Search

Distributed systems have been widely adopted for deep neural networks model training. However, the scalability of distributed training systems is largely bounded by the communication cost. We design a highly efficient collective communication library, namely Alibaba Collective Communication Library (ACCL), to build distributed training systems with linear scalability. ACCL provides optimized algorithms to fully make use of heterogeneous interconnects simultaneously. And the experimental results show significant performance improvement.

In this review, we have summarized the current landscape of therapeutic antibody optimization for successful development. By engineering antibodies with display technology, computer-aided design and site mutagenesis, various properties of the therapeutic antibody candidates can be improved with the purpose of enhancing their safety, efficacy and developability. These properties include antigen binding affinity and specificity, biological efficacy, pharmacokinetics and pharmacodynamics, immunogenicity and physicochemical developability features. A best-in-class strategy may require the optimization of all these properties to generate a good therapeutic antibody.

SARS-CoV-2 and its variants have persisted in this ongoing COVID-19 pandemic. While the vaccines have greatly reduced the COVID-19 cases, hospitalizations, and death, about half of the world remain unvaccinated due to various reasons. Furthermore, the duration of the immunity gained from COVID-19 vaccination is still unclear. Therefore, there is a need for innovative prophylactic and treatment measures. In response to this need, we previously reported on the successful computer-aided development of potent VHH-based multispecific antibodies that were characterized in vitro. Here, we evaluated in vivo efficacy and safety of the lead trispecific VHH-Fc, ABS-VIR-001. Importantly, our data showed that ABS-VIR-001 treatment prevented SARS-CoV-2 infection and death when provided as an intranasal prophylaxis in a humanized ACE-2 mouse model. In addition, ABS-VIR-001 post-exposure treatment was shown to greatly reduce viral loads by as much as 50-fold. A detailed panel of metabolic and cellular parameters demonstrated that ABS-VIR-001 treatment was overall comparable to the PBS treatment, indicating a favorable safety profile. Notably, our inhibition studies show that ABS-VIR-001 continued to demonstrate unwavering efficacy against SARS-CoV-2 mutants, associated with key variants including Delta and Omicron, owing to its multiple epitope design. Lastly, we rigorously tested and confirmed the excellent thermostability of ABS-VIR-001 when heated to 45 °C for up to 4 weeks. Taken together, our study suggests that ABS-VIR-001 is an efficacious and durable prophylaxis and post-exposure treatment for COVID-19 with promising safety and manufacturability features for global distribution.

Background Bispecific T cell engaging antibodies (TEAs) with one arm targeting a cancer antigen and another arm binding to CD3 have demonstrated impressive efficacy in multiple clinical studies. However, establishing a safety/efficacy balance remains challenging. For instance, some TEAs have severe safety issues. Additionally, not all patients or all cancer cells of one patient respond equally to TEAs. Methods Here, we developed a next-generation bispecific TEA with better safety/efficacy balance and expanded mechanisms of action. Using the computer-aided antibody design strategy, we replaced heavy chain complementarity-determining regions (HCDRs) in one Rituximab arm with HCDRs from a CD3 antibody and generated a novel CD20/CD3 bispecific antibody. Results After series of computer-aided sequence optimization, the lead molecule, GB261, showed great safety/efficacy balance both in vitro and in animal studies. GB261 exhibited high affinity to CD20 and ultra-low affinity to CD3. It showed comparable T cell activation and reduced cytokine secretion compared with a benchmark antibody (BM). ADCC and CDC caused by GB261 only killed CD20+ cells but not CD3+ cells. It exhibited better RRCL cell killing than the BM in a PBMC-engrafted, therapeutic treatment mouse model and good safety in cynomolgus monkeys. Conclusions Thus, GB261 is a promising novel TEA against CD20+ cancers.

Carbon emission reduction (CER) is increasingly becoming a global issue. This study explored the impact mechanism of upgrading of consumption structure (UCS) and human capital level (HC) on carbon emissions, and an empirical test was carried out using the provincial panel data from 2000 to 2019 in China. The results show the following: (1) China’s UCS could significantly curb carbon emissions. (2) At present, China’s HC is positively correlated with carbon emissions. The higher the level of human capital, the less conducive to CER. Additionally, the moderating effect of HC could inhibit the CER induced by UCS. (3) Regional heterogeneity analysis showed that the UCS in the central and western regions of China was conducive to CER, while the estimated coefficient of UCS on CER in the eastern region was not significant. (4) The UCS could reduce carbon emissions by stimulating the mediating effect of industrial upgrading. Based on empirical study results, this study proposes policy suggestions that can help reduce China’s carbon emissions.

The Beijing–Tianjin–Hebei region is the most dynamic region and largest economy in northern China; however, the air quality is the worst in the country. The study of the air quality in the cities around Beijing is of great significance for air pollution control. Therefore, this study analyzed whether the COVID-19 pandemic altered the periodic pattern of the air quality in Beijing and its neighboring cities. The study employed continuous wavelet transform to examine the impact of the COVID-19 pandemic on the air quality of Beijing and its neighboring cities. This method reveals the changes in the air quality from a periodic pattern perspective. The results showed that COVID-19 weakened the periodic changes in air quality in Beijing and five neighboring cities, and this effect was most pronounced during the outbreak of the pandemic in early 2020. The cycle synchronization analysis showed that the pandemic weakened the cycle synchronization of air quality of the cities in the north of Beijing, while less impact was found on the cities to the south of Beijing. Moreover, the periodic patterns in 2020 and 2021 were compared with that in 2019 (before the outbreak of the pandemic), and it was found that the periodic patterns during the outbreak of the pandemic in 2020 and 2021 were significantly different from that in the same period in 2019. Therefore, COVID-19 weakened the periodic pattern of air quality in the cities around Beijing and altered the connection to air quality among them. The changes reveal the connections of inter-city air pollutants caused by human economic and social activities in cities around Beijing.

Distributed deep learning (DDL) uses a cluster of servers to train models in parallel. This has been applied to a multiplicity of problems, e.g. online advertisement, friend recommendations. However, the distribution of training means that the communication network becomes a key component in system performance. In this paper, we measure the Alibaba's DDL system, with a focus on understanding the bottlenecks introduced by the network. Our key finding is that the communications overhead has a surprisingly large impact on performance. To explore this, we analyse latency logs of 1.38M Remote Procedure Calls between servers during model training for two real applications of high-dimensional sparse data. We reveal the major contributors of the latency, including concurrent write/read operations of different connections and network connection management. We further observe a skewed distribution of update frequency for individual parameters, motivating us to propose using in-network computation capacity to offload server tasks.

Consolidated server racks are quickly becoming the standard infrastructure for engineering, business, medicine, and science. Such servers are still designed much in the way when they were organized as individual, distributed systems. Given that many fields rely on big-data analytics substantially, its cost-effectiveness and performance should be improved, which can be achieved by flexibly allowing resources to be shared across nodes. Here we describe Venice, a family of data-center server architectures that includes a strong communication substrate as a first-class resource. Venice supports a diverse set of resource-joining mechanisms that enables applications to leverage non-local resources efficiently. We have constructed a hardware prototype to better understand the implications of design decisions about system support for resource sharing. We use it to measure the performance of at-scale applications and to explore performance, power, and resource-sharing transparency tradeoffs (i.e., how many programming changes are needed). We analyze these tradeoffs for sharing memory, accelerators, and NICs. We find that reducing/hiding latency is particularly important, the chosen communication channels should match the sharing access patterns of the applications, and of which we can improve performance by exploiting inter-channel collaboration.

Background: Immune checkpoint interaction helps tumors resist immunity-induced apoptosis. Blocking PD-1/PD-L1 leads to remarkable clinical responses; however, a large portion of patients develop acquired resistance after initial therapy. Many tumor-specific antigens (TSAs) and tumor-associated antigens (TAAs) have been identified. TSAs are restricted to cancer cells only but TAAs are elevated on tumor cells and minimally expressed on healthy cells. Therefore, co-targeting TAAs/TSAs would show benefits over monotherapy. CD55 is a TAA that regulates complement activation pathway and expressed on cells exposed to the complement system to protect against complement attack. Cancer cells escape this attack by raising CD55 expression. Additionally, CD55 acts as a virus receptor for internalization. Based on the roles of CD55, we designed PD-L1/CD55 bispecific antibody (GB262) with weaker CD55 binding arm to rule out the possibility of lethality to healthy cells but maintain its CD55 binding and internalization ability specifically to cancer cells. Moreover, co-binding to PD-L1 and CD55 leads to co-internalization of PD-L1 by CD55, enhancing the blocking of PD-1/PD-L1 interaction. Methods: Co-localization and levels of PD-L1 and CD55 on cancer cells were analyzed by microscopy and FACS. T-cell activation was studied using PD-L1+/CD55+ cells and PBMCs. CD97 and CD55 blocking was performed using FACS. Cancer killing, ADCC and CDC were performed by culturing PDL-1+/CD55+ cells with PBMCs, NK cells and human complement serum (HCS), respectively. GB262 internalization and its effects on intracellular PD-L1 were studied by pHrodo and western blot. Saporin-conjugated GB262 efficacy was evaluated by cell binding, internalization and cancer killing. In vivo study was performed by engrafting PBMCs and PANC-1 in B-NDG mice. Results: PD-L1 significantly co-localized with CD55 on cancer cells from different indications. While high PD-L1 binding was maintained, a significant reduction in CD55 binding was observed by GB262 to their receptors. GB262 also blocked the interaction between CD97 and CD55 more efficiently than monoclonals. GB262 induced stronger T-cell activation and cancer killing in presence of PBMCs. Similarly, higher level of ADCC and CDC were also evident by GB262. The degree of GB262 internalization was CD55 level dependent on cells surface and leads to intracellular PD-L1 degradation. When conjugated to saporin, GB262-saporin significantly induced cancer killing in absence of effector cells or HCS, whereas GB262 or saporin had no effect. Both GB262 and GB262-saporin suppressed tumor growth in B-NDG mice models. Conclusion: GB262 is the first bispecific antibody that not only releases cancer repression on T-cell activation, but also releases cancer repression on CDC. By conjugating to a toxin, GB262 could serve as an ADC to target multiple cancer cells. GB262 has potential to serve as a novel therapy for many cancers. Citation Format: Amit K. Chaudhary, Jiadong Shi, Wenyan Cai, Bo Wang, Mohd S. Zaman, Cai Huang, Jun Lin, Steven Z. Kan, Joe Zhou, Jianbo Dong, Yue Liu. Development of a novel bispecific antibody targeting PD-L1 and CD55 for cancer therapy [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr LB067.

Background: Tyrosine kinase inhibitors (TKIs) have been developed against EGFR. However, in NSCLC (Non-small Cell Lung Cancer) and some other cancers, the occurrence of activating mutations in EGFR leads to acquired resistance. Anomalous cMET signaling additionally contributes to developing resistance to targeted therapies, including ones directed at EGFR. A number of bispecific antibodies targeting EGFR and cMET are in clinical trials and JNJ-61186372 (Amivantamab) from Janssen labs, has been given fast track approval by the FDA. In this study we created bispecific and trispecific antibody formats which target both cMET and EGFR. Trispecific (TsAb), unlike the bispecific (BsAb), binds to two different epitopes on cMET and one on EGFR resulting in enhanced internalization of the receptors. This study demonstrates that our antibodies exhibit enhanced capabilities than the benchmark at clinical stage in cell and non-cell based in vitro assays. Methods: The two cMET binding arms of TsAb were designed to block the cMET-HGF (ligand) interaction and enhance the internalization of the receptors. Cell based assays were performed using a NSCLC cell line NCI-H1975 (ATCC CRL-5908) and HGF-expressing NCI-H1975. Receptor binding and receptor/ligand blocking (cMET/HGF and EGFR/EGF) assays were performed using both ELISA and flow cytometry. Phosphorylation assays were performed using ELISA and Western blotting. Receptor internalization studies were executed using the pHrodo labelling kit and were further assessed through fluorescence microscopy and quantified through flow cytometry. ADCC assays (Antibody dependent cellular cytotoxicity) were performed through the Crystal violet method. Results: BsAb was more effective than benchmark in terms of its binding capabilities to cMET and EGFR receptors and also in blocking the interaction of cMET and EGFR with their ligands. Moreover, functional assays revealed that the new BsAb is in most cases better than benchmark in successfully inhibiting the phosphorylation of not only the cMET and EGFR receptors but also the downstream signaling pathways which are involved in cancer cell proliferation, survival and obstructing the apoptotic pathways. Furthermore, the new TsAb was found to be more effective than BsAb and benchmark in cMET/EGFR receptor internalization which may lead to additional extenuation of the downstream pathways for these receptors. Cancer cell killing assays for ABT BsAb show ADCC comparable to the benchmark. Conclusion: Comparing to BsAbs, TsAb induces enhanced internalization of the cMET/EGFR receptors leading to further extenuation of the downstream oncogenic signaling pathways. These data suggest the development of ABT BsAb and specifically TsAb as a new drug target for patients with lung cancer and NSCLC in particular. Citation Format: Mohd Zaman, Cai Huang, Sachith Gallolu Kankanamalage, Amit K. Chaudhary, Jianbo Dong, Yue Liu. Development of cMET/cMET/EGFR Trispecific antibody as therapeutic modality for Non-small Cell Lung Cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr LB069.

Background: Tumors are heterogeneous in nature with highly suppressive microenvironment. The heterogeneity is distinguished by the presence of distinct cell populations, where many tumors are partial suppressed, thus lead to tumor relapse and develop drug-resistance after initial therapy. Most importantly, the heterogeneity in tumors raises considerable challenges to design effective treatment strategy with predictable outcomes. Paracellular permeability at tight junctions is largely determined by Claudins. Under aberrant condition, Claudins disrupt tight junction functions and promote tumor-initiation. Claudin 18.2 is significantly expressed in gastric and pancreatic cancers, while normal tissue expression is limited to the epithelium of the stomach, making it a promising target for cancer therapy. Although great efficacy has been observed for T-cell engagers, balance of safety-efficacy remains a major concern. Bispecific antibody (BsAb) with lower CD3-binding arm could have a huge potential to minimize unspecific T-cell activation while maintaining better-efficacy. Here, we designed Claudin 18.2/CD3 BsAb with weaker CD3-binding arm and maintain Fc-effector functions to broaden mechanisms of action including antibody-dependent cellular cytotoxicity (ADCC) and complement-dependent cytotoxicity (CDC). Methods: Claudin 18.2 expressing and T-cells were used for Claudin 18.2 and CD3 binding analysis, respectively. T-cell activation was performed by co-culturing T-cells with Claudin 18.2+ and Chaudin 18.2- cells. Pancreatic cancer cell was co-cultured with human PBMCs for ADCC, T-cell activation and T-cell cytotoxicity studies. Co-culture of Jurkat and CD16+ NK cells were used for Fc-mediated T-cell activation study. Claudin 18.2+ and T-cell cells were used for CDC studies in presence of human complement serum. Results: While comparable Claudin 18.2 binding affinity was maintained, a significant reduction in CD3 binding was observed by Claudin 18.2/CD3 BsAb compared to a benchmark antibody. Although weaker CD3 binding, BsAb induced equivalent Claudin 18.2 specific T-cell activation in presence of Jurkat and/or human PBMCs compared to the benchmark. Moreover, it also induced ADCC of pancreatic cancer cell and T-cell activation in presence of PBMCs in dose-dependent manner. Claudin 18.2/CD3 BsAb induced CDC of Claudin 18.2+ cells but not T-cell cells in presence of human complement serum. While Claudin 18.2/CD3 BsAb maintained the Fc-mediated ADCC and CDC functions, it had no significant effect on Fc-mediated T-cell activation compared to the benchmark antibody in presence of CD16+ NK cells. Conclusion: We designed Claudin 18.2/CD3 BsAb with weaker CD3 binding to minimize unspecific T-cell activation while maintaining Fc-mediated-effector functions. These data suggest the development of Claudin 18.2/CD3 BsAb as a drug target for patients with gastric and pancreatic cancers. Citation Format: Amit K. Chaudhary, Mohd S. Zaman, Cai Huang, Jianbo Dong, Yue Liu. Development of Claudin 18.2/CD3 bispecific antibody maintaining effector functions for better cancer therapy [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr LB070.

Background Bispecific T cell engaging antibodies (TEAs) with one arm targeting a cancer antigen and another arm binding to CD3 have demonstrated impressive efficacy in multiple clinical studies. However, establishing a balance of safety versus efficacy remains challenging. For instance, some TEAs have severe safety issues such as causing cytokine release symptom. Additionally, due to the heterogeneity of cancer cells, not all patients or all cancer cells of one patient respond equally to TEA treatments. It has been shown that cancer cells with PD-L1 overexpression respond to TEA therapy worse than other cancer cells. To develop next generation bispecific TEAs with a better balance of safety, efficacy and expanded mechanism of actions, we replaced heavy chain complementarity-determining regions (HCDRs) in one arm of Rituximab with HCDRs from a CD3 antibody to make a novel CD20/CD3 bispecific antibody, Rituximab/CD3 (GB261). After several rounds of computer aided sequence optimization, the final lead molecule demonstrated a superior balance of safety and efficacy in both in vitro and animal studies. Methods The affinities of antibodies to antigens were determined using FACS and SPR. T cell activation was determined by measuring CD69+/CD2+ cell percentage using FACS and cytokine release. Cancer cell killing assays were performed by measuring target cell depletion after incubating with PBMCs, ADCC assays were performed by measuring target cell depletion after incubating with NK92-CD16 cells and CDC was performed by measuring target cell depletion after incubating with complements-enriched human serum. In vivo cancer cell killing was performed using human PBMC engrafted NDG mouse models. Pre-tox and PK study were performed with cynomolgus monkey.Results GB261 exhibited high affinity to CD20 and an extremely low affinity to CD3. It showed similar T cell activation and cancer cell killing in Rituximab resistant Raji cells (RRCL) while reducing cytokine secretion compared to a benchmark CD20/CD3 (BM). GB261-induced ADCC and CDC only killed CD20+ cells but not CD3+ cells. When RRCL and PBMC cells (1:1) were co-injected with antibodies, it exhibited cancer killing effect comparable to the BM in a PBMC engrafted mouse model. However, if RRCL cells were injected first, then PBMC cells and antibodies were co-injected 4 days later, it showed superior cancer cell killing. Moreover, it delayed cancer relapse more effectively than Rutuximab in mice. Pre-tox and PK data in a cyno model also suggested that GB261 has good safety and PK. Finally, GB261 also demonstrated excellent manufacturability based on expression level and bio-analytic data.Conclusions GB261 exhibits enhanced efficacy and safety, and great manufacturability, presenting several advantages over the BM antibody. Thus, GB261 is a promising bispecific therapeutic antibody against CD20+ cancers. Citation Format: Wenyan Cai, Jianbo Dong, Sachith Gallolu Kankanamalage, Allison Titong, Jiadong Shi, Zhejun Jia, Bo Wang, Cai Huang, Jing Zhang, Jun Lin, Steven Z. Kan, Joe Zhou, Yue Liu. A novel CD20/CD3 T cell engager designed from Rituximab targeting CD20+ cancer with multiple mechanisms [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr LB068.

Despite mounting evidence for SARS-CoV-2 engagement with immune cells, most express little, if any, of the canonical receptor of SARS-CoV-2, ACE2. Here, using a myeloid-cell receptor-focused ectopic expression screen, we identified several C-type lectins (DC-SIGN, L-SIGN, LSECtin, ASGR1, and CLEC10A) and Tweety family member 2 (TTYH2) as glycan-dependent binding partners of the SARS-CoV-2 spike. Except for TTYH2, these molecules primarily interacted with spike via regions outside of the receptor-binding domain. Single-cell RNA-sequencing analysis of pulmonary cells from COVID-19 patients indicated predominant expression of these molecules on myeloid cells. Although these receptors do not support active replication of SARS-CoV-2, their engagement with virus induced robust proinflammatory responses in myeloid cells that correlated with COVID-19 severity. We also generated a bispecific anti-spike nanobody that not only blocked ACE2-mediated infection but also the myeloid receptors-mediated proinflammatory responses. Our findings suggest SARS-CoV-2-myeloid receptor interactions promote immune hyper-activation, which represents potential targets for COVID-19 therapy., Graphical Abstract, Most immune cells express little, if any, of the canonical SARS-CoV-2 receptor, ACE2. Lu et al. report that C-type lectins and TTYH2 act as SARS-CoV-2 myeloid cell-interacting partners that trigger immune hyper-activation, but not infection. These findings raise the possibility that these virus-myeloid cell interactions are directly involved in COVID-19 immunopathogenesis and could be targeted for COVID-19 therapy.

Smooth cordgrass Spartina alterniflora has rapidly spread in salt marshes of Yangtze River estuary. However, there is no consistent pattern on the effects of S. alterniflora invasions on microhabitats and benthic communities. The study first compared vegetation, water velocity, sediment deposition rate, sediment properties and benthic communities between invasive S. alterniflora and native Scirpus mariqueter patches using a pair-wise sampling design at 3 sites (Dongtan, Jiuduansha and Nanhui) in four seasons. The stands of S. alterniflora exhibited taller and less dense shoots, greater aboveground biomass and longer phenological phase than did S. mariqueter . S. alterniflora marshes had significantly lower water velocity, greater sediment deposition rates and higher light attenuation relative to S. mariqueter habitats. In general, sediment temperature, salinity, porosity, organic carbon content and microalgal biomass (Chla) did not significantly differ between invaded and uninvaded habitats at any of 3 sites. Thirty-four of 37 macrofauna species recorded did not exhibit density differences between habitat types. Limited influences on macrofauna were evident only at Dongtan where S. alterniflora marshes had significantly greater gastropod densities in the autumn. In 8 of 12 observations, benthic community structure did not differ between invaded and uninvaded habitats. It was concluded that the differences in physical structure were not enough to significantly change sediment properties, macrofaunal density and community structure.

The limited write endurance is one of the major obstacles for phase-change random access memory (PRAM)-based main memory. Traditionally, wear-leveling (WL) techniques were proposed to enhance its lifetime by balancing write traffic. However, these techniques do not concern the endurance variation in PRAM chips. When different PRAM cells have distinct endurance, balanced writes results in lifetime degradation due to the weakest cells. In this paper, we first define a new metric-wear rate (i.e., writes/endurance) considering both the write traffic and endurance distribution from application and hardware, respectively. After investigating the writing behavior of applications and endurance variation, we propose an architecture-level leveling mechanism to balance wear rate of cells across the PRAM chip. Hardware and algorithm to support the proposed leveling mechanism are presented. Moreover, there is an important tradeoff between endurance improvement and swapping data volume. To co-optimize endurance and swapping, this situation is formulated as a maximum weight perfect matching problem in bipartite graph. Thereafter, a novel algorithm that minimizes wear-rate and swapping by employing Kuhn–Munkras algorithm is proposed to maximize PRAM lifetime and minimize performance degradation. The experimental results show $\sim 17 \times $ lifetime improvement over prior WL.

Efficient resource utilization requires that emerging datacenter interconnects support both high performance communication and efficient remote resource sharing. These goals require that the network be more tightly coupled with the CPU chips. Designing a new interconnection technology thus requires considering not only the interconnection itself, but also the design of the processors that will rely on it. In this paper, we study memory hierarchy implications for the design of high-speed datacenter interconnects — particularly as they affect remote memory access — and we use PCIe as the vehicle for our investigations. To that end, we build three complementary platforms: a PCIe-interconnected prototype server with which we measure and analyze current bottlenecks; a software simulator that lets us model microarchitectural and cache hierarchy changes; and an FPGA prototype system with a streamlined switchless customized protocol Thunder with which we study hardware optimizations outside the processor. We highlight several architectural modifications to better support remote memory access and communication, and quantify their impact and limitations.

Data centers require efficient, low-cost, flexible interconnects to manage the rapidly growing internal traffic generated by an increasingly diverse set of applications. To meet these requirements, data center networks are increasingly employing alternatives such as RapidIO, Freedom, and PCIe, which require fewer physical devices and/or have simpler protocols than more traditional interconnects. These networks offer raw high performance communication capabilities, but simply using them for conventional TCP/IP-based communication fails to realize the potential performance of the physical network. Here we analyze causes for this performance loss for the TCP/IP protocol over one such fabric, PCIe, and we explore a hardware/software solution that mitigates overheads and exploits PCIe's advanced features. The result is P-Socket, an efficient library that enables legacy socket applications to run without modification. Our experiments show that P-Socket achieves an end-to-end latency of 1.2μs and effective bandwidth of up to 2.87GB/s (out of a theoretical peak of 3.05GB/s).

Consolidated server racks are quickly becoming the backbone of IT infrastructure for science, engineering, and business, alike. These servers are still largely built and organized as when they were distributed, individual entities. Given that many fields increasingly rely on analytics of huge datasets, it makes sense to support flexible resource utilization across servers to improve cost-effectiveness and performance. We introduce Venice, a family of data-center server architectures that builds a strong communication substrate as a first-class resource for server chips. Venice provides a diverse set of resource-joining mechanisms that enables user programs to efficiently leverage non-local resources. To better understand the implications of design decisions about system support for resource sharing we have constructed a hardware prototype that allows us to more accurately measure end-to-end performance of at-scale applications and to explore tradeoffs among performance, power, and resource-sharing transparency. We present results from our initial studies analyzing these tradeoffs when sharing memory, accelerators, or NICs. We find that it is particularly important to reduce or hide latency, that data-sharing access patterns should match the features of the communication channels employed, and that inter-channel collaboration can be exploited for better performance.

Thread-level redundancy is an efficient approach for transient fault detection and recovery in Chip Multiprocessors (CMPs), in which two adjacent cores are statically coupled to form a functional Dual Modular Redundancy (DMR). Manufacturing process variations cause core-to-core (C2C) performance asymmetry across the chip, which can be further divided into the asymmetry among core-pairs and the asymmetry within a core-pair. We call them inter- and intra-pair asymmetries, respectively, both of which should be taken into considerations in application scheduling for CMPs with static core coupling. In this paper, we first formulate the above scheduling problem as a 0-1 programming problem to maximize the system Weighted Throughput. An efficient IVF&AppSen algorithm is then proposed, which we prove to be optimal when the number of applications equals to that of core-pairs. We also adapt the Simulated Annealing technique to tackle this problem when applications are less than core-pairs on chip. Simulations on a 64-core CMP shows that the proposed algorithms achieve 2.5-9.3% improvement in Weighted Throughput when compared to prior VarF&AppIPC algorithm.

The standard of treatment for botulism, equine antitoxin, is a foreign protein with associated safety issues and a short serum half-life which excludes its use as a prophylactic antitoxin and makes it a less-than-optimal therapeutic. Due to these limitations, a recombinant monoclonal antibody (mAb) product is preferable. It has been shown that combining three mAbs that bind non-overlapping epitopes leads to highly potent botulinum neurotoxin (BoNT) neutralization. Recently, a triple human antibody combination for BoNT/A has demonstrated potent toxin neutralization in mouse models with no serious adverse events when tested in a Phase I clinical trial. However, a triple antibody therapeutic poses unique development and manufacturing challenges. Thus, potentially to streamline development of BoNT antitoxins, we sought to achieve the potency of multiple mAb combinations in a single IgG-based molecule that has a long serum half-life. The design, production, and testing of a single tri-epitopic IgG1-based mAb (TeAb) containing the binding sites of each of the three parental BoNT/A mAbs yielded an antibody of nearly equal potency to the combination. The approach taken here could be applied to the design and creation of other multivalent antibodies that could be used for a variety of applications, including toxin elimination.

Existing antibodies (Abs) used to treat botulism cannot enter the cytosol of neurons and bind to botulinum neurotoxin (BoNT) at its site of action, and thus cannot reverse paralysis. However, Abs targeting the proteolytic domain of the toxin could inhibit the proteolytic activity of the toxin intracellularly and potentially reverse intoxication, if they could be delivered intracellularly. As such, antibodies that neutralize toxin activity could serve as potent inhibitory cargos for therapeutic antitoxins against botulism. BoNT serotype B (BoNT/B) contains a zinc endopeptidase light chain (LC) domain that cleaves synaoptobrevin-2, a SNARE protein responsible for vesicle fusion and acetylcholine vesicle release. To generate monoclonal Abs (mAbs) that could reverse paralysis, we targeted the protease domain for Ab generation. Single-chain variable fragment (scFv) libraries from immunized mice or humans were displayed on yeast, and 19 unique BoNT/B LC-specific mAbs isolated by fluorescence-activated cell sorting (FACS). The equilibrium dissociation constants (KD) of these mAbs for BoNT/B LC ranged from 0.24 nM to 14.3 nM (mean KD 3.27 nM). Eleven mAbs inhibited BoNT/B LC proteolytic activity. The fine epitopes of selected mAbs were identified by alanine-scanning mutagenesis, revealing that inhibitory mAbs bound near the active site, substrate-binding site or the extended substrate-binding site. The results provide mAbs that could prove useful for intracellular reversal of paralysis and identify epitopes that could be targeted by small molecules inhibitors.

Rab7 is a key regulatory protein governing early to late endocytic membrane transport. In this study the proteasome alpha-subunit XAPC7 (also known as PSMA7, RC6-1, and HSPC in mammals) was identified to interact specifically with Rab7 and was recruited to multivesicular late endosomes through this interaction. The protein interaction domains were localized to the C terminus of XAPC7 and the N terminus of Rab7. XAPC7 was not found on early or recycling endosomes, but could be recruited to recycling endosomes by expression of a Rab7-(1-174)Rab11-(160-202) chimera, establishing a central role for Rab7 in the membrane recruitment of XAPC7. Although XAPC7 could be shown to associate with membranes bearing ubiquitinated cargo, overexpression had no impact on steady-state ubiquitinated protein levels. Most notably, overexpression of XAPC7 was found to impair late endocytic transport of two different membrane proteins, including EGFR known to be highly dependent on ubiquitination and proteasome activity for proper endocytic sorting and lysosomal transport. Decreased late endocytic transport caused by XAPC7 overexpression was partially rescued by coexpression of wild-type Rab7, suggesting a negative regulatory role for XAPC7. Nevertheless, Rab7 itself was not subject to XAPC7-dependent proteasomal degradation. Together the data establish the first direct molecular link between the endocytic trafficking and cytosolic degradative machineries.

The exploding growth of digitalized information has led to the rapid growth of data centers, both in numbers and in size. Cluster has been the dominating system architecture used in most data centers. However, the increasingly diversified data center applications have requirements beyond what the cluster architecture can deliver. For instance, clouding computing requires flexible sharing of all data center resources. Big data applications often need large memory capacity. A few applications can use GPGPU effectively. Existing system might be extended to a certain degree to meet those needs. Those extensions however would often be prohibitively expensive. The paper presents our attempt to design a system using commodity products that can meet the varying needs of many emerging data center applications in a cost-effective way. Our attempt is to create a system by connecting multiple nodes through a PCIe switch and then extend the software stack to support resource sharing among these nodes. In particular, a node can directly use the memory, NIC, and GPGPU of other nodes through the PCIe switch with no or little involvement from other nodes. We build a prototype as our evaluation platform. Our evaluation results indicate that those resources can be shared effectively in many cases. For using remote memory as block device, our prototype system has 5 times bandwidth, 11 times IOPS and 1/12 latency compared with the system connected by 10GigE in average for Orion benchmark; Using remote GPGPU via PCIe switch achieves average 60 times speedup than the case without GPGPU, and the performance loss is also acceptable (its average execution time is 1/3 of local GPGPU) for micro-benchmarks from GPU computing SDK; And using remote NIC via PCIe switch achieves average 95% bandwidth and 1.4 times latency of local NIC in httperf testing. While our prototype system offers multiple benefits, it is not perfect and has a lot room for further optimization and extension. We hope the outcome presented in this paper will encourage more researchers to join us in designing highly efficient and cost-effective servers.

The limited write endurance of phase change random access memory (PRAM) is one of the major obstacles for PRAM-based main memory. Wear leveling techniques were proposed to extend its lifetime by balancing writes traffic. Another important concern that need to be considered is endurance variation in PRAM chips. When different PRAM cells have distinct endurance, balanced writes will result in lifetime degradation due to the weakest cells. Instead of balancing writes traffic, in this paper we propose wear rate leveling (WRL), a variant of wear leveling, to balance wear rates (i.e., writes traffic/edudrance) of cells across the PRAM chip. After investigating writing behavior of applications and endurance variation, we propose an architecture-level WRL mechanism. Moreover, there is an important tradeoff between endurance improvement and swapping data volume. To co-optimize endurance and swapping, a novel algorithm, Max Hyper-weight Rematching, is proposed to maximize PRAM lifetime and minimize performance degradation. Experimental results show 19x endurance improvement to prior Wear Leveling.

Thread-Level Redundancy in Chip Multiprocessors(TLR-CMP) is efficient for soft error tolerance. Process variation causes core-to-core (C2C) performance asymmetry across a chip, which should be taken into consideration for application scheduling. In this paper, two types of variations beyond C2C are introduced, i.e., inter-pair and intra-pair variation in TLR-CMP. Intra-pair performance asymmetry can affect the performance of applications differently. Based on the above observation, we firstly formalize the variation aware scheduling in TLR-CMP as a 0-1 programming problem,to maximize the system weighted throughput. An efficient scheduling algorithm, named IntraVarF&AppSen, is then proposed to tackle this problem, which can be proved to be optimal when the number of applications to be scheduled is equal to the number of core pairs. Simulation on a 64-core CMP shows 2.8%-4% improvement in weighted throughput when compared to prior VarF&AppIPC algorithm.

Proteasomes have long been known to mediate the degradation of polyubiquitinated proteins in the cytoplasm and the nucleus. Additionally, proteasomes have been identified as participating in cellular degradative pathways involving the endomembrane system. In conjunction with the endoplasmic reticulum, proteasomes serve as a quality control mechanism for disposing of malfolded newly synthesized proteins, while on the endocytic pathway they serve to facilitate the degradation of key signaling and nutrient receptors as well as the destruction of phagocytosed pathogens. Our laboratory has identified a direct interaction between the late endocytic Rab7 GTPase and the alpha-proteasome subunit, XAPC7, thus providing the first molecular link between the endocytic trafficking and cytosolic degradative machineries. In this chapter reagents and methods for studying the regulation and interactions between XAPC7, the 20S proteasome, and Rab7 are described.

MthK is a Ca2+-gated K+ channel from Methanobacterium autotrophicum. The crystal structure of the MthK channel in a Ca2+-bound open state was previously determined at 3.3 A and revealed an octameric gating ring composed of eight intracellular ligand-binding RCK (regulate the conductance of K+) domains. It was suggested that Ca2+ binding regulates the gating ring conformation, which in turn leads to the opening and closing of the channel. However, at 3.3 AA resolution, the molecular details of the structure are not well defined, and many of the conclusions drawn from that structure were hypothetical. Here we have presented high resolution structures of the MthK RCK domain with and without Ca2+ bound from three different crystals. These structures revealed a dimeric architecture of the RCK domain and allowed us to visualize the Ca2+ binding and protein-protein contacts at atomic detail. The dimerization of RCK domains is also conserved in other RCK-regulated K+ channels and transporters, suggesting that the RCK dimer serves as a basic unit in the gating ring assembly. A comparison of these dimer structures confirmed that the dimer interface is indeed flexible as suggested previously. However, the conformational change at the flexible interface is of an extent smaller than the previously hypothesized gating ring movement, and a reconstruction of these dimers into octamers by applying protein-protein contacts at the fixed interface did not generate enclosed gating rings. This indicated that there is a high probability that the previously defined fixed interface may not be fixed during channel gating. In addition to the structural studies, we have also carried out biochemical analyses and have shown that near physiological pH, isolated RCK domains form a stable octamer in solution, supporting the notion that the formation of octameric gating ring is a functionally relevant event in MthK gating. Additionally, our stability studies indicated that Ca2+ binding stabilizes the RCK domains in this octameric state.

Proteasomes have long been known to mediate the degradation of polyubiquitinated proteins in the cytoplasm and the nucleus. Additionally, proteasomes have been identified as participating in cellular degradative pathways involving the endomembrane system. In conjunction with the endoplasmic reticulum, proteasomes serve as a quality control mechanism for disposing of malfolded newly synthesized proteins, while on the endocytic pathway they serve to facilitate the degradation of key signaling and nutrient receptors as well as the destruction of phagocytosed pathogens. Our laboratory has identified a direct interaction between the late endocytic Rab7 GTPase and the alpha-proteasome subunit, XAPC7, thus providing the first molecular link between the endocytic trafficking and cytosolic degradative machineries. In this chapter reagents and methods for studying the regulation and interactions between XAPC7, the 20S proteasome, and Rab7 are described.

Rab GTPases serve as master regulators of vesicular membrane transport on both the exo- and endocytic pathways. In their active forms, rab proteins serve in cargo selection and as scaffolds for the sequential assembly of effectors requisite for vesicle budding, cytoskeletal transport, and target membrane fusion. Rab protein function is in turn tightly regulated at the level of protein expression, localization, membrane association, and activation. Alterations in the rab GTPases and associated regulatory proteins or effectors have increasingly been implicated in causing human disease. Some diseases such as those resulting in bleeding and pigmentation disorders (Griscelli syndrome), mental retardation, neuropathy (Charcot-Marie-Tooth), kidney disease (tuberous sclerosis), and blindness (choroideremia) arise from direct loss of function mutations of rab GTPases or associated regulatory molecules. In contrast, in a number of cancers (prostate, liver, breast) as well as vascular, lung, and thyroid diseases, the overexpression of select rab GTPases have been tightly correlated with disease pathogenesis. Unique therapeutic opportunities lie ahead in developing strategies that target rab proteins and modulate the endocytic pathway.