AuraTracer智迹闻
中文

EVENT DOSSIER

How can a Chinese company, established only 4 months ago, develop a globally leading full-atom biological large model?

2026-09-07 08:00 Models 🔥 42.2 heat score
1sources
1days unfolding
42.2heat score
4mentions
SummaryAI generated

On September 7, 2026, Atomelody, a Chinese company founded just 4 months prior, released Melo-1 Preview, its own developed universal biomolecular base model. This model surpassed current best practices in tasks involving proteins and peptides, as well as proteins and small molecules. Most of its performance results exceeded those of AlphaFold3, and it demonstrated absolute leadership in the Protein-peptide and Protein-ligand scenarios. Melo-1 addresses the shortcomings of mainstream full-atom models in peptide complexes, focusing on atomic-level interaction modeling for highly flexible molecules and incorporating physical and chemical constraints into its design to ensure structural validity. The core team members had previously tackled world-class targets such as GPCRs and delivered results from wet lab experiments.

Related eventsRELATED EVENTS
Key entitiesKEY ENTITIES
AlphaFold3Deep Spin TechnologyMelo-1Zhang Xiaonan

Coverage · reports per dayLANGUAGE SPLIT

Entity relations
AlphaFold3 × Deep Spin …1AlphaFold3 × Melo-11AlphaFold3 × Zhang Xiao…1Deep Spin Technology × …1Deep Spin Technology × …1Melo-1 × Zhang Xiaonan1

SignalsSIGNALS

Keyword heat
  • Deep Spin Technology1
  • Melo-11
  • AlphaFold31
  • Zhang Xiaonan1

All reports (1)SOURCES

凤凰网·科技 zh 2026-09-07 08:00

How can a Chinese company, established only 4 months ago, develop a globally leading full-atom biological large model?

Atomelody released its self-developed universal biomolecular base model, Melo-1 Preview, for the first time. This model has improved SOTA metrics in tasks such as protein-polypeptides and protein-small molecules, outperforming AlphaFold3 in most evaluation scenarios and leading significantly in the Protein-peptide and Protein-ligand scenarios. Melo-1 addresses the shortcomings of mainstream full-atom models in peptide complexes, focusing on atomic-level interaction modeling for high-degree-of-freedom molecules and incorporating physical and chemical constraints into its design to ensure structural validity. The core team members have successfully tackled world-class targets such as GPCR and delivered results from wet experiment validation.