XtalPi Names First In-House siRNA Candidate for IgA Nephropathy in Nine Months

A nine-month program built on generative AI

XtalPi, an AI- and robotics-driven drug and materials discovery company, announced the nomination of a preclinical candidate (PCC) for immunoglobulin A nephropathy (IgAN). The company completed the program in nine months with Kodexia, which it describes as its proprietary siRNA discovery and development platform and "the world's first to integrate generative AI and first-principles modeling in a closed loop of computational design, experimental validation, and iterative learning."

According to XtalPi, the milestone shows its end-to-end ability to move AI-guided molecular designs from target-region discovery to PCC nomination, turning platform capabilities into a proprietary RNA therapeutic asset.

IgAN is one of the world's most common primary glomerular diseases, affecting an estimated 10 million people globally, the company says. It is driven by the deposition of immune complexes that inflame the kidney's filtering structures and can lead to progressive renal decline and kidney failure.

A constrained design problem

siRNA therapeutics use the cell's natural RNA interference machinery to degrade specific messenger RNA (mRNA) and reduce production of a disease-associated protein. XtalPi notes that chemical modifications can improve stability and extend the duration of action, which creates the potential for less frequent dosing.

The company described this IgAN target as a highly constrained engineering challenge. The target mRNA offered limited viable sequence options, while competing clinical programs and an established patent landscape narrowed the design space further.

According to XtalPi, Kodexia responded by identifying a previously unexplored active target region and jointly engineering the siRNA sequence with sequence-specific chemical modifications. Freedom-to-operate considerations were addressed early in design. After iterative wet-lab evaluation of the initial AI-generated molecules, the company arrived at a candidate it says is distinguished mainly by potency, durability and safety, while preserving freedom to operate within the scope of the current intellectual property assessment.

Reported preclinical results

XtalPi reports the following data for the candidate:

• In vitro: subnanomolar silencing potency, with an IC50 tens of times lower than two clinical-stage reference molecules under corresponding assay conditions.
• Non-human primates: in head-to-head studies, a single dose sustained high-level target-protein suppression over a multi-month observation period, showing greater activity and durability than the reference molecules.
• Toxicology: studies indicated a broad safety window.

On this basis, the company says the candidate has best-in-class potential. These are company claims from preclinical work, and the release does not include clinical data.

How the platform works

Kodexia combines generative AI, first-principles modeling and automated experimentation. Rather than applying a fixed modification template after sequence selection, it jointly optimizes sequence and chemistry against potency, durability, off-target activity and safety. XtalPi says this approach has increased molecular design efficiency nearly threefold compared with conventional workflows, and that more than 50% of first-round designs tested across multiple programs showed stronger in vivo activity than their positive controls.

High-throughput experimentation and robotics support a continuous design, test and learn cycle, enabling hundreds of in vitro experiments and dozens of in vivo experiments each week. Results feed back into the models. In the IgAN program, the model's predicted activity ranking across the candidate set was confirmed in vivo, according to the company.

What comes next

XtalPi says it will keep advancing the IgAN candidate toward clinical development and wants to work with global partners on next-generation RNA interference therapies. It is also applying the platform to harder problems, including co-optimization of single-molecule, dual-target siRNAs, and exploring new conjugate strategies to deliver siRNA therapeutics beyond the liver.

For the market, the announcement is a data point on how AI-driven design may shorten early siRNA discovery, and it shows XtalPi building a proprietary pipeline rather than only offering platform services. Whether the candidate lives up to its preclinical profile will depend on clinical development, which has not yet begun.