Search

Computing-In-memory (CIM) accelerators have the characteristics of storage and computing integration, which has the potential to break through the limit of Moore’s law and the bottleneck of Von-Neumann architecture for convolutional neural networks (CNN) implementation improvement. However, the performance of CIM accelerators is still limited by conventional CNN architectures and inefficient readouts. To increase energy-efficient performance, an optimized CNN model is required and a low-power column parallel readout is necessary for edge-computing hardware. In this work, an ReRAM-based CNN accelerator is designed. Mixed-bit operations from 1 bit to 8 bits are supported by an effective bitwidth configuration scheme to implement Neural Architecture Search (NAS)-optimized layer-wise multi-bit CNNs. Besides, column-parallel readout is achieved with excellent energy-efficient performance by a variation-reduction accumulation mechanism and low-power readout circuits. Additionally, we further explore systolic data reuse in an ReRAM-based PE array. Experiments are implemented on NAS-optimized ResNet-18. Benchmarks show that the proposed ReRAM accelerator can achieve peak energy efficiency of 2490.32 TOPS/W for 1-bit operation and average energy efficiency of 479.37 TOPS/W for $1\sim 8$ -bit operations with evaluating NAS-optimized multi-bitwidth CNNs. When compared with the state-of-the-art works, the proposed accelerator shows at least $14.18{\times }$ improvement on energy efficiency.

The flaviviruses dengue virus (DENV) and Zika virus (ZIKV) are severe health threats with rapidly expanding ranges. To identify the host cell dependencies of DENV and ZIKV, we completed orthologous functional genomic screens using RNAi and CRISPR/Cas9 approaches. The screens recovered the ZIKV entry factor AXL as well as multiple host factors involved in endocytosis (RAB5C and RABGEF), heparin sulfation (NDST1 and EXT1), and transmembrane protein processing and maturation, including the endoplasmic reticulum membrane complex (EMC). We find that both flaviviruses require the EMC for their early stages of infection. Together, these studies generate a high-confidence, systems-wide view of human-flavivirus interactions and provide insights into the role of the EMC in flavivirus replication.

Zika virus has emerged as a severe health threat with a rapidly expanding range. The IFITM family of restriction factors inhibits the replication of a broad range of viruses, including the closely related flaviruses West Nile virus and dengue virus. Here, we show that IFITM1 and IFITM3 inhibit Zika virus infection early in the viral life cycle. Moreover, IFITM3 can prevent Zika-virus-induced cell death. These results suggest that strategies to boost the actions and/or levels of the IFITMs might be useful for inhibiting a broad range of emerging viruses.

No relevant conflicts of interest to declare.

The myeloproliferative neoplasms (MPNs) are hematologic malignancies with a chronic clinical course and a risk of developing thrombosis and acute leukemia. JAK2 is a member of the Janus family of non-receptor tyrosine kinases, and the somatic activating point mutation of JAK2(JAK2V617F) has been found in Philadelphia chromosome-negative MPNs in approximately 95% of polycythemia (PV), 60% of essential thrombocytosis (ET), and 50% of JAK2 V617F-positive MPD (PMF) patients. Although several JAK2 inhibitors are in the early stage of clinical trials and have been shown to help to improve symptoms and quality of life in patients, their long-term effectiveness in patients remains to be determined. New therapeutic strategies need to be developed for treating PV more effectively. Because hematopoietic stem cells (HSCs) harbor JAK2V617Fin PV patients, PV should be compared to Philadelphia chromosome-positive chronic myeloid leukemia (CML) that is also derived from HSCs and has a myeloproliferative phenotype similar to PV in mice. Therefore, we reasoned that a gene essential for CML development might also play a critical role in PV development. We have shown that the survival and self-renewal of CML-initiating cells require the arachidonate 5-lipoxygenase gene (Alox5) and that Alox5is essential for CML development. Therefore, in this study we investigated the role of Alox5in pathogenesis of PV in mice. We showed that JAK2V617Fupregulates Alox5 expression both in vitro(Ba/F3 cells) and in vivo(the spleen cells in PV mice). To examine whether Alox5 is required for induction of PV by JAK2V617F, we transduced bone marrow (BM) cells from wild type (WT) or Alox5homozygous knockout (Alox5-/-) mice with JAK2V617F-GFP, followed by transplantation into lethally-irradiated recipient mice. We found that the average percentage of JAK2V617F-expressing white blood cell counts (GFP+WBCs) in recipients of JAK2V617F-transduced Alox5-/-donor BM cells was significantly lower than that in recipients of JAK2V617F-transduced WT donor BM cells (P< 0.001). Total numbers of WBCs, red blood cells (RBCs), hemoglobin (HGB) and hematocrit (HCT) were also significantly lower in the absence of Alox5. With time (at 9 months after induction of PV), the effect of Alox5deficiency on PV development became more significant. Because JAK2V617F-transformed HSCs is shown to be a PV-initiating cell population, we further tested whether loss of Alox5affects this cell population in PV mice. We found that the percentage and total number of HSCs (Lin-Sca-1+c-Kit+) in mice receiving JAK2V617F-transduced Alox5-/-BM cells was significantly lower than that in mice receiving JAK2V617F-transduced WT BM cells. These findings suggest that Alox5 could be an effective target gene for PV. We tested this idea by treating PV mice with an Alox5 inhibitor zileuton (300 mg/kg, once a day) or a placebo. Western blot analysis of protein lysates from the spleens of PV mice showed that 5-lipoxygines (5-LO, a protein product of the Alox5gene) is upregulated by JAK2V617F, and inhibition of Alox5 function by zileuton restored the 5-LO level back to the control level. In addition, the number of WBCs in placebo-treated mice rose significantly with time, whereas the number of WBCs in zileuton-treated mice did not. The inhibition of PV development by zileuton correlated with much smaller infiltrated spleen in zileuton-treated PV mice than in placebo-treated PV mice. Furthermore, we compared the levels of JAK2V617F-expressing cells (GFP+) in PV mice. FACS analysis showed that the percentage of Gr-1+or Mac+PV cells in peripheral blood was significantly lower in zileuton-treated PV mice than in placebo-treated PV mice. Consistent with the inhibition of PV cells by zileuton in mice, survival of zileuton-treated PV mice was significantly improved compared to placebo-treated PV mice (Fig. 3I). At 5 months after induction of PV, about 40% placebo-treated PV mice died, but all zileuton-treated mice survived. To understand the underlying mechanisms, we treated JAK2V617F-expressing BaF/3 cells with zileuton, and found that zileuton inhibited activation of AKT and β-catenin by JAK2V617F, and did not affect activation of PI3K, MAPK and JAK2. Together, these results demonstrate that Alox5 represents a critical pathway in JAK2V617F-induced PV and that targeting of Alox5 provides a new strategy for treating PV.

No relevant conflicts of interest to declare.

Abstract 1666