Figure 3.

The predictive power for loss-of-function phenotypes stems from a wide variety of data types integrated into HumanNet. Prediction both of (A) genes associated with mouse phenotypes and (B) of genes associated with human diseases are supported by diverse lines of evidence, including, for example, fly and worm data contributing strongly to mouse phenotypes, and yeast data contributing to human diseases. Grayscale indicates percentage contribution of a given data set's naive Bayes score to a phenotype's total AUC.