Multiple Data Types Research Articles

A LoRa-based safety monitoring system for hydrogen compressed natural gas (HCNG)

Abstract Hydrogen compressed with natural gas pipeline for gas mixture delivery is an effective way to realize large-scale, low-cost delivery of hydrogen. The Internet of Things (IoT) technology is one of the important supporting technologies for smart energy transformation, and the low power wide area network (LPWAN) represented by LoRa will play a key role. In this paper, a LoRa-based hydrogen compressed natural gas safety monitoring system is designed, which can collect, transmit, and analyze multiple types of data, such as ambient temperature and humidity, gas concentration, pipeline and equipment vibration, etc., and store the data and configure the alarm thresholds through the cloud server, and finally complete the real-time safety monitoring in the application layer. The system is characterized by high efficiency, high precision, low power consumption, and intelligence, which will have a positive effect on the safe transportation of hydrogen.

Integration of multi-modal datasets to estimate human aging

Aging involves complex biological processes leading to the decline of living organisms. As population lifespan increases worldwide, the importance of identifying factors underlying healthy aging has become critical. Integration of multi-modal datasets is a powerful approach for the analysis of complex biological systems, with the potential to uncover novel aging biomarkers. In this study, we leveraged publicly available epigenomic, transcriptomic and telomere length data along with histological images from the Genotype-Tissue Expression project to build tissue-specific regression models for age prediction. Using data from two tissues, lung and ovary, we aimed to compare model performance across data modalities, as well as to assess the improvement resulting from integrating multiple data types. Our results demostrate that methylation outperformed the other data modalities, with a mean absolute error of 3.36 and 4.36 in the test sets for lung and ovary, respectively. These models achieved lower error rates when compared with established state-of-the-art tissue-agnostic methylation models, emphasizing the importance of a tissue-specific approach. Additionally, this work has shown how the application of Hierarchical Image Pyramid Transformers for feature extraction significantly enhances age modeling using histological images. Finally, we evaluated the benefits of integrating multiple data modalities into a single model. Combining methylation data with other data modalities only marginally improved performance likely due to the limited number of available samples. Combining gene expression with histological features yielded more accurate age predictions compared with the individual performance of these data types. Given these results, this study shows how machine learning applications can be extended to/in multi-modal aging research. Code used is available at https://github.com/zroger49/multi_modal_age_prediction.

The contribution of water-energy-food nexus governance to sustainability: A case study of Singapore

The water-energy-food (WEF) nexus has been promoted as an approach to integrated governance that can increase system sustainability. However, there have been limited empirical studies that demonstrate how WEF nexus governance leads towards sustainability. Therefore, the study investigates how collaborative governance of the WEF nexus connects to sustainability in Singapore, considered one of the most sustainable cities in the world. The objective is to understand the structure of WEF nexus governance in Singapore and explore how this structure contributes towards sustainability. This study uses a case study approach with qualitative text analysis on multiple data types to address this research objective. The results show two key findings. First, the findings show that WEF nexus governance in Singapore has a collaborative governance approach at the national level but is less integrated at lower levels of governance. Second, the results show that while governance integration, including WEF nexus connections, contributes to sustainability, there are also factors beyond WEF nexus governance that contribute towards sustainability in Singapore. The study concludes that WEF nexus governance provides an opportunity for improved sustainability within the system, but WEF nexus governance alone may not be sufficient for sustainability to be realized.

Monitoring Cover Crop Biomass in Southern Brazil Using Combined PlanetScope and Sentinel-1 SAR Data

Precision agriculture integrates multiple sensors and data types to support farmers with informed decision-making tools throughout crop cycles. This study evaluated Aboveground Biomass (AGB) estimates of Rye using attributes derived from PlanetScope (PS) optical, Sentinel-1 Synthetic Aperture Radar (SAR), and hybrid (optical plus SAR) datasets. Optical attributes encompassed surface reflectance from PS’s blue, green, red, and near-infrared (NIR) bands, alongside the Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI). Sentinel-1 SAR attributes included the C-band Synthetic Aperture Radar Ground Range Detected, VV and HH polarizations, and both Ratio and Polarization (Pol) indices. Ground reference AGB data for Rye (Secale cereal L.) were collected from 50 samples and four dates at a farm located in southern Brazil, aligning with image acquisition dates. Multiple linear regression models were trained and validated. AGB was estimated based on individual (optical PS or Sentinel-1 SAR) and combined datasets (optical plus SAR). This process was repeated 100 times, and variable importance was extracted. Results revealed improved Rye AGB estimates with integrated optical and SAR data. Optical vegetation indices displayed higher correlation coefficients (r) for AGB estimation (r = +0.67 for both EVI and NDVI) compared to SAR attributes like VV, Ratio, and polarization (r ranging from −0.52 to −0.58). However, the hybrid regression model enhanced AGB estimation (R2 = 0.62, p < 0.01), reducing RMSE to 579 kg·ha−1. Using only optical or SAR data yielded R2 values of 0.51 and 0.42, respectively (p < 0.01). In the hybrid model, the most important predictors were VV, NIR, blue, and EVI. Spatial distribution analysis of predicted Rye AGB unveiled agricultural zones associated with varying biomass throughout the cover crop development. Our findings underscored the complementarity of optical with SAR data to enhance AGB estimates of cover crops, offering valuable insights for agricultural zoning to support soil and cash crop management.

Online Learning Based on Learning Analytics in Big Data for College English Language Teaching

This study addresses the challenges of enhancing the quality of education and improving the overall student experience in online English language teaching sessions. Current approaches often struggle with session initiation, real-time data processing, and personalized learning experiences. To tackle these issues, the study proposes a manifold learning data analytics model (MLDAM). This innovative method leverages classifier tree learning to distinguish between trivial and non-trivial aspects of student learning experiences and session data. MLDAM integrates multi-dimensional data extraction, classification learning, and impact evaluation to enhance the effectiveness of online English language teaching. The model analyzes data from 176,009 English phrases across 36 online teaching sessions, focusing on improving session accessibility, student comprehension, and suggestion accuracy. Using an iterative training process based on student performance and feedback, it continuously extracts and processes multiple types of data to refine teaching strategies. Results show notable improvements: a 14.74% increase in classification accuracy, an 8.73% increase in data extraction ratio, an 11.84% reduction in feature discard, a 10.57% decrease in initialization time, and a 13.17% reduction in classification time. These metrics demonstrate MLDAM’s ability to efficiently process and analyze session data, enabling real-time adjustments during lessons. The model optimizes data utilization, making learning more responsive and adaptable. It enhances the precision of impact evaluations, facilitating targeted course adjustments and personalized learning experiences. This research presents a comprehensive, data-driven methodology for improving teaching quality and student outcomes in virtual English classrooms.

DORA: an interactive map for the visualization and analysis of ancient human DNA and associated data.

The ability to sequence ancient genomes has revolutionized the way we study evolutionary history by providing access to the most important aspect of evolution-time. Until recently, studying human demography, ecology, biology, and history using population genomic inference relied on contemporary genomic datasets. Over the past decade, the availability of human ancient DNA (aDNA) has increased rapidly, almost doubling every year, opening the way for spatiotemporal studies of ancient human populations. However, the multidimensionality of aDNA, with genotypes having temporal, spatial and genomic coordinates, and integrating multiple sources of data, poses a challenge for developing meta-analyses pipelines. To address this challenge, we developed a publicly-available interactive tool, DORA, which integrates multiple data types, genomic and non-genomic, in a unified interface. This web-based tool enables browsing sample metadata alongside additional layers of information, such as population structure, climatic data, and unpublished samples. Users can perform analyses on genotypes of these samples, or export sample subsets for external analyses. DORA integrates analyses and visualizations in a single intuitive interface, resolving the technical issues of combining datasets from different sources and formats, and allowing researchers to focus on the scientific questions that can be addressed through analysis of aDNA datasets.

Steckman Ridge: A naturally fractured underground gas storage field

This paper presents a field development case study of a gas storage field in Pennsylvania within a fractured reservoir. Integration of multiple data sets was required for the project's success. During the study, the interpretation of 3D isofrequency seismic volumes, generated through spectral decomposition, was shown to be the key step in this data integration. The interpretation of these isofrequency volumes successfully identified previously unresolved subtle structural features, which controlled the natural fracture system. The existence of these subtle structural features was subsequently confirmed with eight horizontal gas storage wells. Accurate characterization of a fracture trend requires the integration of seismic data, available well data, surface geology, drilling, and production data. This study was performed to develop the Oriskany reservoir in the Steckman Ridge underground gas storage field located in Bedford County, Pennsylvania. This field had undergone rapid depletion during its primary production phase. During the early portion of the study, the reservoir was determined to be a type 1 fractured reservoir with little to no gas storage available in the sandstone matrix. Multiple data types were acquired and integrated to detect and characterize the structural features that compose the reservoir. Isofrequency volumes, developed through spectral decomposition of 3D seismic data over the field, helped identify and map subtle shear faults that proved to be controlling the location and orientation of the dominant fracture trend in the field. Horizontal boreholes, needed to convert the field from gas production to gas storage, were planned and drilled based on this seismic interpretation. The existence of these shear faults, their location, and their orientation were confirmed by image logs acquired along the length of six horizontal gas storage boreholes. Additionally, flow tests performed to determine gas deliverability supported this interpretation.

Deciphering spatial domains from spatial multi-omics with SpatialGlue

Advances in spatial omics technologies now allow multiple types of data to be acquired from the same tissue slice. To realize the full potential of such data, we need spatially informed methods for data integration. Here, we introduce SpatialGlue, a graph neural network model with a dual-attention mechanism that deciphers spatial domains by intra-omics integration of spatial location and omics measurement followed by cross-omics integration. We demonstrated SpatialGlue on data acquired from different tissue types using different technologies, including spatial epigenome–transcriptome and transcriptome–proteome modalities. Compared to other methods, SpatialGlue captured more anatomical details and more accurately resolved spatial domains such as the cortex layers of the brain. Our method also identified cell types like spleen macrophage subsets located at three different zones that were not available in the original data annotations. SpatialGlue scales well with data size and can be used to integrate three modalities. Our spatial multi-omics analysis tool combines the information from complementary omics modalities to obtain a holistic view of cellular and tissue properties.

STEM-03. A NOVEL MOUSE MODEL OF CHOROID PLEXUS CARCINOMA UNRAVELS SEQUENTIAL DEDIFFERENTIATION OF HINDBRAIN CHOROID PLEXUS TISSUE AND REVEALS NEW THERAPEUTIC PROSPECTS

Abstract BACKGROUND Choroid Plexus tumors constitute 10-20% of brain tumors in the first year of life. Among those, Choroid Plexus Carcinomas (CPC) pose a significant challenge due to their aggressiveness, resulting in poor survival rates and long-term adverse effects stemming from intense therapeutic approaches. Although loss of p53 and Sonic hedgehog signaling have been implicated in CPC development, a deeper biological understanding is needed to expand mechanistic insights and advance targeted therapies. METHODS We generated a novel mouse model with a simultaneous loss of TP53 and overexpression of MYCN and Gli2 in central nervous system precursors (hGFAP-cre::P53Fl/Fl::lsl-MYCN::lsl-Gli2). Resulting tumors were histologically characterized and compared to human CPC. Moreover, spatial transcriptomic data were generated for molecular characterization. Finally, mouse tumor-derived cell lines were employed for differentiation analysis and therapeutic testing. RESULTS Examinations at different developmental stages revealed brain tumors in the choroid plexus as well as smaller lesions in the subcallosal cortex with a penetrance of 80% within 18 days. We observed a gradual dedifferentiation of hindbrain CP tissue from day 5 onwards, indicating a continuous developmental sequence corresponding with expanding tumor growth. The mouse model recapitulates early tumor development, aggressive and metastatic growth along with histological features of human tumors. Spatial transcriptomic analysis highlighted chromosomal aberrations as well as other features known from human CPCs and emphasized the disturbance of ciliary pathways in tumor development. In vitro studies revealed rapidly growing tumor cells with stem cell-like differentiation potential and high sensitivity to targeted therapies. CONCLUSION Through in-depth analysis, including multiple data types, we show that our novel CPC mouse model closely resembles human tumors. Our tumors reveal a strict pattern of dedifferentiation implying similar mechanisms in humans. Additionally, primary mouse tumor cells in culture facilitate further in vitro analysis, and newly generated spatial transcriptomic data enables further insights into tumor biology.

Multimodal deep learning for dementia classification using text and audio

Dementia is a progressive neurological disorder that affects the daily lives of older adults, impacting their verbal communication and cognitive function. Early diagnosis is important to enhance the lifespan and quality of life for affected individuals. Despite its importance, diagnosing dementia is a complex process. Automated machine learning solutions involving multiple types of data have the potential to improve the process of automated dementia screening. In this study, we build deep learning models to classify dementia cases from controls using the Pitt Cookie Theft dataset from DementiaBank, a database of short participant responses to the structured task of describing a picture of a cookie theft. We fine-tune Wav2vec and Word2vec baseline models to make binary predictions of dementia from audio recordings and text transcripts, respectively. We conduct experiments with four versions of the dataset: (1) the original data, (2) the data with short sentences removed, (3) text-based augmentation of the original data, and (4) text-based augmentation of the data with short sentences removed. Our results indicate that synonym-based text data augmentation generally enhances the performance of models that incorporate the text modality. Without data augmentation, models using the text modality achieve around 60% accuracy and 70% AUROC scores, and with data augmentation, the models achieve around 80% accuracy and 90% AUROC scores. We do not observe significant improvements in performance with the addition of audio or timestamp information into the model. We include a qualitative error analysis of the sentences that are misclassified under each study condition. This study provides preliminary insights into the effects of both text-based data augmentation and multimodal deep learning for automated dementia classification.

Efficient and Portable Vectorized Sweep Kernels for the Transport Equation on 3D Cartesian Grids Using the Kokkos Framework

This paper describes the implementation of efficient and portable vectorized sweep kernels as part of the resolution of the neutron transport equation on three-dimensional Cartesian grids using the discrete ordinates ( S n ) method for the angular variable and the diamond differencing (DD) scheme for the spatial discretization. Vectorization is set up along the directions within the same octant and is independent of the spatial discretization order; therefore, the extension of this technique to high-order DD or discontinuous Galerkin schemes is immediate. Our implementation is written in C++17 and relies on the Kokkos performance portability framework. This library allows one to express shared-memory parallelism (including vectorization) in a machine-independent way and supports many backends including CUDA and OpenMP. Our vectorization procedure relies on the portable single instruction multiple data types provided by Kokkos. The method has been implemented for DD schemes up to order 2 and yields promising results on CPUs supporting standard vector instructions.

Transfer contrast learning based on model-level data enhancement for cross-domain recommendation

A cross-domain recommendation system is an intelligent recommendation technology that integrates multiple fields or types of data. It can cross independent information islands, effectively integrate and complement data resources, and improve recommendation performance. This paper proposes a transfer contrast learning method based on model-level data enhancement for cross-domain recommendations. This method first obtains the initial embeddings of the two domains using item-based collaborative filtering, after which it enhances the transformer network with model-level data through contrastive learning to pre-train the source domain data. The pre-trained transformer network parameters are then transferred and fine-tuned before being applied to tasks on the target domain data. The information link from the source domain to the target domain is effectively constructed, and it has been proven to improve the accuracy and effectiveness of the target domain on real datasets.

Using expression data to fine map QTL associated with fertility in dairy cattle

BackgroundFemale fertility is an important trait in dairy cattle. Identifying putative causal variants associated with fertility may help to improve the accuracy of genomic prediction of fertility. Combining expression data (eQTL) of genes, exons, gene splicing and allele specific expression is a promising approach to fine map QTL to get closer to the causal mutations. Another approach is to identify genomic differences between cows selected for high and low fertility and a selection experiment in New Zealand has created exactly this resource. Our objective was to combine multiple types of expression data, fertility traits and allele frequency in high- (POS) and low-fertility (NEG) cows with a genome-wide association study (GWAS) on calving interval in Australian cows to fine-map QTL associated with fertility in both Australia and New Zealand dairy cattle populations.ResultsVariants that were significantly associated with calving interval (CI) were strongly enriched for variants associated with gene, exon, gene splicing and allele-specific expression, indicating that there is substantial overlap between QTL associated with CI and eQTL. We identified 671 genes with significant differential expression between POS and NEG cows, with the largest fold change detected for the CCDC196 gene on chromosome 10. Our results provide numerous candidate genes associated with female fertility in dairy cattle, including GYS2 and TIGAR on chromosome 5 and SYT3 and HSD17B14 on chromosome 18. Multiple QTL regions were located in regions with large numbers of copy number variants (CNV). To identify the causal mutations for these variants, long read sequencing may be useful.ConclusionsVariants that were significantly associated with CI were highly enriched for eQTL. We detected 671 genes that were differentially expressed between POS and NEG cows. Several QTL detected for CI overlapped with eQTL, providing candidate genes for fertility in dairy cattle.

Liam tackles complex multimodal single-cell data integration challenges.

Multi-omics characterization of single cells holds outstanding potential for profiling the dynamics and relations of gene regulatory states of thousands of cells. How to integrate multimodal data is an open problem, especially when aiming to combine data from multiple sources or conditions containing both biological and technical variation. We introduce liam, a flexible model for the simultaneous horizontal and vertical integration of paired single-cell multimodal data and mosaic integration of paired with unimodal data. Liam learns a joint low-dimensional representation of the measured modalities, which proves beneficial when the information content or quality of the modalities differ. Its integration accounts for complex batch effects using a tunable combination of conditional and adversarial training, which can be optimized using replicate information while retaining selected biological variation. We demonstrate liam's superior performance on multiple paired multimodal data types, including Multiome and CITE-seq data, and in mosaic integration scenarios. Our detailed benchmarking experiments illustrate the complexities and challenges remaining for integration and the meaningful assessment of its success.

A time series analysis of disruptions to maternal and child health care in northern Togo during the COVID-19 pandemic in the context of an integrated primary care program

There are significant concerns that COVID-19-related disruptions in routine health care will reverse the gains made over the past 2 decades in reducing maternal and under-5 mortality in low- and middle-income countries. The objective of this study was to examine the extent of disruptions in community-based maternal and child health services and explore community perceptions and experiences with health care use in the year following the start of the COVID-19 pandemic in one district in northern Togo, West Africa. This study included analysis of multiple types of data including: (1) routine health management information system data from 18 public sector health centers, (2) programmatic data collected through an integrated health center- and community-based primary care program at 5 health centers, and (3) semi-structured interviews with health center managers, community health workers (CHWs), and community members at those 5 health centers. We found only short-term declines in service utilization coinciding with the COVID-19 pandemic and no decline in the delivery of community-based care by CHWs. Qualitative data were consistent with the quantitative results, reporting sustained use of health-care services. Multilevel factors related to the continued provision of care during the COVID-19 pandemic included regular and clear communication from CHWs as trusted community sources, risk minimization at the health-center level, continued provision of community-based care by CHWs, and collaboration between community leaders and health-care workers to limit COVID-19 transmission. Findings demonstrate resiliency of an integrated primary care system when equipped with an adequately trained, supervised, and supplied health workforce, implementation of infection prevention and control measures, communication by trusted community sources, and adaptations to health-care delivery that enable the continued provision of care.

Prediction of prunoideae fruit quality characteristics based on machine learning and spectral characteristic acquisition optimization

Machine learning algorithms have been widely used in the estimation and prediction of various quality characteristics of food and agricultural products. The purpose of this study is to study the availability and accuracy of the prediction model of Prunoideae fruit quality characteristics based on machine learning algorithms. XGBoost, LightGBM, CatBoost, and Random Forest (RF) are the four machine learning algorithms used in this work to create prediction models for the soluble solids content (SSC) and titratable acidity (TA) of peach, apricot, and cherry. The study investigated three hyperspectral denoising methods and used two feature extraction techniques to lower the dimensionality of hyperspectral data. The models' accuracy was further improved by adding multiple types of spectral data. Based on opposite spectral data input, the MLP-SG-XGBoost model produced the best SSC predictions for peach and apricot, with R2 values of 0.9162 and 0.9251, respectively, according to experimental results. When inputting opposite spectral data, the LGR-SG-LightGBM model showed the highest accuracy in predicting TA for peach and apricot, with R2 values of 0.9193 and 0.9206, respectively. The outcomes imply that the models can accurately forecast the qualities of Prunoideae fruit.

Modelling and Validation of the Derna Dam Break Event

The catastrophic failure of two dams in Libya on 10 and 11 September 2023 resulted in the devastating flooding of the city of Derna, which is located downstream of the dams, causing more than 6000 fatalities and displacing thousands of residents. The failure was attributed to heavy rainfall from Storm Daniel, leading to the dams reaching full capacity and subsequently overflowing and failing. This paper presents an analysis of the dam break, including the modelling of flow discharge and the resulting flooding of Derna. For validation purposes, this study compares the modelled quantities with post-event satellite imagery from UNOSAT and Copernicus, local reports, and data collected from social media using AI detection. The findings provide valuable insights into the dynamics of the dam break and its initial parameters, as well as an assessment of the accuracy of the results. The analysis is performed using a rapid estimation technique developed by JRC to provide the international emergency community with a swift overview of the impact and damage assessment of potential or actual dam break events. The use of all available data shows a satisfactory comparison with the calculated quantities. The rapid modelling of dam break events and combined analysis of multiple data types are proven suitable for promptly assessing the expected dynamic of the event, as well as reconstructing the unknown initial conditions before the break. Incorporating sensitivity analyses provides an estimate of the uncertainties associated with the deduced values of the unknown parameters and their relative importance in the analysis.

DCNeT: A disease comorbidity network-based temporal deep learning framework to predict cardiovascular risk in patients with mental disorders

Patients with mental disorders (MDs) are at higher subsequent risk of developing cardiovascular diseases (CVDs) than the general population. Early identification of cardiovascular risk in patients with MDs is beneficial for timely intervention and reducing disease burden. Recently, deep learning approaches have been increasingly applied in CVDs risk prediction. However, these methods have three major issues: 1) mostly relying on multiple types of clinical data, 2) not sufficiently mining and utilizing comorbidity patterns hidden in complex correlations among various diseases, and 3) not fully leveraging the time information, including the irregular intervals. To address these issues, we propose a disease comorbidity network-based temporal deep learning framework (DCNeT) to predict the subsequent CVDs risk for patients with MDs based on routinely collected administrative health data. Firstly, to identify the comorbidity patterns of MDs, we construct a disease comorbidity network (DCN) for MDs and apply graph embedding method to generate disease embeddings for each disease in the DCN. Then, a code attention mechanism is proposed to obtain the weight of each disease which is embedded into dense vectors based on the structure of DCN. We present a view attention mechanism to compute the attention weights of different types of features including disease embeddings, basic features, and disease indicators for generating the final representations of patients’ hospitalizations. Furthermore, to fully utilize the information on the irregular time intervals between hospitalizations, a time encoding module is designed, and the time-aware LSTM is adopted to model the irregular time intervals and capture the temporal patterns of patients’ hospitalizations. The experimental results show that DCNeT outperforms the state-of-the-art methods, with the area under the receiver operating characteristic curve of 0.7658, 0.8143, 0.8110, and 0.7839 on four datasets, respectively. The ablation experiments further demonstrate that each module of DCNeT, including the code attention, view attention, and time encoding module, contributes to its superior performance, with average improvements of 1.20 %, 1.65 %, and 1.13 % in accuracy, respectively. Our DCNeT could be utilized as an efficient framework for identifying high-risk groups of CVDs among patients with MDs that may benefit from screening and preventive strategies.

Evaluating the Effect of Multiple Data Types on a Large Scale Static Origin–Destination Matrix Estimation, Case Study of Tehran, Iran

Evaluating the Effect of Multiple Data Types on a Large Scale Static Origin–Destination Matrix Estimation, Case Study of Tehran, Iran

Multi-view data visualisation via manifold learning

Non-linear dimensionality reduction can be performed by manifold learning approaches, such as stochastic neighbour embedding (SNE), locally linear embedding (LLE) and isometric feature mapping (ISOMAP). These methods aim to produce two or three latent embeddings, primarily to visualise the data in intelligible representations. This manuscript proposes extensions of Student’s t-distributed SNE (t-SNE), LLE and ISOMAP, for dimensionality reduction and visualisation of multi-view data. Multi-view data refers to multiple types of data generated from the same samples. The proposed multi-view approaches provide more comprehensible projections of the samples compared to the ones obtained by visualising each data-view separately. Commonly, visualisation is used for identifying underlying patterns within the samples. By incorporating the obtained low-dimensional embeddings from the multi-view manifold approaches into the K-means clustering algorithm, it is shown that clusters of the samples are accurately identified. Through extensive comparisons of novel and existing multi-view manifold learning algorithms on real and synthetic data, the proposed multi-view extension of t-SNE, named multi-SNE, is found to have the best performance, quantified both qualitatively and quantitatively by assessing the clusterings obtained. The applicability of multi-SNE is illustrated by its implementation in the newly developed and challenging multi-omics single-cell data. The aim is to visualise and identify cell heterogeneity and cell types in biological tissues relevant to health and disease. In this application, multi-SNE provides an improved performance over single-view manifold learning approaches and a promising solution for unified clustering of multi-omics single-cell data.