Then, while querying the database and reviewing the graphical output, you will soon realize that the number of available links for your query is usually too large. This is indicated with a message at the result web page: "Note that more links were found at this confidence. They were removed...". Essentially, neither the browser could visualize them, nor a human being would be able to grasp the displayed complexity. For this reason, a number of built-in procedures is trying to anticipate this overload and pre-filter the retrieved network -- hence the message. However, the procedures have a common feature: the filtering is confidence-driven and is not aware of your research focus. As a result, hundreds or even thousands of potentially important links are not shown. The following measures may help you re-focus the query around the gene(s) of your interest (let's call it X):

• Add some pertinent genes, or group IDs (GO category, KEGG pathway, OMIM disease -- pick these from the area just below the "Run" button) to the yellow box - the default algorithm (1) will prioritize direct links between them and X.
• Add such identifiers to the magenta box. The search will finish by looking for any links between its content and the already found network environment of X.
• Try to increase/decrease the confidence score threshhold (values 0.02...0.99 from the red select box at "Sub-network reduction")
• Try ARACNE: algorithms (3) and (4) at "Sub-network reduction". At the cost of re-wiring the network, it shows all found genes.
• Redefine the evidence that controls querying with the box "Base network links on evidence". It may help if you are e.g. sceptical about miRNA, or phylogenetic profiling, or fungal+plant data etc. Note that discarded evidence categories are ignored rather than prohibited, i.e. you may still see respective lines on the screen.
• Upon receiving the result page, you can use checkboxes of jSquid applet -- specified links will disappear. But you change only the visual representation of links. The difference with the previous method is that they have influenced the sub-network retrieval.
• Employ the multi-species view at "Show sub-network(s) in". Seeing a cross-species conservation pattern can be useful.
• Note that setting a too strict confidence cut-off would eventually leave you with commonplace textbook sub-networks and no room for your future discovery!