Yuxuan Zhong

Increasing the binding affinity of an antibody to its target antigen is a crucial task in antibody therapeutics development. This paper presents a pretrainable geometric graph neural network, GearBind, and explores its potential in in silico affinity maturation. Leveraging multi-relational graph construction, multi-level geometric message passing and contrastive pretraining on mass-scale, unlabeled protein structural data, GearBind outperforms previous state-of-the-art approaches on SKEMPI and an independent test set. A powerful ensemble model based on GearBind is then derived and used to successfully enhance the binding of two antibodies with distinct formats and target antigens. ELISA EC50 values of the designed antibody mutants are decreased by up to 17 fold, and KD values by up to 6.1 fold. These promising results underscore the utility of geometric deep learning and effective pretraining in macromolecule interaction modeling tasks. Increasing the binding affinity of an antibody to its target antigen is key for antibody therapeutics. Here the authors report a pretrainable geometric graph neural network, GearBind, and explore its potential in in silico antibody affinity maturation.

Abstract Increasing the binding affinity of an antibody to its target antigen is a crucial task in antibody therapeutics development. This paper presents a pretrainable geometric graph neural network, GearBind, and explores its potential in in silico affinity maturation. Leveraging multi-relational graph construction, multi-level geometric message passing and contrastive pretraining on mass-scale, unlabeled protein structural data, GearBind outperforms previous state-of-the-art approaches on SKEMPI and an independent test set. A powerful ensemble model based on GearBind is then derived and used to successfully enhance the binding of two antibodies with distinct formats and target antigens. ELISA EC 50 values of the designed antibody mutants are decreased by up to 17 fold, and K D values by up to 6.1 fold. These promising results underscore the utility of geometric deep learning and effective pretraining in macromolecule interaction modeling tasks.

As an emerging space propulsion technology, atmosphere-breathing electric propulsion (ABEP) system can operate for a long time in ultra-low earth orbit (ULEO) without carrying propellant from ground. However, the erosion of intake wall caused by the oxygen-contained atmosphere in ULEO will alter the wall characteristics to some degree, which influences the intake performance significantly. To analyze the influence of wall characteristics on intake performance, the direct simulation Monte Carlo (DSMC) method is used to simulate the gas flow in intake chamber under the mixed inflow of N₂ and atomic oxygen (AO). The density and velocity distribution of flow field for three intake configurations (parabola, cone and cylinder-cone type) are analysed under the varying wall characteristics (reflection coefficient σ from 0 to 1), and the corresponding variations of the collection and compression performance are also obtained. Results show that the increasing of σ leads to the decline of collection efficiency for all three intake configurations. The parabola type intake shows the highest collection efficiency of 93.67% when σ=0 under the specular reflection condition. The compression ratio increases with the value of σ for both parabola and cylinder-cone type, while the compression ratio for cone type intake initially increases then decreases, which reaches the maximum value of 44.36 when σ=0.5.

Atmosphere-Breathing Electric Propulsion (ABEP) is an international research focus. ABEP systems work in very low earth orbit (VLEO) can realize orbit maintenance of spacecrafts with little or no fuel. In this field, radio-frequency inductively coupled plasma (RF-ICP) thrusters can solve the problems of electrode corrosion and produce a dense plasma under thin atmosphere condition of VLEO. In this research, a RF-ICP source for the thruster configuration is designed, and a multi-physical coupling model is established. A complete simulation process is constructed to analyze the spatial distribution of the particles (e, N+, N2+, N, N2, N2S) in the RF-ICP. The results show that the particles in RF-ICP are axisymmetric in space, the electrons are mainly constrained in the central region of the ionization chamber by the magnetic field, and the plasma region expands and the electron density increases with the increase of the coil power. The other particles (N+, N2+, N, N2, N2S) are mainly distributed near the chamber wall as reactants or products of the surface reactions. And the distribution range of particles has a certain negative correlation with the electron energy required for excitation.

Abstract Background As one of the most promising classes of next-generation antibody therapeutics, bispecific antibodies (bsAbs) have gained increasing attention owing to their unique dual-targeting mechanisms. However, current bsAb formats often face challenges such as low expression levels, poor homogeneity, and unstable therapeutic efficacy due to their complex structures. Therefore, it is urgent to overcome the current technical limitations and develop novel formats of bsAbs with more stable structures and improved expression efficiency. Methods Through rational design and phage display-based screening, we engineered a novel light-chain single-domain antibody (VHHL). Using modular assembly and replacement strategies, the VHHL was reconstituted into conventional immunoglobulin G (IgG)s and the resulting bsAbs were comprehensively characterized by size-exclusion high-performance liquid chromatography, biolayer interferometry binding assay, enzyme-linked immunosorbent assay, and flow cytometry. Results A light chain engineering strategy combining complementarity-determining region 3 (CDR3)-grafting with site-directed mutagenesis of CDR1/CDR2 was developed to generate VHHLs. Through phage screening, two mouse CD16-specific VHHL candidates with favorable binding affinities and biophysical properties were identified, and one of which was structurally resolved via X-ray crystallography (3.05 Å resolution). When incorporated into full-length IgGs, the resulting bsAbs retained high structural similarity to natural monoclonal antibodies and maintained dual antigen-binding capabilities through their respective light and heavy chains. Conclusions Consequently, this study presents a novel IgG-format bsAb platform enabled by the integration of a rationally designed antigen-binding VHHL, providing a streamlined and versatile strategy for the development of multifunctional antibodies. Statement of Significance We developed a novel class of light-chain single-domain antibody (VHHL). By incorporating VHHL into full-length IgG, the resulting bispecific antibody (bsAb) retained the structural integrity of native IgG while preserving the independent antigen-binding activities of both the heavy and light chains, offering a simplified and flexible approach for bsAb generation.