Insider Brief
- Therna Biosciences introduced RNA-Logix, an AI platform designed to investigate RNA biology, evaluate therapeutic hypotheses against proprietary experimental data and design candidate medicines.
- BioReason-RNA, developed with NVIDIA, generates step-by-step explanations of how RNA sequences may function in cells, producing hypotheses for scientists to test.
- The platform draws on tens of millions of experimental measurements to refine hypotheses and select candidates for laboratory testing across mRNA, antisense and small interfering RNA therapies.
PRESS RELEASE — Therna Biosciences, a biotechnology company translating RNA language into genomic medicines, today showcased RNA-Logix™, its Recursive AI RNA Biologist platform. RNA-Logix brings together Therna’s RNA foundation models, proprietary experimental data, and an AI scientist that investigates biological questions, uncovers new biology, and designs candidate medicines.
Starting with a scientific question, RNA-Logix selects the appropriate tools from Therna’s AI stack, reasons over the biological mechanisms underlying disease, and identifies novel therapeutic avenues. It proposes novel molecular mechanisms and therapeutic hypotheses and verifies them against Therna’s large proprietary data lake. When the evidence conflicts with an explanation, the platform revisits and refines it. RNA-Logix also leverages the rapid verification loop that this strategy offers to further train its core reasoning models via reinforcement learning, establishing a landmark application of recursive self-improvement in biology.
“We built RNA-Logix to work through the questions an RNA biologist asks: What is happening, why is it happening, and how could we change it to treat disease?” said Nazli Azimi, Ph.D., Chief Executive Officer of Therna. “The platform brings our models and experimental data together to investigate those questions, challenge its own explanations, and design candidate medicines. Our collaboration with NVIDIA supports that work, from analyzing biological data to developing BioReason-RNA.”
BioReason-RNA, developed in collaboration with NVIDIA and built on NVIDIA Nemotron, is the reasoning brain within RNA-Logix. The RNA-native multimodal language model predicts how an RNA sequence functions within the cell through a causal mechanistic chain of molecular interactions. It works directly from sequence to produce a step-by-step biological explanation, giving scientists novel and readily testable hypotheses.
What sets RNA-Logix apart is its ability to verify these hypotheses against Therna’s proprietary experimental data lake, which contains tens of millions of measurements across hundreds of thousands of synthetic designs and dozens of cellular contexts. This enables scientists to refine hypotheses and select candidate medicines for laboratory testing. The platform applies to both RNA as medicine and RNA as a target; its capabilities span mRNA, antisense oligonucleotides, and small interfering RNAs.
“Most AI models in biology are black boxes: they map an RNA sequence to a predicted outcome but cannot articulate the why,” said Hani Goodarzi, Ph.D., Scientific Co-Founder of Therna, Core Investigator at the Arc Institute, and faculty at the University of California, San Francisco. “Our goal with BioReason-RNA was different. We wanted a model that reasons over the biology and lays out a causal chain of hypotheses behind each prediction, from sequence to molecular mechanism to downstream effects on diseased cells. Every link in that chain can be verified and tested against our large data corpus. When the chain holds, we learn new biology; when it breaks, we can see exactly where. That is how we get to the mechanisms behind diseases we still cannot treat and propose new genomic medicines.”
“Reasoning models are changing what scientists can ask of technology, and RNA is one of the hardest places to apply them,” said Anthony Costa, Director of Digital Biology, NVIDIA. “Therna built the foundation that makes reasoning over RNA possible. Working with their team on BioReason-RNA showed what becomes possible when deep biology and accelerated computing are built together.”
Therna Biosciences will provide hosted access to BioReason-RNA through selected platforms for research and industry partners, with documentation and demonstration cases available at https://gtcberlindemo.therna.
Therna’s collaboration with NVIDIA extends beyond BioReason-RNA to the computing tools that support RNA-Logix. The platform uses CodonFM, a BioNeMo open model for RNA design, rapids-singlecell, co-developed with scverse, for single-cell analysis, and NVIDIA Parabricks for genomic analysis. These capabilities support Therna’s experimental data layer, enabling RNA-Logix to examine biological evidence, check proposed mechanisms, and guide the design of candidate medicines.
Organizations interested in applying RNA-Logix™ to their own biology are invited to contact Therna about collaboration.
