A COVID-19 DNA Vaccine Candidate Elicits Broadly Neutralizing Antibodies Against Multiple SARS-CoV-2 Variants Including the Currently Circulating Omicron BF.5, BF.7, BQ.1 and XBB (preprint)

Waves of breakthrough infections by SARS-CoV-2 Omicron subvariants pose a global challenge to pandemic control today. We have previously reported a pVAX1-based DNA vaccine candidate, pAD1002, which encodes a receptor-binding domain (RBD) chimera of SARS-CoV-1 and Omicron BA.1. In mouse and rabbit models, pAD1002 plasmid induced cross-neutralizing Abs against heterologous Sarbecoviruses, including SARS-CoV-1 and SARS-CoV-2 prototype, Delta and Omicron variants. However, these antisera failed to block the recent emerging Omicron subvariants BF.7 and BQ.1. To solve this problem, we replaced the BA.1-encoding DNA sequence in pAD1002 with that of BA.4/5. The resulting construct, namely pAD1016, elicited SARS-CoV-1 and SARS-CoV-2 RBD-specific IFN-gamma+ cellular responses in BALB/c and C57BL/6 mice. More importantly, pAD1016 vaccination in mice and rabbits generated serum Abs capable of neutralizing pseudoviruses representing multiple SARS-CoV-2 Omicron subvariants including BA.2, BA.4/5, BF.7, BQ.1 and XBB. As a booster vaccine for inactivated SARS-CoV-2 virus preimmunization in C57BL/6 mice, pAD1016 broadened the serum Ab neutralization spectrum to cover the Omicron BA.4/5, BF7 and BQ.1. These data highlight the potential benefit of pAD1016 in eliciting neutralizing Abs against broad spectrum Omicron subvariants in individuals previously vaccinated with inactivated prototype SARS-CoV-2 virus and suggests that pAD1016 is worthy further translational study as a COVID-19 vaccine candidate.

Potent Immunogenicity and Broad-Spectrum Protection Potential of Microneedle Array Patch-Based COVID-19 DNA Vaccine Candidates Encoding Dimeric RBD Chimera of SARS-CoV and SARS-CoV-2 Variants (preprint)

Breakthrough infections by SARS-CoV-2 variants pose a global challenge to pandemic control, and the development of more effective vaccines of broad- spectrum protection is needed. In this study, we constructed pVAX1-based plasmids encoding heterodimeric receptor-binding domain (RBD) chimera of SARS-CoV and SARS-CoV-2 Omicron BA.1 (RBDSARS/BA1), SARS-CoV and SARS- CoV-2 Beta (RBDSARS/Beta), or Omicron BA.1 and Beta (RBDBA1/Beta) in secreted form. When i.m. injected in mice, RBDSARS/BA1 and RBDSARS/Beta encoding plasmids (pAD1002 and pAD131, respectively) were by far more immunogenic than RBDBA1/Beta plasmid (pAD1003). Dissolvable microneedle array patches (MAP) laden with these DNA plasmids were fabricated. All 3 resulting MAP-based vaccine candidates, namely MAP-1002, MAP1003 and MAP-131, were comparable to i.m. inoculated plasmids with electroporation assistance in eliciting strong and durable IgG responses in BALB/c and C57BL/6 mice as well as rabbits, while MAP-1002 was comparatively the most immunogenic. More importantly, MAP-1002 significantly outperformed inactivated SARS-CoV-2 virus vaccine in inducing RBD-specific IFN-g+ T cells. Moreover, MAP-1002 antisera effectively neutralized pseudo- viruses displaying spike proteins of SARS-CoV, prototype SARS-CoV-2 or Beta, Delta, Omicron BA1, BA2 and BA4/5 variants. Collectively, MAP-based DNA constructs encoding chimeric RBDs of SARS-CoV and SARS-CoV-2 variants, as represented by MAP-1002, are potential COVID-19 vaccine candidates worthy further translational study.

Advanced Graph and Sequence Neural Networks for Molecular Property Prediction and Drug Discovery (preprint)

Properties of molecules are indicative of their functions and thus are useful in many applications. With the advances of deep learning methods, computational approaches for predicting molecular properties are gaining increasing momentum. However, there lacks customized and advanced methods and comprehensive tools for this task currently. Here we develop a suite of comprehensive machine learning methods and tools spanning different computational models, molecular representations, and loss functions for molecular property prediction and drug discovery. Specifically, we represent molecules as both graphs and sequences. Built on these representations, we develop novel deep models for learning from molecular graphs and sequences. In order to learn effectively from highly imbalanced datasets, we develop advanced loss functions that optimize areas under precision-recall curves. Altogether, our work not only serves as a comprehensive tool, but also contributes towards developing novel and advanced graph and sequence learning methodologies. Results on both online and offline antibiotics discovery and molecular property prediction tasks show that our methods achieve consistent improvements over prior methods. In particular, our methods achieve #1 ranking in terms of both ROC-AUC and PRC-AUC on the AI Cures Open Challenge for drug discovery related to COVID-19. Our software is released as part of the MoleculeX library under AdvProp.