AI-Designed HIV Candidate Reaches Preclinical Stage in Record Seven Months

A biotech company called Viva-Thera, incubated and backed by the Hong Kong-listed firm XtalPi, has moved an experimental HIV drug into preclinical testing after seven months of discovery work—a timeline that would normally take years. The speed came from combining two things: Viva-Thera's deep structural understanding of how HIV infects cells, and XtalPi's artificial intelligence platform that can design molecules and test them at scale.

The target is a region of the HIV virus called MPER, short for membrane-proximal external region. It sits on the outer surface of the virus and is part of the machinery the virus uses to fuse with and enter human cells. What makes MPER attractive to researchers is that it barely changes across different strains of HIV—it retains about 90 percent genetic identity no matter which variant you're looking at. That consistency means a drug that works against one strain could work against many. But the region's position against the viral membrane and its shifting shape have made it notoriously difficult to target precisely. For decades, researchers have struggled to design molecules that can bind to it effectively.

Viva-Thera's founder and chief scientific officer, Qingshan Fu, had spent years studying MPER. His team achieved something that hadn't been done before: they obtained the first atomic-resolution structure of the full MPER region, including the transmembrane domain and cytoplasmic tail, using nuclear magnetic resonance. They then placed this structure on nanoparticle surfaces in membrane-like environments, so any drug candidate would be tested against the target as it actually exists in the virus, not in isolation.

With that structural foundation in place, XtalPi's generative AI models took over the screening work. The system navigated a chemical space of roughly 10 million compounds, identifying five validated hits that showed both desirable drug properties and strong antiviral activity. The team then used AI continuously to engineer and optimize those leads, pushing the program to a preclinical candidate in just seven months. The resulting molecule targets the fusion mechanism at MPER and binds to a site that existing HIV drugs—reverse-transcriptase inhibitors, integrase inhibitors, protease inhibitors—do not address. That means it could be combined with current treatments and potentially restore options for patients whose virus has developed resistance.

Viva-Thera is not betting everything on a single molecule. The company is pursuing four separate approaches to HIV simultaneously. The vaccine program uses the same MPER structural insights. In rhesus macaques, Viva-Thera's MPER immunogens triggered antibody levels about five times higher than a conventional HIV vaccine comparator. More importantly, the antibodies retained functional antiviral activity even when diluted 1,000-fold, suggesting the immune response could genuinely interfere with viral entry.

A third program targets the viral reservoir—the integrated HIV DNA that persists in cells even when antiviral drugs suppress active virus, and which can cause rebound infection if treatment stops. Viva-Thera is developing a non-replicating HIV-derived vector to deliver CRISPR-Cas9 gene-editing tools into CD4+ T cells, the immune cells that HIV infects. The team has completed vector construction and tested it in humanized mice; they are now preparing studies with live virus.

The fourth modality involves cell therapy, though fewer details were disclosed. Across all four programs, XtalPi's platform does the computational work: generative design for small molecules, structural analysis of immunogens, dosing models for gene therapy, and optimization of immunotherapies. That work is paired with robotic automation for high-throughput testing and rapid iteration. Viva-Thera provides the biological strategy, structural expertise, and experimental validation.

Fu framed the ambition plainly: to prevent HIV infection and ultimately cure it by combining interventions that block entry and erase the viral reservoir. XtalPi's chairman, Shuhao Wen, positioned the partnership as proof that AI's value in drug discovery should be measured by what it makes possible for patients. The company is betting that exceptional science, paired with computational power and automation, can accelerate progress on a disease that has resisted decades of conventional research. Whether the preclinical candidate advances to human trials, and whether the vaccine and gene-therapy programs deliver on their promise, will take years to determine. But the speed of the initial discovery phase suggests the partnership has found a way to compress timelines that have historically been a bottleneck in HIV drug development.