Single Nucleotide Polymorphism (SNP) is a common form of genetic variation, and genomic prediction is an emerging technique that utilizes SNP information to predict phenotypes in animals and plants. It is gradually being applied in animal and plant breeding as well as human disease risk assessment. Traditional statistical learning methods can only focus on linear interactions between the genome and phenotypes. Machine learning methods and deep learning methods have become popular due to their ability to recognize non-linear interactions between SNPs. However, existing deep learning methods often only learn short-distance interactions between SNPs and overlook long-distance interactions. Therefore, we propose a genomic prediction method called DCNNCSA (DualCNN Channel Spatial Attention) based on a channel spatial attention mechanism. DCNNCSA first uses a dual-branch convolutional neural network to extract features in the channel and spatial dimensions. Then, it utilizes a channel spatial attention mechanism to identify long-distance interactions between SNPs, thereby improving the accuracy of genomic prediction. Experimental evaluations are conducted on datasets including wheat 2000 and wheat 599 datasets. The results demonstrate that DCNNCSA outperforms rrBLUP, LASSO, XGBoost, Random Forest, DeepGS, and DLGWAS methods in terms of prediction accuracy.
Synthetic lethal interactions are critical genetic interactions that have been discovered for identifying new drug targets and potential cancer drug combination strategies. As a targeted approach to selectively kill cancer cells, it has attracted great attention in the field of cancer treatment. However, with the rapid growth of high-throughput data, effective identification of synthetic lethal interactions remains challenging. Although many graph topology methods focus on predicting SL interactions, heterogeneous graphs that exist in the real world tend to be scale-free, where the number of high-order neighbor nodes of high-degree nodes grows exponentially. The graph embedding method based on Euclidean space may not be able to effectively capture the internal hierarchical structure of the scale-free network, and there is not enough space to accommodate the nodes, making the node embedding highly distorted, which greatly limits the modeling ability. In this paper, we propose a model named HLASL based on the hyperbolic Lorentz attention mechanism, it ensures that the learned node features follow a Lorentz manifold. Specifically, in hyperbolic space, we use Lorentzian manifolds to encode heterogeneous information from KGs and information about synthetic lethality. In addition, we introduce the Lorenz knowledge graph attention mechanism to distinguish the weights of different information from the hyperbolic space to optimize gene embedding. Experiments on two datasets show that HLASL outperforms several state-of-the-art baseline methods in synthetic lethality prediction.
Single-cell RNA sequencing allows to discovery of new cell subtypes based on transcriptomic information. Clustering analysis is an effective approach for exploring single-cell heterogeneity. Nevertheless, current single-cell clustering methodologies are challenged by the high-dimensional, sparse, and dropouts on the one hand, and on the other hand, they neglect the thorough exploration of potential relationships between cells and noise removal. To tackle these challenges, this study introduces a novel self-optimized single-cell clustering algorithm named scZVEA, which combines ZeroInflated Negative Binomial (ZINB) model and variational graph attention autoencoder. The scZVEA framework comprises three key modules. Firstly, a Deep Count Autoencoder (DCA) is designed to model data distribution and eliminate noise. Then, a variational graph attention autoencoder to extract latent features from the data. Lastly, the self-optimized clustering module enables the two previously independent clustering and feature modules to mutually benefit each other by iteratively updating cluster centers to further enhance clustering performance. Experimental findings based on six authentic scRNAseq datasets illustrate that the proposed clustering algorithm significantly enhances clustering accuracy when compared to alternative methods.
Phenotypic prediction before crop planting and harvest facilitates plant phenotyping analysis and the implementation of precision agriculture, which is crucial for food security policy formulation, crop management, and food security. Environmental data collected in the field, as important factors affecting phenotypes, can be used together with genotypes to train phenotypic prediction models to improve their prediction capabilities. Although the rise of deep learning has provided a powerful tool for predicting phenotypes using genotype and environmental data, current research only performs simple splicing operations when processing genotype and environmental data and fails to effectively model the impact of genotype and environment on phenotype. Therefore, it is still challenging to accurately predict phenotype using genotype and environment. To solve the problems mentioned above, we design a phenotype prediction model named FF-LSTM. In this method, LSTM is used to extract features from SNP and environmental, and the feature fusion method weights the extracted features to simulate the influence of genotype and environment on phenotype. We conducted experiments on the dataset and verified the effectiveness of this method by analyzing and evaluating the experimental results.
Early detection and treatment of esophageal cancer may improve the survival rate of patients, despite its high incidence and mortality. The use of computer technology can assist in the diagnosis of esophageal cancer. RNA-Seq gene expression data can be used for the diagnosis of esophageal cancer, but it is difficult to analyze directly because of its high dimension and small sample size. Applying computer technology to this data can solve these problems. In our work, we used the RNA-Seq gene expression dataset and considered the specificity of the sample, proposed an artificial intelligence approach for esophageal cancer classification through selecting the comprehensive features of RNA-Seq gene expression data using mutual information feature selection and obtaining a set of sample specific features by generating adversarial examples using one-pixel attack method to reduce the dimensionality of the dataset. Finally, the deep learning method is used to construct a deep neural network as the classifier. The experimental results reveal that this method outperforms other state-of-the-art algorithms in terms of accuracy and other metrics.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.