← Back to Home

Targeting HIV Entry into Cells, Viva-Thera Uses AI to Select a Preclinical Drug Candidate

Molecular discovery and optimization took seven months, combining viral structural research, membrane environment reconstruction, and AI design. However, selecting a candidate is only the starting point for development. Whether it can safely block infection remains to be answered through comprehensive experiments and human trials.

By SURL BioNews

The moment when HIV fuses with the cell membrane is one checkpoint researchers are trying to intercept to prevent the virus from infecting cells. Viva-Thera, which was incubated by XtalPi and in which XtalPi holds an equity stake, is designing small-molecule drugs targeting a viral protein involved in this process, hoping to add an option with a different mechanism of action to future antiviral treatment.

XtalPi announced on September 22 that the two companies had selected a preclinical drug candidate targeting the membrane-proximal external region (MPER) of the HIV-1 gp41 protein; BioPharma APAC also reported the news the following day. The company said molecular discovery and optimization took seven months. This means the team selected a molecule for further development, not that it has received approval to enter human trials.

MPER is located near the viral membrane and forms part of HIV’s membrane fusion machinery. Effectively disrupting this region could block the virus from entering cells. However, its proximity to the membrane surface and changing conformation also make it difficult for drugs to access precisely. The company hopes this mechanism could complement existing treatment combinations in the future, but publicly available data do not yet support its effectiveness against drug-resistant viruses.

To make screening more closely reflect the target’s actual state, Viva-Thera used solution nuclear magnetic resonance to study structures encompassing MPER, the transmembrane region, and the cytoplasmic tail, then reconstructed a membrane environment containing the target protein on the surface of nanoparticles. This biological foundation is key to understanding the AI application here: the algorithms must work with a protein region embedded at the membrane boundary whose shape can change.

According to XtalPi, generative AI explored a chemical space of approximately 10 million compounds, yielding five experimentally validated initial hit compounds. Further design and optimization then led to the selection of the drug candidate. This figure describes the scope of computational exploration and should not be interpreted as meaning that all 10 million molecules were synthesized and experimentally tested. The seven-month timeline also did not include all the preliminary structural research supporting the program.

The evidence currently comes mainly from company disclosures, and BioPharma APAC’s coverage is consistent with the announcement, so it cannot yet be regarded as independent validation. Although the announcement described the hit compounds as having antiviral activity and desirable drug properties, it did not provide specific results for the candidate’s potency, toxicity, pharmacokinetics, or drug-resistance testing. Nor did it provide comparative data sufficient to assess the efficiency of AI against conventional workflows.

This development therefore more clearly illustrates how structural research, computational design, and experimental screening can be connected. Whether the candidate molecule can next maintain sufficient exposure in vivo while meeting safety requirements, and accumulate evidence to support an application for human trials, will determine whether this faster discovery process can lead to a treatment available to patients.

References

  1. BioPharma APAC
  2. XtalPi