Jianbo Dong

Coronaviruses cause severe human viral diseases including SARS, MERS and COVID-19. Most recently SARS-CoV-2 virus (causing COVID-19) has led to a pandemic with no successful therapeutics. The SARS-CoV-2 infection relies on trimeric spike (S) proteins to facilitate virus entry into host cells by binding to ACE2 receptor on host cell membranes. Therefore, blocking this interaction with antibodies are promising agents against SARS-CoV-2. Here we describe using humanized llama antibody VHHs against SARS-CoV-2 that would overcome the limitations associated with polyclonal and monoclonal combination therapies. From two llama VHH libraries, unique humanized VHHs that bind to S protein and block the S/ACE2 interaction were identified. Furthermore, pairwise combination of VHHs showed synergistic blocking. Multi-specific antibodies with enhanced affinity and avidity, and improved S/ACE2 blocking are currently being developed using an in-silico approach that also fuses VHHs to Fc domains. Importantly, our current bi-specific antibody shows potent S/ACE2 blocking (KD – 0.25 nM, IC100 ∼ 36.7 nM, IC95 ∼ 12.2 nM, IC50 ∼ 1 nM) which is significantly better than individual monoclonal VHH-Fcs. Overall, this design would equip the VHH-Fcs multiple mechanisms of actions against SARS-CoV-2. Thus, we aim to contribute to the battle against COVID-19 by developing therapeutic antibodies as well as diagnostics.

At fertilization in mammals, the sperm activates the egg by inducing a series of oscillations in the intracellular free Ca(2+) concentration. There is evidence showing that this oscillatory event is triggered by a sperm-derived protein factor which diffuses into egg cytoplasm after gamete membrane fusion. At present the identity of this factor and its precise mechanism of action is unknown. Here, we studied the specificity of action of the sperm factor in triggering Ca(2+) oscillations in mammalian eggs. In doing so, we examined the patterns of Ca(2+) signaling in mouse eggs, zygotes, parthenogenetic eggs and maturing oocytes following the stimulation of bovine sperm extracts which contain the sperm factor. It is observed that the sperm factor could induce Ca(2+) oscillations in metaphase eggs, maturing oocytes and parthenogenetically activated eggs but not in the zygotes. We present evidence that Ca(2+) oscillations induced by the sperm factor require a maternal machinery. This machinery functions only once in mammalian oocytes and eggs, and is inactivated by sperm-derived components but not by parthenogenetic activation. In addition, it is found that neither InsP(3) receptor sensitivity to InsP(3) nor Ca(2+) pool size are the determinants that cause the fertilized egg to lose its ability to generate sperm-factor-induced Ca(2+) oscillations at metaphase. In conclusion, our study suggests that the orderly sequence of Ca(2+) oscillations in mammalian eggs at fertilization is critically dependent upon the presence of a functional maternal machinery that determines whether the sperm-factor-induced Ca(2+) oscillations can persist.

Deep neural networks (DNNs) have gained tremendous attractions as compelling solutions for applications such as image classification, object detection, speech recognition, and so forth. Its great success comes with excessive trainings to make sure the model accuracy is good enough for those applications. Nowadays, it becomes challenging to train a DNN model because of 1) the model size and data size keep increasing, which usually needs more iterations to train; 2) DNN algorithms evolve rapidly, which requires the training phase to be short for a quick deployment. To address those challenges, distributed training platforms have been proposed to leverage massive server nodes for training, with the hope of significant training time reduction. Therefore, scalability is a critical performance metric to evaluate a distributed training platform. Nevertheless, our analysis reveals that traditional server clusters have poor scalability for training due to the traffic congestions within the server and beyond. The intra-server traffic on the I/O fabric can result in severe congestions and skewed quality of service as high performance devices are competing with each other. Moreover, the traffic congestions on the Ethernet for inter-server communication could also incur significant performance degradation. In this work, we devise a novel distributed training platform, EFLOPS, that adopts an algorithm and system co-design methodology to achieve good scalability. A new server architecture is proposed to alleviate the intra-server congestions. Moreover, a new network topology, BiGraph, is proposed to divide the network into two separate parts, so that there is always a direct connection between any nodes from different parts. Finally, accompany with BiGraph, a topology-aware allreduce algorithm is proposed to eliminate the traffic congestion on the direct connection. The experimental results show that eliminating the congestions on network interface can gain up to 11.3xcommunication speedup. The proposed algorithm and topology can provide further improvement up to 6.08x. The overall performance of ResNet-50 training achieves near-linear scalability, and is competitive to the top-rankings of MLPerf results.