Functionalities

how can i distinguish the up and downregulated genes? mavatar discovery does not directly label genes as upregulated or downregulated, because the platform is built on co expression and correlation rather than differential expression however, the network structure itself encodes expression directionality in ways that can be leveraged with the right approach here we give you some tips to help you interpret your results note when uploading a gene list file, you can label the regulation of your genes with 1 (upregulated) and 1 (downregulated), or 0 (not deregulated) to highlight them in different colors in your graph you can modify the theme colors in options – theme gene expression behavior is shown in co expression clusters genes that tend to be upregulated together in each condition will appear correlated and clustered together in the network most of the times likewise, genes that are co downregulated tend to form their own clusters this means the modular structure of your graph already carries information about coordinated expression changes genes within the same tightly connected cluster tend to behave similar across the samples that built the network, whether the shared behavior is up or downregulation nevertheless, context changes what correlated means two genes might appear strongly correlated in a general network but show no meaningful connection in a disease specific network, and vice versa this shift is revealing something important about the biology and molecular topology of your disease of interest the co expression relationship is condition dependent, suggesting that the coordinated regulation of those genes is driven by the disease state rather than by the baseline tissue biology this is a more complex and robust signal than a simple up/down label, because it captures the rewiring of regulatory relationships in disease you can use the dual network graph generation to make this visible the platform allows you to plot two networks simultaneously for instance, a general brain network and an alzheimer’s specific brain network by overlaying these, you can directly compare which gene gene correlations are present in both contexts, and which appear only in the disease network edges that exist exclusively in the disease specific network represent interactions that emerge under disease conditions (i e , genes involved are likely co regulated (either co upregulated or co downregulated) in a disease dependent manner) edges that are present in the general network but absent in the disease network may reflect healthy regulatory programs that are disrupted in disease this comparative approach gives you a functional readout of expression directionality without needing explicit fold change values to confirm the direction of change for individual genes, the conditions expression chart lets you examine expression levels across disease versus healthy conditions if you identify a cluster of co correlated genes in your disease specific network and then check each in the conditions expression chart, you can determine whether that cluster represents a group of co upregulated or co downregulated genes in your disease of interest what kind of information can i get from querying two networks at the same time? what does it mean? dual network querying is one of the most powerful features in mavatar discovery it allows you to overlay two networks simultaneously and compare their topologies side by side, revealing differences in gene interaction topology that would be invisible if you examined each network in isolation the biological insight you gain depends on which two networks you choose to compare same tissue, different conditions this is the most common use case for disease mechanism exploration by overlaying a general network and a disease specific network for the same tissue, for example, a general brain network and an alzheimer's specific brain network, you can see which gene interactions are present across all conditions and which emerge only in disease edges that appear exclusively in the disease specific network represent co expressions that are driven by the disease state, meaning those genes are co regulated in a way that doesn't occur under normal physiological conditions conversely, edges present in the general network but absent in the disease network may reflect healthy regulatory programs that are disrupted or lost in disease this comparison gives you a direct view of how the regulatory architecture rewires in disease, going beyond simple gene lists to capture changes in the relationships between genes different tissues, same disease if your disease of interest affects multiple organs or tissue compartments, comparing networks across tissues can reveal which molecular mechanisms are tissue specific and which are shared for instance, overlaying a brain network and a blood network for the same disease could show you whether the inflammatory signaling module you observe in the brain is also detectable in peripheral blood, which has direct implications for biomarker accessibility interactions that appear in both tissues may reflect systemic disease biology, while those unique to one tissue highlight compartment specific mechanisms that would require tissue specific therapeutic approaches linear versus ai non linear networks this comparison unlocks an entirely different layer of biological information standard networks in the platform are built on linear correlation methods, which capture proportional co expression relationships the ai networks, by contrast, detect non linear relationships where the connection between two genes follows a more complex pattern by overlaying a linear and an ai network for the same tissue, you can categorize interactions into three groups those captured by both methods (robust linear relationships that also hold non linearly), those present only in the linear network (straightforward correlation that doesn't extend to more complex modeling), and those present only in the ai network (non linear interactions that conventional statistical approaches would never detect) this last category, which consists of regulatory relationships that exist in the data but are invisible to standard correlation analysis, is particularly valuable for discovering novel biologal connections cell type versus tissue level perspectives if a cell type specific network is available for your tissue of interest, overlaying it with the broader tissue network lets you determine which interactions in the tissue level graph are driven by a specific cell population for example, comparing a macrophage/monocyte network with a general blood network could reveal that a particular inflammatory module in the tissue level graph is almost entirely attributable to macrophage specific gene regulation, or alternatively, that it reflects a coordinated signal across multiple cell types general versus general across tissues comparing two general networks from different tissues, for example, brain versus liver, can reveal the tissue specificity of gene interactions, independent of any disease this is useful for understanding whether a gene pair you're interested in interacts broadly across the body or whether its co regulation is restricted to a specific organ, which has implications for off target effects and tissue specific therapeutic strategies this is also useful for the study of systemic diseases that might affect more than one organ expression directionality through network comparison dual network comparison also provides an indirect but powerful way to infer expression directionality genes that are tightly correlated in a disease specific network but uncorrelated in the general network are likely co regulated, either co upregulated or co downregulated, specifically under disease conditions pairing this observation with the conditions expression chart for individual genes lets you confirm the direction and build a complete picture of coordinated expression changes in disease what would be the best way to initiate an analysis? what would a workflow for discovery look like? should i start with general networks or disease specific networks and why? what parameters would you include? distance and neighbors, tissue choice, etc getting the most out of mavatar discovery depends on thoughtful decisions about where to start and how to progressively layer on analytical depth here's a recommended workflow that moves from initial exploration to more in depth insights choose your starting network model begin with a general network in your tissue of interest general networks integrate data from all available conditions different diseases, healthy controls, infections, and more, which gives you the broadest view of how your genes interact in that tissue context this is important as a first step because it lets you see the full landscape before narrowing down you can observe which conditions contribute to each interaction and whether your genes of interest form connections that are broadly present or condition specific after this initial exploration, you can move into a disease specific network within the same tissue to see how the interaction structure changes when the data is restricted to your condition of interest if an ai derived network is available for your tissue, setting it as a second network layer adds an additional dimension the ai networks capture non linear correlations that conventional linear statistical methods would miss, so comparing the two can reveal interactions that only emerge through non linear modeling set your graph parameters to balance focus and discovery when generating your initial graph, start with a short distance (e g , 1 or 2) which keeps the network tightly centered around your queried genes and their closest interactions set the number of neighbors to 5 or 6 this combination gives you a manageable, interpretable graph that emphasizes the most direct and significant interactions without overwhelming you with distant or weakly connected genes you can always expand later by increasing distance or neighbors once you understand the core structure around your input genes explore how your genes relate to each other and to their context once your graph is generated, the first question to ask is whether your genes of interest are direct neighbors in the network or separated by intermediary genes this tells you whether they are part of the same tightly co regulated module or connected through shared regulatory mechanisms from there, examine the functional annotation card to see which datasets and disease cohorts contribute to the edges in your graph, revealing whether the interactions you're seeing are driven by your condition of interest or by unrelated biology then move to the conditions expression chart to evaluate how each gene behaves across conditions from rna seq studies is its expression specific to your disease, or is it broadly altered across multiple conditions? apply the same question to neighboring genes if a neighbor shows strong disease specificity alongside your seed gene, it becomes a much more interesting candidate if an ai network is available, compare linear and non linear interactions setting the ai network as your second network allows you to identify correlations that exist in both the standard (linear) and ai (non linear) networks, as well as interactions that appear only in the ai network these non linear only interactions are particularly valuable because they represent relationships that conventional statistical approaches would never surface, potentially uncovering regulatory mechanisms that depend on other complex mechanism identify enriched pathways and functional modules use the functional enrichment card to determine which biological pathways and processes are overrepresented among your network genes you can run enrichment on the full graph to get a broad functional overview or select a specific cluster or module within the network to interrogate its biological role in isolation this step transforms a visual network into an interpretable set of biological functions and helps you prioritize which regions of the graph are most relevant to your research question expand your query and layer on translational context once you've characterized the core network around your initial gene set, consider modifying your query to include additional genes (e g , known disease markers, genes with established mutations in your condition, or genes known to be deregulated) adding these to your network can reveal previously unseen links between your original hits and established disease biology, helping you identify which of your candidates sit closest to known mechanistic nodes and which might represent novel connections worth pursuing at this stage, you can also overlay drug target information to assess which genes in your network are pharmacologically tractable, closing the loop from discovery to potential therapeutic relevance in summary , a productive workflow moves through six defined stages start broad in a general network with focused parameters, explore gene relationships and condition specificity, compare linear and non linear interaction layers, identify enriched functions, then iteratively expand your query and add translational overlays each step builds on the previous one, progressively sharpening your understanding of the biology encoded in your data once i see a network, what is next? apart from this graph, is there anything else i can do or visualize? is there any other functionality that i need to be aware of? what else can i do beyond the graph? the network on your screen is just the starting point what you're looking at is a visual representation of gene interactions, but underneath it sits a robust layer of analytical tools designed to let you interrogate, reframe, and build on that network in multiple directions without writing any code some functionalities to be aware of explore expression patterns across patients and conditions the patient stratification heatmap lets you visualize how the genes in your network behave across different patient groups and conditions clustering on both the sample and gene axes reveals which genes share expression profiles and which patient subgroups drive those patterns you can select specific cohorts to zoom into the comparisons that matter most to your research disease versus control, treatment responders versus non responders, or any other grouping available in the data examine cell type specific expression profiles the cell type explorer allows you to explore how your network genes are expressed at the single cell level, showing you whether a gene or set of genes is concentrated in specific cell types this adds cellular resolution to what is otherwise a tissue level network and can help you determine, for instance, whether a hub gene in your network is primarily active in t cells, b cells, or macrophages and monocytes evaluate condition specificity the conditions expression chart lets you assess how individual genes perform across disease categories, drawing on all the rna seq data contributing to your network this is where you test whether a gene's behavior is truly specific to your condition of interest or broadly altered across many diseases annotate interactions with biological and dataset context the functional annotation card reveals which datasets, diseases, and sample cohorts contribute to each edge in your graph, giving you more insight for every interaction the functional enrichment card identifies overrepresented pathways and biological processes among your network genes, either across the full graph or within a selected subset identify drug targets overlaying chembl drug target information highlights which genes in your network are targeted by existing compounds, connecting your biological findings to drug repurposing opportunities iterate, refine and save your query you're never locked into your initial graph you can expand your network by adding new genes of interest, simplify it by filtering edges or removing core genes, increase or decrease distance and neighbor parameters, or switch between general, disease specific, and ai networks to see how the interaction landscape changes under different data contexts each modification gives you a different lens on the same biology, and you can save each query independently to come back to your findings later what are network core genes? and if i hide them, does this mean that the genes that remain are more specific for the disease i am studying? network core genes are genes whose wiring patterns are conserved across multiple networks they show significantly similar connectivity regardless of tissue or condition think of them as the molecular infrastructure that is maintained across biological contexts pan tissue housekeeping pathways, fundamental cellular processes, and evolutionarily conserved modules when you hide them, the remaining genes and interactions are enriched for context specific biology, meaning they are more likely to reflect the distinctive wiring of your particular network however, whether that means "disease specific" depends on which network you're working in if you're in a disease specific network, then yes, removing core genes will emphasize the interactions and modules that are unique to that disease context if you're in a general tissue network, hiding core genes will surface tissue specific biology, but that may include signal from multiple conditions and not just one disease to identify the strongest disease specific signals, it is recommended to combine core gene filtering with a disease specific network that way you're both restricting the data context to your disease and removing the conserved wiring that would otherwise dilute the condition specific patterns what are mavatar curated gene lists? how can i explore them or include them in my workflow? mavatar curated gene lists are pre calculated, context specific gene signatures derived entirely from network topology each list captures genes whose connectivity patterns within a specific network differ significantly from their wiring in other networks, meaning these genes are differentially wired in that tissue or condition they are identified through iterative subnetwork comparisons using the deltacon method across all available networks in practical terms, these are ready to use gene sets that represent the distinctive molecular fingerprint of each network context you can use them to generate network specific graphs directly, giving you a starting point that is already enriched for the biology that makes that context unique you can also run functional enrichment to identify which pathways and processes are captured by the signature, explore the genes in the conditions expression chart to assess their expression profiles across diseases, or overlay drug target information to evaluate pharmacological relevance they are also useful as a benchmarking tool for your own data if you have a list of differentially expressed genes from your own experiment or even your generated graph from your genes of interest, you can compare it against the mavatar curated gene list for the relevant context to see how much overlap exists using the functional enrichment – mavatar curated list function