What am I looking at?
When the model predicts a gene’s expression in a tissue, it looks at other genes. Attention is how much it looks. Each edge here is one of the strongest of those look-ups in the aggregated {{ tissueName }} bulk RNA-seq profile; each node is a gene.
Cluster the strongest edges and the islands that appear are complexes and pathways: the ribosome, the proteasome, oxidative phosphorylation, complement. The model was never told any of that. White edges have pair annotations in the reference databases, matched through the encoded proteins and their UniProt accessions. Grey edges are unannotated relationships: model hypotheses that need validation. Edge details distinguish missing pair annotations from missing database coverage. Partners change between tissues: compare the same gene in brain and liver.
Clustersthe community map: top 3,000 edges, coloured by Leiden community, named by its best KEGG enrichment.
Networkthe raw graph; choose a preset of 500, 1,500, 3,000 or 5,000 strongest edges.
Double-clicka node, or search any of the {{ nProteins }} genes in this tissue, to see its 25 strongest partners and which databases already know them.
