Mavatar Discovery

The platform contains an extensive catalog of DINA (Deep Integrated Network Analysis) resources from a wide spectrum of tissues, and disease-specific curated resources from transcriptomics data: microarray, bulk and single cell RNA-seq. Networks represent strongly preserved gene-gene interactions, with readily available detailed functional characterization, based on a database of curated meta data.

At the core of Mavatar Discovery is DINA (Deep Integrated Network Analysis), our proprietary framework built on 20+ years of scientific expertise. DINA integrates and analyzes thousands of transcriptomic datasets to model disease mechanisms at a systems level—delivering:

Our approach also ensures high reproducibility, enables more accurate translation between animal and human studies, allows us to compare different diseases affecting the same tissue, and how the same disease is present in different tissues, etc.

Yes, right click on any figure you want to download, and select the image format you prefer. You can choose either .png or vectorized .svg format. All images are exported in high quality, making them suitable for presentations, publications, posters, and other types of sharing. You can also customize pre-set color theme – to do so go to Options tab visible in the left upper corner of Mavatar Discovery and select Theme.

We currently support the secure upload of gene lists and their regulation status (e.g., upregulated or downregulated labeled genes) for generating graphs. Learn how to upload a gene search file in our video tutorialvideo tutorial or 15. How do I search using a list of genes?x.

Depending on the complexity of the query, results can be generated in mere seconds, significantly reducing the time required for traditional analysis.

Conduct a new search by clicking the “Generate Graph” button in the top left-hand corner of the canvas, selecting a tissue, general or disease-specific network model. Submit your gene names, choose from our suggested gene sets, search by drug, or search for top interactions in the selected model.

Is the tissue / disease you are interested in not precalculated in Mavatar Discovery yet? Contact us and we will be happy to arrange a project together for a custom data survey and network construction.

The space where your network is generated is called the canvas – this is where you will explore & discover new connections between genes.

Floating on top of the canvas Mavatar Discovery presents cards with various categories of information. A card will always contain information about the network generated in the canvas – eg about the genes, edges or other information regarding what you see in the canvas.

At the top of the Generate graph window, Click "Top interactions" and choose the number of strongest edges you want to generate the network for. The graph will then be constructed based on the edges with the highest T-Values, which represent how reproducible the gene co-expression was found across included datasets (the higher T-Value the more consistent co-expression was observed).

After an initial network has been drawn, a histogram of the edge weights among the edges drawn can be seen in the filter card. Adjust the sliders to limit the edges further within this edge weight span.

We provide a description of datasets contributing to each network. Further information on datasets contributing to specific edges can be found in the DINA Functional Annotation card connected to the graph.

To explore drug-targets, click Generate Graph and search for your drug of interest using the Drugs search bar. All drugs and their associated targets listed in ChEMBL are currently available for querying, with plans to expand the drug library in future updates.

You can upload a gene list as a .csv, .txt, or Excel (.xlsx) file. A template is available for download from the File search bar or through this downloadable archive . Open the template and modify it to include your genes of interest. For details on input format and other specifications, refer to the README file included in the archive.

Once your file is ready, head back to Mavatar Discovery and upload your gene list. The graph will update just as it does when entering individual genes, generating a network view that reflects the regulation status of your queried genes.

When clicking a gene in the drawn network, this card provides a description and some synonyms of this gene, along with the gene type as defined by Ensembl.

Yes. Mavatar Discovery supports both symbol and Ensembl ID.

In the top right-hand corner of your screen, just above the canvas you have the option to save the current generated network. To find your saved searches, simply click the dropdown of saved searches in the top right-hand corner of the canvas.

The Patient Stratification Heatmap displays normalized and scaled expression levels from high throughput sequencing experiments contributing to the present network. Only genes appearing in the drawn network are included in the heatmap. In the drop-down menus, you can select which patient groups to include in the heatmap, as well as your preferred sample clustering approach. Genes, and by unsupervised clustering samples, are clustered by UPGMA method using Euclidean distance and samples are annotated based on our curated metadata.

The Cell Type Explorer displays coordinates in UMAP space, of cells from an example single-cell RNA sequencing dataset from the same tissue as the tissue of the current network. Cells can be colored by a gene’s expression level by clicking that gene node in the drawn network.

Drawing two networks at once (toggle “Secondary” in the generate graph modal) draws the edges from both networks, in different colors. This provides an overview of their overlap. In addition, any single network drawn will be compared to our other networks (using mantel tests on the node-node distances) – and the results visualised in the card “DINA Network Similarity”. This provides a measure of topology overlap for this gene set in pairs of networks.

The gene type filter gives you the option to filter your analysis by gene/transcript biotype. Classification provided by Ensembl (1x).

Yes, you can. We employ our pipeline to create a custom DINA resource for any combination of public domain, or user-provided datasets. This will allow a user with a novel dataset from any disease domain to be integrated and compared to existing datasets from the same disease domain or tissue context. The most seamless custom DINA resource integration requires RNA sequencing data (bulk or single cell), but our pipeline is compatible with proteomic or any quantitative data. Contact sales  for more information.

The graph structure generated will depend on the number of genes and clusters but can be modified manually by selecting and pulling the nodes in your desired directions (click and drag). If you wish to select and move a full cluster, right click on one gene in the cluster and “Select Local Cluster”. You can also select multiple nodes by holding the Control key (Mac) or Alt key (Windows) and clicking and dragging, or by selecting each node individually one at a time.

The network structure is automatically generated, but you can identify and move a full cluster by right clicking one gene within that cluster and "Select Local Cluster". Then pull the selected cluster away from the other.

The edge thickness represents the Normalized T value, where a higher correlation T value gives a thicker edge. Thicker edges therefore indicate greater confidence in the reproducibility of gene co-expression across the datasets used to train the network.

The query genes are the input genes from which your graph is generated. They can be a specifically defined set of genes, drug targets, or genes associated with GO terms, diseases, or included in other Mavatar Curated gene lists. Within the graph, these genes are highlighted in gray using default theme settings and tagged as Queried Gene within the Gene Information card.