The change from the temperature independence of the primary (1°) H/D kinetic isotope effects (KIEs) in wild-type enzyme-catalyzed H-transfer reactions (ΔE a = E aD – E aH ∼ 0) to a strong temperature dependence with the mutated enzymes (ΔE a ≫ 0) has recently been frequently observed. This has prompted some enzymologists to develop new H-tunneling models to correlate ΔE a with the donor–acceptor distance (DAD) at the tunneling-ready state (TRS) as well as the protein thermal motions/dynamics that sample the short DADTRS’s for H-tunneling to occur. While extensive evidence supporting or disproving the thermally activated DAD sampling concept has emerged, a comparable study of the simpler bimolecular H-tunneling reactions in solution has not been carried out. In particular, small ΔE a’s (∼0) have not been found. In this paper, we report a study of the hydride-transfer reactions from four NADH models to the same hydride acceptor in acetonitrile. The ΔE a’s were determined: 0.37 (small), 0.60, 0.99, and 1.53 kcal/mol (large). The α-secondary (2°) KIEs on the acceptor that serve as a ruler for the rigidity of reaction centers were previously reported or determined. All possible productive reactant complex (PRC) configurations were computed to provide insight into the structures of the TRS’s. Relationships among structures, 2° KIEs, DADPRC’s, and ΔE a’s were discussed. The more rigid system with more suppressed 2° C–H vibrations at the TRS and more narrowly distributed DADPRC’s in PRCs gave a smaller ΔE a. The results replicated the trend observed in enzymes versus mutated enzymes and appeared to support the concepts of different thermally activated DADTRS sampling processes in response to the rigid versus flexible donor–acceptor centers.

The α-H/D (if available) and remote β-type N-CH 3 /CD 3 2°kinetic isotope effects (KIEs) on 10-methylacridine (MAH), 9,10-dimethylacridine (DMAH), 1,3-dimethyl-2-phenylbenzimidazoline (DMPBIH) and on the oxidized forms MA + and DMA + , in their hydride transfer reactions with several hydride acceptors/donors in acetonitrile, were determined. The corresponding equilibrium isotope effects (EIEs) were computed. Hammett correlations of several closely related hydride transfer reactions were constructed using the literature data. The α-2°KIEs on both MAH and MA + are inflated relative to the semiclassical prediction on the basis of the KIE/EIE comparison and Hammond's postulate. This together with previously published unusual 1°and 2°K IE behaviors strongly suggest a H-tunneling mechanism. By comparing with the EIEs, the α-2°KIEs were used to analyze the rehybridization of the reaction center C and the N-CH 3 /CD 3 2°KIEs to calculate the charge distribution on the structure containing N during H-tunneling. The rehybridization appears to lag behind the charge development in the donor moiety. The charge distribution at the tunneling ready transition state is in agreement with the Hammett correlations; the donor is productlike, and the acceptor is reactant-like, indicative of a partial negative charge borne by the "in-flight" nucleus being "hydridic" in character. Results were compared with the α-2°KIEs on NADH/NAD + and the Hammett correlations in closely related enzymes. The comparison implicates that the H-tunneling probability would be enhanced by these enzymes.

As convolution has empowered many smart applications, dynamic convolution further equips it with the ability to adapt to diverse inputs. However, the static and dynamic convolutions are either layout-agnostic or computationheavy, making it inappropriate for layout-specific applications, e.g., face recognition and medical image segmentation. We observe that these applications naturally exhibit the characteristics of large intra-image (spatial) variance and small cross-image variance. This observation motivates our efficient translation variant convolution (TVConv) for layout-aware visual processing. Technically, TVConv is composed of affinity maps and a weight-generating block. While affinity maps depict pixel-paired relationships gracefully, the weight-generating block can be explicitly overparameterized for better training while maintaining efficient inference. Although conceptually simple, TVConv significantly improves the efficiency of the convolution and can be readily plugged into various network architectures.Extensive experiments on face recognition show that TV-Conv reduces the computational cost by up to 3.1× and improves the corresponding throughput by 2.3× while maintaining a high accuracy compared to the depthwise convolution. Moreover, for the same computation cost, we boost the mean accuracy by up to 4.21%. We also conduct experiments on the optic disc/cup segmentation task and obtain better generalization performance, which helps mitigate the critical data scarcity issue.

No abstract