scPerb: Predict single-cell perturbation via style transfer-based variational autoencoder.

INTRODUCTION: Traditional methods for obtaining cellular responses after perturbation are usually labor-intensive and costly, especially when working with multiple different experimental conditions. Therefore, accurate prediction of cellular responses to perturbations is of great importance in computational biology. Existing methodologies, such as graph-based approaches, vector arithmetic, and neural networks, either mix perturbation-related variances with cell-type-specific patterns or implicitly distinguish them within black-box models. OBJECTIVES: This study aims to introduce and demonstrate a novel framework, scPerb, which explicitly extracts perturbation-related variances and transfers them from unperturbed to perturbed cells to accurately predict the effect of perturbation in single-cell level. METHODS: scPerb utilizes a style transfer strategy by incorporating a style encoder into the architecture of a variational autoencoder. The style encoder captures the differences in latent representations between unperturbed and perturbed cells, enabling accurate prediction of post-perturbation gene expression data. RESULTS: Comprehensive comparisons with existing methods demonstrate that scPerb delivers improved performance and higher accuracy in predicting cellular responses to perturbations. Notably, scPerb outperforms other methods across multiple datasets, achieving superior R2 values of 0.98, 0.98, and 0.96 on three benchmarking datasets. CONCLUSION: scPerb offers a significant advancement in predicting cellular responses by effectively separating and transferring perturbation-related variances. This framework not only enhances prediction accuracy but also provides a robust tool for computational biology, addressing the limitations of current methodologies.

PINet: Privileged Information Improve the Interpretablity and generalization of structural MRI in Alzheimer's Disease.

The irreversible and progressive atrophy by Alzheimer's Disease resulted in continuous decline in thinking and behavioral skills. To date, CNN classifiers were widely applied to assist the early diagnosis of AD and its associated abnormal structures. However, most existing black-box CNN classifiers relied heavily on the limited MRI scans, and used little domain knowledge from the previous clinical findings. In this study, we proposed a framework, named as PINet, to consider the previous domain knowledge as a Privileged Information (PI), and open the black-box in the prediction process. The input domain knowledge guides the neural network to learn representative features and introduced intepretability for further analysis. PINet used a Transformer-like fusion module Privileged Information Fusion (PIF) to iteratively calculate the correlation of the features between image features and PI features, and project the features into a latent space for classification. The Pyramid Feature Visualization (PFV) module served as a verification to highlight the significant features on the input images. PINet was suitable for neuro-imaging tasks and we demonstrated its application in Alzheimer's Disease using structural MRI scans from ADNI dataset. During the experiments, we employed the abnormal brain structures such as the Hippocampus as the PI, trained the model with the data from 1.5T scanners and tested from 3T scanners. The F1-score showed that PINet was more robust in transferring to a new dataset, with approximatedly 2% drop (from 0.9471 to 0.9231), while the baseline CNN methods had a 29% drop (from 0.8679 to 0.6154). The performance of PINet was relied on the selection of the domain knowledge as the PI. Our best model was trained under the guidance of 12 selected ROIs, major in the structures of Temporal Lobe and Occipital Lobe. In summary, PINet considered the domain knowledge as the PI to train the CNN model, and the selected PI introduced both interpretability and generalization ability to the black box CNN classifiers.

Photocatalysis triggered CVD synthesis of graphene at low temperature.

Conceiving the coupling strategy of photothermal catalysis, in which a CeO2 based photocatalyst facilitates the conversion of CH4 into hydrocarbon species with light irradiation and hydrocarbon species act for the subsequent CVD growth, high quality graphene has been successfully synthesized at 700 °C by photocatalysis triggered CH4-CVD over a Cu substrate.