ISMI-VAE: A deep learning model for classifying disease cells using gene expression and SNV data.

Various studies have linked several diseases, including cancer and COVID-19, to single nucleotide variations (SNV). Although single-cell RNA sequencing (scRNA-seq) technology can provide SNV and gene expression data, few studies have integrated and analyzed these multimodal data. To address this issue, we introduce Interpretable Single-cell Multimodal Data Integration Based on Variational Autoencoder (ISMI-VAE). ISMI-VAE leverages latent variable models that utilize the characteristics of SNV and gene expression data to overcome high noise levels and uses deep learning techniques to integrate multimodal information, map them to a low-dimensional space, and classify disease cells. Moreover, ISMI-VAE introduces an attention mechanism to reflect feature importance and analyze genetic features that could potentially cause disease. Experimental results on three cancer data sets and one COVID-19 data set demonstrate that ISMI-VAE surpasses the baseline method in terms of both effectiveness and interpretability and can effectively identify disease-causing gene features.

Iterated Clique Reductions in Vertex Weighted Coloring for Large Sparse Graphs.

The Minimum Vertex Weighted Coloring (MinVWC) problem is an important generalization of the classic Minimum Vertex Coloring (MinVC) problem which is NP-hard. Given a simple undirected graph G=(V,E), the MinVC problem is to find a coloring s.t. any pair of adjacent vertices are assigned different colors and the number of colors used is minimized. The MinVWC problem associates each vertex with a positive weight and defines the weight of a color to be the weight of its heaviest vertices, then the goal is the find a coloring that minimizes the sum of weights over all colors. Among various approaches, reduction is an effective one. It tries to obtain a subgraph whose optimal solutions can conveniently be extended into optimal ones for the whole graph, without costly branching. In this paper, we propose a reduction algorithm based on maximal clique enumeration. More specifically our algorithm utilizes a certain proportion of maximal cliques and obtains lower bounds in order to perform reductions. It alternates between clique sampling and graph reductions and consists of three successive procedures: promising clique reductions, better bound reductions and post reductions. Experimental results show that our algorithm returns considerably smaller subgraphs for numerous large benchmark graphs, compared to the most recent method named RedLS. Also, we evaluate individual impacts and some practical properties of our algorithm. Furthermore, we have a theorem which indicates that the reduction effects of our algorithm are equivalent to that of a counterpart which enumerates all maximal cliques in the whole graph if the run time is sufficiently long.

ProgClust: A progressive clustering method to identify cell populations.

Identifying different types of cells in scRNA-seq data is a critical task in single-cell data analysis. In this paper, we propose a method called ProgClust for the decomposition of cell populations and detection of rare cells. ProgClust represents the single-cell data with clustering trees where a progressive searching method is designed to select cell population-specific genes and cluster cells. The obtained trees reveal the structure of both abundant cell populations and rare cell populations. Additionally, it can automatically determine the number of clusters. Experimental results show that ProgClust outperforms the baseline method and is capable of accurately identifying both common and rare cells. Moreover, when applied to real unlabeled data, it reveals potential cell subpopulations which provides clues for further exploration. In summary, ProgClust shows potential in identifying subpopulations of complex single-cell data.

DeepMPM: a mortality risk prediction model using longitudinal EHR data.

BACKGROUND: Accurate precision approaches have far not been developed for modeling mortality risk in intensive care unit (ICU) patients. Conventional mortality risk prediction methods can hardly extract the information in longitudinal electronic medical records (EHRs) effectively, since they simply aggregate the heterogeneous variables in EHRs, ignoring the complex relationship and interactions between variables and the time dependence in longitudinal records. Recently deep learning approaches have been widely used in modeling longitudinal EHR data. However, most existing deep learning-based risk prediction approaches only use the information of a single disease, neglecting the interactions between multiple diseases and different conditions. RESULTS: In this paper, we address this unmet need by leveraging disease and treatment information in EHRs to develop a mortality risk prediction model based on deep learning (DeepMPM). DeepMPM utilizes a two-level attention mechanism, i.e. visit-level and variable-level attention, to derive the representation of patient risk status from patient's multiple longitudinal medical records. Benefiting from using EHR of patients with multiple diseases and different conditions, DeepMPM can achieve state-of-the-art performances in mortality risk prediction. CONCLUSIONS: Experiment results on MIMIC III database demonstrates that with the disease and treatment information DeepMPM can achieve a good performance in terms of Area Under ROC Curve (0.85). Moreover, DeepMPM can successfully model the complex interactions between diseases to achieve better representation learning of disease and treatment than other deep learning approaches, so as to improve the accuracy of mortality prediction. A case study also shows that DeepMPM offers the potential to provide users with insights into feature correlation in data as well as model behavior for each prediction.

Fog-Based Two-Phase Event Monitoring and Data Gathering in Vehicular Sensor Networks.

Vehicular nodes are equipped with more and more sensing units, and a large amount of sensing data is generated. Recently, more and more research considers cooperative urban sensing as the heart of intelligent and green city traffic management. The key components of the platform will be a combination of a pervasive vehicular sensing system, as well as a central control and analysis system, where data-gathering is a fundamental component. However, the data-gathering and monitoring are also challenging issues in vehicular sensor networks because of the large amount of data and the dynamic nature of the network. In this paper, we propose an efficient continuous event-monitoring and data-gathering framework based on fog nodes in vehicular sensor networks. A fog-based two-level threshold strategy is adopted to suppress unnecessary data upload and transmissions. In the monitoring phase, nodes sense the environment in low cost sensing mode and generate sensed data. When the probability of the event is high and exceeds some threshold, nodes transfer to the event-checking phase, and some nodes would be selected to transfer to the deep sensing mode to generate more accurate data of the environment. Furthermore, it adaptively adjusts the threshold to upload a suitable amount of data for decision making, while at the same time suppressing unnecessary message transmissions. Simulation results showed that the proposed scheme could reduce more than 84 percent of the data transmissions compared with other existing algorithms, while it detects the events and gathers the event data.

Adaptive Data Gathering in Mobile Sensor Networks Using Speedy Mobile Elements.

Data gathering is a key operator for applications in wireless sensor networks; yet it is also a challenging problem in mobile sensor networks when considering that all nodes are mobile and the communications among them are opportunistic. This paper proposes an efficient data gathering scheme called ADG that adopts speedy mobile elements as the mobile data collector and takes advantage of the movement patterns of the network. ADG first extracts the network meta-data at initial epochs, and calculates a set of proxy nodes based on the meta-data. Data gathering is then mapped into the Proxy node Time Slot Allocation (PTSA) problem that schedules the time slots and orders, according to which the data collector could gather the maximal amount of data within a limited period. Finally, the collector follows the schedule and picks up the sensed data from the proxy nodes through one hop of message transmissions. ADG learns the period when nodes are relatively stationary, so that the collector is able to pick up the data from them during the limited data gathering period. Moreover, proxy nodes and data gathering points could also be timely updated so that the collector could adapt to the change of node movements. Extensive experimental results show that the proposed scheme outperforms other data gathering schemes on the cost of message transmissions and the data gathering rate, especially under the constraint of limited data gathering period.