Supported Au nanoclusters are well-known for their unusual properties in catalysis. We describe here that nanostructured porous Au made via dealloying represents a new class of unsupported catalysts with extraordinary activities in important reactions such as CO oxidation. Although nanoporous Au may contain some oxides on the surface, our results demonstrate that it is metallic Au that plays the main role in this catalytic reaction. Furthermore, this material has good low-temperature catalytic stability and is extremely CO tolerant.

Dealloying single phase alloys is known to generate a type of nanostructured porous metals with intriguing properties. In this study, nanoporous gold (NPG) made by dealloying Au-Ag is investigated as a novel electrode material for methanol electro-oxidation. Compared to bulk Au electrode, oxidation and subsequent reduction of NPG occur at significantly negative potentials in both acid and alkaline solutions. NPG shows great catalytic activity for methanol electro-oxidation, but the structure quickly coarsens upon long time potential cycling. Interestingly, after surface modification with only a tiny amount of platinum, NPG exhibits greatly enhanced electrocatalytic activity toward methanol oxidation in the alkaline solutions, which is exemplified by a broad and high anodic peak during the positive scan and two secondary oxidation peaks in the subsequent reverse scan. At the same time, SEM observation and long-time potential cycling both prove that Pt-NPG has much enhanced structure stability as compared with bare NPG.

INTRODUCTORY Re-expression of the paralogous γ-globin genes ( HBG1/2 ) could be a universal strategy to ameliorate the severe β-globin disorders sickle cell disease (SCD) and β-thalassemia by induction of fetal hemoglobin (HbF, α 2 γ 2 ) 1 . Previously we and others have shown that core sequences at the BCL11A erythroid enhancer are required for repression of HbF in adult-stage erythroid cells but dispensable in non-erythroid cells 2 – 6 . CRISPR-Cas9 mediated gene modification has demonstrated variable efficiency, specificity, and persistence in hematopoietic stem cells (HSCs). Here we demonstrate that Cas9:sgRNA ribonucleoprotein (RNP) mediated cleavage within a GATA1 binding site at the +58 BCL11A erythroid enhancer results in highly penetrant disruption of this motif, reduction of BCL11A expression, and induction of fetal γ-globin. We optimize conditions for selection-free on-target editing in patient-derived HSCs as a nearly complete reaction lacking detectable genotoxicity or deleterious impact on stem cell function. HSCs preferentially undergo nonhomologous as compared to microhomology mediated end-joining repair. Erythroid progeny of edited engrafting sickle cell disease (SCD) HSCs express therapeutic levels of fetal hemoglobin (HbF) and resist sickling, while those from β-thalassemia patients show restored globin chain balance. NHEJ-based BCL11A enhancer editing approaching complete allelic disruption in HSCs is a practicable therapeutic strategy to produce durable HbF induction.

We present a self-supervised learning approach for optical flow. Our method distills reliable flow estimations from non-occluded pixels, and uses these predictions as ground truth to learn optical flow for hallucinated occlusions. We further design a simple CNN to utilize temporal information from multiple frames for better flow estimation. These two principles lead to an approach that yields the best performance for unsupervised optical flow learning on the challenging benchmarks including MPI Sintel, KITTI 2012 and 2015. More notably, our self-supervised pre-trained model provides an excellent initialization for supervised fine-tuning. Our fine-tuned models achieve stateof-the-art results on all three datasets. At the time of writing, we achieve EPE=4.26 on the Sintel benchmark, outperforming all submitted methods.

We present DDFlow, a data distillation approach to learning optical flow estimation from unlabeled data. The approach distills reliable predictions from a teacher network, and uses these predictions as annotations to guide a student network to learn optical flow. Unlike existing work relying on handcrafted energy terms to handle occlusion, our approach is data-driven, and learns optical flow for occluded pixels. This enables us to train our model with a much simpler loss function, and achieve a much higher accuracy. We conduct a rigorous evaluation on the challenging Flying Chairs, MPI Sintel, KITTI 2012 and 2015 benchmarks, and show that our approach significantly outperforms all existing unsupervised learning methods, while running at real time. * Work mainly done during an internship at Tencent AI Lab.