Strategic Multi-Omics Data Integration via Multi-Level Feature Contrasting and Matching.

The analysis and comprehension of multi-omics data has emerged as a prominent topic in the field of bioinformatics and data science. However, the sparsity characteristics and high dimensionality of omics data pose difficulties in terms of extracting meaningful information. Moreover, the heterogeneity inherent in multiple omics sources makes the effective integration of multi-omics data challenging To tackle these challenges, we propose MFCC-SAtt, a multi-level feature contrast clustering model based on self-attention to extract informative features from multi-omics data. MFCC-SAtt treats each omics type as a distinct modality and employs autoencoders with self-attention for each modality to integrate and compress their respective features into a shared feature space. By utilizing a multi-level feature extraction framework along with incorporating a semantic information extractor, we mitigate optimization conflicts arising from different learning objectives. Additionally, MFCC-SAtt guides deep clustering based on multi-level features which further enhances the quality of output labels. By conducting extensive experiments on multi-omics data, we have validated the exceptional performance of MFCC-SAtt. For instance, in a pan-cancer clustering task, MFCC-SAtt achieved an accuracy of over 80.38%.

CD-Net: Cascaded 3D Dilated convolutional neural network for pneumonia lesion segmentation.

COVID-19 is a global pandemic that has caused significant global, social, and economic disruption. To effectively assist in screening and monitoring diagnosed cases, it is crucial to accurately segment lesions from Computer Tomography (CT) scans. Due to the lack of labeled data and the presence of redundant parameters in 3D CT, there are still significant challenges in diagnosing COVID-19 in related fields. To address the problem, we have developed a new model called the Cascaded 3D Dilated convolutional neural network (CD-Net) for directly processing CT volume data. To reduce memory consumption when cutting volume data into small patches, we initially design a cascade architecture in CD-Net to preserve global information. Then, we construct a Multi-scale Parallel Dilated Convolution (MPDC) block to aggregate features of different sizes and simultaneously reduce the parameters. Moreover, to alleviate the shortage of labeled data, we employ classical transfer learning, which requires only a small amount of data while achieving better performance. Experimental results conducted on the different public-available datasets verify that the proposed CD-Net has reduced the negative-positive ratio and outperformed other existing segmentation methods while requiring less data.

FedMCC: Federated multi-center clustering algorithm to improve privacy healthcare.

In recent years, healthcare data from various sources such as clinical institutions, patients, and pharmaceutical industries have become increasingly abundant. However, due to the complex healthcare system and data privacy concerns, aggregating and utilizing these data in a centralized manner can be challenging. Federated learning (FL) has emerged as a promising solution for distributed training in edge computing scenarios, utilizing on-device user data while reducing server costs. In traditional FL, a central server trains a global model sampled client data randomly, and the server combines the collected model from different clients into one global model. However, for not independent and identically distributed (non-i.i.d.) datasets, randomly selecting users to train server is not an optimal choice and can lead to poor model training performance. To address this limitation, we propose the Federated Multi-Center Clustering algorithm (FedMCC) to enhance the robustness and accuracy for all clients. FedMCC leverages the Model-Agnostic Meta-Learning (MAML) algorithm, focusing on training a robust base model during the initial training phase and better capturing features from different users. Subsequently, clustering methods are used to ensure that features among users within each cluster are similar, approximating an i.i.d. training process in each round, resulting in more effective training of the global model. We validate the effectiveness and generalizability of FedMCC through extensive experiments on public healthcare datasets. The results demonstrate that FedMCC achieves improved performance and accuracy for all clients while maintaining data privacy and security, showcasing its potential for various healthcare applications.

Spatiotemporal variability of glacier changes and their controlling factors in the Kanchenjunga region, Himalaya based on multi-source remote sensing data from 1975 to 2015.

Spatio-temporal behavior of glaciers in the Himalayas has varied greatly in response to reported climate warming and other modulating factors such as topography, debris cover, and glacier morphology. In this paper, 429 glaciers were examined in the Kanchenjunga region in the middle of the Himalayas. Geodetic methods, feature-based image matching, and multi-parametric integrated approaches were used to detect differences of glacier change and the dominant characteristics driving these differences based on digital elevation models (DEMs), Landsat TM/ETM+/OLI images, Envisat/ASAR and Sentinel-1 data. The results showed that the average change rates in glacier area and surface elevation in 1975-2015 were -0.18 ± 0.07% a-1 and - 0.32 ± 0.02 m a-1, respectively. The rates of areal shrinkage of glaciers and the glacier surface velocity on the northern side of the Himalayan crest were 1.25 and 1.7 times larger than those of the glaciers on the southern slopes, respectively, whereas the rates of glacier thinning were lower in the north than in the south. The temperature increase from 1975 to 2015 caused an overall widespread glacier retreat in the region. However, differences in the topography of the Kanchenjunga region led to spatial variability in glacier changes with discrepancies as large as several times. The features of individual glaciers, such as glacier size, debris cover, and development of ice-contact glacial lakes enhanced the local complexity of glacier change and elusive response behaviors of the glaciers to climate warming led by the different topographic conditions.

A novel graph attention adversarial network for predicting disease-related associations.

Identifying complex human diseases at molecular level is very helpful, especially in diseases diagnosis, therapy, prognosis and monitoring. Accumulating evidences demonstrated that RNAs are playing important roles in identifying various complex human diseases. However, the amount of verified disease-related RNAs is still little while many of their biological experiments are very time-consuming and labor-intensive. Therefore, researchers have instead been seeking to develop effective computational algorithms to predict associations between diseases and RNAs. In this paper, we propose a novel model called Graph Attention Adversarial Network (GAAN) for the potential disease-RNA association prediction. To our best knowledge, we are among the pioneers to integrate successfully both the state-of-the-art graph convolutional networks (GCNs) and attention mechanism in our model for the prediction of disease-RNA associations. Comparing to other disease-RNA association prediction methods, GAAN is novel in conducting the computations from the aspect of global structure of disease-RNA network with graph embedding while integrating features of local neighborhoods with the attention mechanism. Moreover, GAAN uses adversarial regularization to further discover feature representation distribution of the latent nodes in disease-RNA networks. GAAN also benefits from the efficiency of deep model for the computation of big associations networks. To evaluate the performance of GAAN, we conduct experiments on networks of diseases associating with two different RNAs: MicroRNAs (miRNAs) and Long non-coding RNAs (lncRNAs). Comparisons of GAAN with several popular baseline methods on disease-RNA networks show that our novel model outperforms others by a wide margin in predicting potential disease-RNAs associations.

Mass Change of Glaciers in Muztag Ata-Kongur Tagh, Eastern Pamir, China from 1971/76 to 2013/14 as Derived from Remote Sensing Data.

The assessment of glacier mass budget is crucial for assessing water reserves stored in glaciers. Derived glacier mass changes in the Muztag Ata and Kongur Tagh (MAKT) region in the eastern Pamir, northwestern China, is helpful in improving our knowledge of the dynamics of glaciers under a changing climate in High Mountain Asia. Here, glacier area and mass changes derived from remote sensing data are investigated for the period 1971/76-2013/14 for glaciers in MAKT. We have used ASTER images (2013/14), Cartosat-1 (2014) and Landsat, SRTM (Shuttle Radar Terrain Mission) digital elevation model (DEM) (2000), topographic maps (1971/76) and the first and second Chinese glacier inventories (CGIs). Our results indicated that the glacier area of MAKT decreased from 1018.3 ± 12.99 km(2) in 1971/76 to 999.2 ± 31.22 km(2) in 2014 (-1.9 ± 0.2%). Weak area shrinkage of glaciers by 2.5 ± 0.5 km(2) (0.2 ± 0.1%) happened after 2000 and the period 2009-2014 even saw a slight expansion by 0.5 ± 0.1 km(2) (0.1 ± 0.0%). The glaciers in this region have experienced an overall loss of -6.99 ± 0.80 km(3) in ice volume or -0.15 ± 0.12 m water equivalent (w.e.) a-1 from 1971/76 to 2013/14. The mass budget of MAKT was -0.19 ± 0.19 m w.e. a-1 for the period ~1971/76-1999 and -0.14 ± 0.24 m w.e. a-1 during 1999-2013/2014. Similar to previous studies, there has been little mass change in the Pamir over recent decades despite such uncertainties. Glacier mass change showed spatial and temporal heterogeneity, with strong mass loss on debris-covered glaciers with an average of -0.32 ± 0.12 m w.e. a-1 from the 1970s to 2013/14.