Engineering topics
Wang, Yixin
Publications and source records attributed to Wang, Yixin.
Clarifying trust of materials property predictions using neural networks with distribution-specific uncertainty quantification
It is critical that machine learning (ML) model predictions be trustworthy for high-throughput catalyst discovery approaches. Uncertainty quantification (UQ) methods allow estimation of the trustworthiness of an ML model, but these methods have not been well explored in the field of heterogeneous catalysis. Herein, we investigate different UQ methods applied to a crystal graph convolutional neural network to predict adsorption energies of molecules on alloys from the Open Catalyst 2020 dataset, the largest existing heterogeneous catalyst dataset. We apply three UQ methods to the adsorption energy predictions, namely k-fold ensembling, Monte Carlo dropout, and evidential regression. The effectiveness of each UQ method is assessed based on accuracy, sharpness, dispersion, calibration, and tightness. Evidential regression is demonstrated to be a powerful approach for rapidly obtaining tunable, competitively trustworthy UQ estimates for heterogeneous catalysis applications when using neural networks. Recalibration of model uncertainties is shown to be essential in practical screening applications of catalysts using uncertainties.
Construction of chicken infectious anemia virus infectious clone and study on its pathogenicity
Chicken infectious anemia virus (CIAV) can be transmitted through contaminated live poultry vaccine. However, the pathogenicity of contaminated CIAV strains is rarely reported. Previously, the chickens showed the typical symptoms of anemia after using the attenuated live fowl pox virus (FPV) vaccine. Therefore, exogenous CIAV contamination was suspected. We detected anti-CIAV antibodies in SPF chicks vaccinated with the FPV vaccine. CIAV contamination was confirmed in the FPV vaccine, and the CIAV strain was named JS2020-FPV. This study aims to rescue JS2020-FPV by reverse genetic assays and investigate its pathogenicity. Firstly, double-copies infectious clone of JS2020-FPV was constructed. For the pathogenicity study, infectious clone of JS2020-FPV was used to inoculate 1-day-old SPF chicks. The typical symptoms of anemia were observed in the JS2020-PFV group 14 days post inoculation. The hematocrit and body weight of chicks in the JS2020-PFV group were significantly lower than those in the mock group. Notably, the thymus development index and antibody levels of NDV were lower in chicks in the JS2020-PFV group than those in the mock group. Different degrees of apoptosis of MSB1 and DF-1 were observed after inoculated with the JS2020-FPV VP3 recombinant fusion protein expressed by E. coli system, indicating that VP3 induced apoptosis in the transformed cells. Overall, the pathogenicity of the CIAV detected in the contaminated vaccine was confirmed by inoculating SPF chicks with the double-copies infectious DNA clone in this study. Our findings indicate that the dangers of vaccine contamination cannot be ignored.