knockdown predictions.

Unrivaled Predictive Performance

When evaluated across 299 held-out test screens, OligoAI significantly outclasses legacy thermodynamic-based predictive algorithms:

Real-World Wet-Lab Validation

• PSD3 Gene Screening: In a held-out evaluation of 2,234 ASOs targeting PSD3, the overall screen reached a median knockdown efficacy of 65.0%. The top 1% highest-scored ASOs by OligoAI achieved a remarkable median knockdown efficacy of 92.5%.

• KCNT2 Experimental Campaign: OligoAI was deployed to screen a virtual library of 200,374 unique 20-mer gapmer ASOs targeting KCNT2. Top-ranked candidates chosen by OligoAI achieved superior target knockdown (median = 81%) compared to randomly sampled sequences (median = 36%).

• 5.72-Fold Reduction in Screening Burden: Bootstrap analysis shows that matching the efficacy of OligoAI’s top-selected candidates using conventional random selection would require testing 103 ASOs instead of 18. This provides a 5.72-fold reduction in required experimental efforts, drastically reducing costs and accelerating therapeutic candidate discovery.

What is Tamarind Bio?

Tamarind Bio is a pioneering no-code bioinformatics platform built to democratize access to powerful computational tools for life scientists and researchers. Recognizing that many cutting-edge machine learning models are often difficult to deploy and use, Tamarind provides an intuitive, web-based environment that completely abstracts away the complexities of high-performance computing, software dependencies, and command-line interfaces.

Key features include a user-friendly graphical interface for setting up and launching experiments, a robust API for integration into existing research pipelines, and an automated system for managing and scaling computational resources. By handling the technical heavy lifting, Tamarind empowers researchers to concentrate on their scientific questions and accelerate the pace of discovery. The Tamarind team holds information/data security as a top priority, as detailed in our Trust Center & Terms of Service, ensuring your data is safe on the platform.

Accelerating Discovery with OligoAI on Tamarind Bio

Deploying OligoAI within the Tamarind Bio platform streamlines modern oligonucleotide therapeutic workflows:

• Overcoming Heuristic Hurdles: Traditional ASO design relies on independent, bifurcated sequence and chemistry optimization. OligoAI jointly models non-linear cross-modal sequence-chemistry interactions, capturing critical chemistry-specific sequence preferences.

• Context-Aware Therapeutics: OligoAI automatically factoring in genomic target region characteristics (such as the high efficacy of exonic and 3'UTR targets over splice junctions or introns) ensures that your designs target highly accessible regions.

• Dramatically Lower R&D Costs: By isolating the top 1% of optimal gapmer candidates in silico, you eliminate thousands of dollars in wasted wet-lab synthesis and screening steps.

How to Use OligoAI on Tamarind Bio

To leverage OligoAI's power on Tamarind Bio, a researcher can follow this streamlined workflow:

1. Access the Platform: Begin by logging in to the app.tamarind.bio website.

2. Select OligoAI: From the list of available computational tools, choose OligoAI.

3. Input ASO Sequences: Enter the candidate gapmer ASO nucleotide sequences you wish to evaluate.

4. Provide Target RNA Context: Input the pre-mRNA transcript sequence context (the hybridisation site along with flanking nucleotides).

5. Specify Chemical Modifications: Annotate position-specific modifications, selecting sugar chemistries (such as 2'-MOE or cEt) and backbone linkages (such as PS or PO).

6. Set Experimental Parameters: Define the prospective dosage (in nM) and intended transfection method (such as electroporation, lipofection, or gymnosis).

7. Predict and Prioritize: Run the model to predict the quantitative in vitro target RNA knockdown percentages. Export your results and filter for the top-scoring candidates to guide your downstream wet-lab validation.

Source