Huge portions of the genomes of even the most well-studied organisms remain poorly understood. Accordingly, computational methods have attempted to predict functions and phenotypes associated with genes. Previously, our lab has constructed so-called functional networks--networks designed to predict the function of genes based on their similarity to other genes on a number of biological measures--and used these networks to show that even complex phenotypes such as longevity are relatively predictable. Here we take a new approach with different data, combining widely varied large publicly available gene-phenotype association datasets for human, mouse, worm, yeast, and more into a single inference procedure termed phenologs to predict phenotypes across species. In-silico validation results for predicting long-lived mutants indicates that this method shares similar promise in predicting even complex phenotypes. Experimental validation results for other predicted phenotypes have been highly encouraging, and lifespan experiments are planned.