The tendency for a human gene to be essential correlates well with the number of its protein-interaction partners, suggesting that essential human genes can be identified directly from protein-interaction networks. (a) For a set of around 31,000 human protein interactions , the number of interactions per protein (the 'degree') is plotted for 907 essential vertebrate proteins known from mouse knockouts , human small interfering (si)RNA screens [51,52], and zebrafish random mutagenesis  and for the remaining 6,661 proteins in the network, considering only the largest connected network component. (b) The likelihood of being essential increases with increasing degree. Proteins were sorted by degree and divided into bins of 100 proteins each (filled diamonds). The observed frequency of essential genes in the bin is plotted against the average degree of the proteins in the bin, showing high correlation (R2 = 0.78) between degree and essentiality.