PacBio and AI drug discovery company TychoBio are collaborating on a large-scale sequencing effort aimed at improving the design of RNA-based treatments for rare diseases. Under the collaboration, TychoBio plans to generate full-length RNA sequencing data from more than 10,000 samples using PacBio’s HiFi and Kinnex long-read sequencing technologies. The resulting datasets will be used to train computational models that could help researchers predict the activity, durability and potential unintended effects of RNA therapeutics.

The initial focus will be on steric blocking antisense oligonucleotides (SBOs), a class of RNA-targeting compounds designed to influence gene expression or RNA processing. TychoBio expects to eventually extend the approach to small interfering RNA (siRNA) and other RNA-targeting modalities. A central goal of the collaboration is to capture how candidate therapeutics alter RNA activity throughout the cell rather than evaluating their effects solely at an intended molecular target.

TychoBio will test candidate SBOs in multiple cell types and use PacBio sequencing to characterize resulting changes across the transcriptome. The company plans to incorporate these measurements into training datasets for AI models capable of associating the design of individual compounds with their biological effects.

The models are intended to help identify candidates with stronger on-target activity while also providing insight into how long their effects could persist and where unintended transcriptomic changes might occur. For rare disease drug developers, such information could be particularly valuable because RNA therapies often need to achieve a precise biological effect while limiting off-target activity.

“AI models are only as good as the data they are trained on,” says Felix Raimundo, Founder and CEO of TychoBio. He adds that PacBio’s HiFi technology provides access to full-length transcripts, allowing researchers to examine changes in RNA isoforms and splicing that may otherwise be difficult to capture.

According to TychoBio, a more detailed view of transcriptomic responses across different cell lineages could provide additional information about how SBOs work and help guide the development of more effective therapeutic candidates. The companies expect the sequencing program to produce a substantial dataset connecting RNA therapeutic designs with their downstream effects. Rather than relying solely on conventional measurements of target engagement, the approach is designed to give AI models a broader picture of cellular responses.

PacBio’s Kinnex technology is intended to facilitate high-throughput analysis of full-length RNA molecules, while HiFi sequencing provides long-read data with high accuracy. Together, the technologies will be used to profile thousands of sequences across multiple cell types.

The resulting data could also support the development of predictive models for both efficacy and toxicity — two factors that can determine whether an experimental RNA therapy progresses toward clinical testing.

PacBio has been involved in several initiatives focused on applying genomic and sequencing technologies to rare diseases, including collaborations with the HiFiSolves Consortium, n-Lorem Foundation, EspeRare, the GREGoR Consortium, Care4Rare Canada Consortium, Genetic Alliance and Genomics England. The company has also highlighted the role of high-quality biological datasets in the development of AI-based approaches to drug discovery.

“Our work with TychoBio highlights the critical value that high-quality sequencing data plays in generating the foundational data that will power biological modeling in the AI ecosystem,” says Mark Van Oene, President and CEO of PacBio.

The collaboration reflects a growing effort to combine high-throughput biological experimentation with AI-based drug discovery. For RNA therapeutics in particular, understanding effects across the transcriptome may help researchers move beyond simple target identification toward more predictive models of drug behavior.

For TychoBio, the immediate objective is to use the PacBio-generated data to improve its models for RNA therapeutic design. If successful, the approach could help researchers prioritize compounds based on predicted efficacy and duration while identifying potential unwanted effects earlier in the discovery process. The companies ultimately expect the integration of long-read sequencing and AI modeling to support the development of RNA therapies for rare diseases and potentially other conditions where precise control of RNA biology is therapeutically relevant.

www.tychobio.ai