Search

Molecular property prediction is essential for drug screening and reducing the cost of drug discovery. Current approaches combined with deep learning for drug prediction have proven their viability. Based on the previous deep learning networks, we propose the Molecular Information Fusion Neural Network (MIFNN). The features of MIFNN are as follows: (1) we extracted directed molecular information using 1D-CNN and the Morgan fingerprint using 2D-CNN to obtain more comprehensive feature information; (2) we fused two molecular features from one-dimensional and two-dimensional space, and we used the directed message-passing method to reduce the repeated collection of information and improve efficiency; (3) we used a bidirectional long short-term memory and attention module to adjust the molecular feature information and improve classification accuracy; (4) we used the particle swarm optimization algorithm to improve the traditional support vector machine. We tested the performance of the model on eight publicly available datasets. In addition to comparing the overall classification capability with the baseline model, we conducted a series of ablation experiments to verify the optimization of different modules in the model. Compared with the baseline model, our model achieved a maximum improvement of 14% on the ToxCast dataset. The performance was very stable on most datasets. On the basis of the current experimental results, MIFNN performed better than previous models on the datasets applied in this paper.

Infrared images can provide clear contrast information to distinguish between the target and the background under any lighting conditions. In contrast, visible images can provide rich texture details and are compatible with the human visual system. The fusion of a visible image and infrared image will thus contain both comprehensive contrast information and texture details. In this study, a novel approach for the fusion of infrared and visible images is proposed based on a dual-discriminator generative adversarial network with a squeeze-and-excitation module (DDGANSE). Our approach establishes confrontation training between one generator and two discriminators. The goal of the generator is to generate images that are similar to the source images, and contain the information from both infrared and visible source images. The purpose of the two discriminators is to increase the similarity between the image generated by the generator and the infrared and visible images. We experimentally demonstrated that using continuous adversarial training, DDGANSE outputs images retain the advantages of both infrared and visible images with significant contrast information and rich texture details. Finally, we compared the performance of our proposed method with previously reported techniques for fusing infrared and visible images using both quantitative and qualitative assessments. Our experiments on the TNO dataset demonstrate that our proposed method shows superior performance compared to other similar reported methods in the literature using various performance metrics.