Skip to main content

Zhejiang Lab

2 related articles

Subscribe to updates

Nature Communications Oct 31, 2026
DeepDelivery mines the human proteome for mRNA delivery scaffolds: 512 candidates, and 12 of 15 tested form particles that deliver
DeepDelivery treats the human proteome as a candidate pool for mRNA delivery scaffolds: deep learning models trained on viral capsid proteins, filtered by a dual-model screen, nominated 512 candidates from 109,375 human sequences. Of 15 candidates tested experimentally, 12 formed mRNA-encapsulating particles and mediated functional delivery; RTL8B reached 73.2% transduction in HEK293T reporter cells, above the reference proteins Arc (56.5%), PEG10 (53.1%) and PNMA2 (67.4%). The results support human proteins as a source of modularly engineerable delivery scaffolds, while immunogenicity and in vivo behavior remain untested.
DeepDelivery treats the human proteome as a candidate pool for mRNA delivery scaffolds: deep learning models trained on viral capsid proteins, filtered by a dual-model screen, nominated 512 candidates from 109,375 human sequences. Of 15 candidates tested experimentally, 12 formed mRNA-encapsulating particles and mediated functional delivery; RTL8B reached 73.2% transduction in HEK293T reporter cells, above the reference proteins Arc (56.5%), PEG10 (53.1%) and PNMA2 (67.4%). The results support human proteins as a source of modularly engineerable delivery scaffolds, while immunogenicity and in vivo behavior remain untested.
Chinese University of Hong Kong Zhejiang Lab ShanghaiTech University Chinese Academy of Sciences Shenzhen Institutes of Advanced Technology
Jie Zou（First author）Lanqing Li（First author, Corresponding author）Dongmin Xun（First author）Chao Zhong（Corresponding author）Pheng-Ann Heng（Corresponding author）Bolin An（Corresponding author）
DeepDelivery Nanocage-forming proteins mRNA delivery 1D-CNN Layer-wise Relevance Propagation
Nature Communications Oct 25, 2026
Permeability stops being one bulk number: cell-by-cell maps show spin-wave interference carving out local negative loss
High-frequency permeability, long treated as a single bulk parameter, is now computed as a field that varies with position and frequency. The approach discretizes a sample into micromagnetic cells, extracts a permeability spectrum per cell through time-to-frequency conversion, and suppresses Fourier error with a verified pulse window. A planar square model shows localized negative imaginary permeability at 14.10 GHz, where magnetization leads the field by 9π/10; on nanoscale iron spheres, machine-learning predictions of defect-to-spectrum trends agree with broadband measurements on four etched samples.
High-frequency permeability, long treated as a single bulk parameter, is now computed as a field that varies with position and frequency. The approach discretizes a sample into micromagnetic cells, extracts a permeability spectrum per cell through time-to-frequency conversion, and suppresses Fourier error with a verified pulse window. A planar square model shows localized negative imaginary permeability at 14.10 GHz, where magnetization leads the field by 9π/10; on nanoscale iron spheres, machine-learning predictions of defect-to-spectrum trends agree with broadband measurements on four etched samples.
Fudan University Zhejiang Lab
Wenbin You（First author）Ke Pei（First author）Ruixuan Zhang（First author）Xuebing Zhao（Corresponding author）Renchao Che（Corresponding author）
Spatial visualization of magnetic permeability Micromagnetic simulation Spin-wave propagation and interference Nanoscale iron spheres

You’re all caught up