San Diego – October 01, 2026 -- Avimer Bio and structural biologists at Spain's National Cancer Research Centre (CNIO) have published computational methods for building self-assembling protein cages with atomic precision, detailed in the Journal of the American Chemical Society (JACS). The study tackles a long-standing design challenge: predicting protein architecture when building blocks bend during assembly.
New algorithms predict how protein helices bend during assembly
Avimer Bio's Chief Scientific Officer Todd Yeates, who introduced geometric methods for protein cage nanoparticles decades ago at UCLA, said earlier techniques required precise, rigid shapes that made structural flexibility a design obstacle. The new computer algorithms predict how spiral-shaped alpha helices bend, letting researchers use molecular flexibility rather than work around it. Applying the method to four trimeric protein blocks with minimal mutations, the team produced cubic protein cages of twelve subunits and molecular masses exceeding 600 kilodaltons, stable in solution.
Cryo-EM confirms computational models to within 2 angstroms
High-resolution cryo-electron microscopy at CNIO in Madrid, led by Pablo San Segundo-Acosta and Roger Castells-Graells, validated the physical cages against computational predictions. Structures reached resolutions of 3.0 to 3.9 angstroms, with deviations between modeled and measured structures as low as 2 angstroms across the protein backbone, confirming the predicted helix-bending patterns.
Platform targets autoimmune disease and oncology applications
Designed protein cages present multiple copies of therapeutic proteins in controlled arrangements to engage cell receptors that require clustering to activate signaling pathways, extending beyond the two binding arms available on a conventional antibody. Potential applications include autoimmune conditions such as rheumatoid arthritis and ulcerative colitis, where targeted engagement of immune cells could help control inflammation, and oncology, where cage-displayed proteins could cluster receptors that stimulate immune cells against tumors.
Design approach preserves near-native human protein sequences
The platform retains much of the starting protein sequence, including in designs built from human protein building blocks, offering a basis for evaluating immune response risk. Each cage is built from a single type of protein subunit, simplifying manufacturing and maintaining consistent composition compared with multi-chain particle designs.
Robin Aglietti and Peter Bowers conducted design work at Avimer Bio alongside Yeates, while Castells-Graells directed CNIO's structural biology analysis. The study, "Design and Structure of Protein Cages Based on Helical Fusion and Machine Learning," appears in JACS.