In the post-genome era, disease-relevant gene finding and prioritization have focused on genome-wide association studies and molecular interaction networks, due to their power in characterizing the functions of genes/proteins in genomics and network biology contexts. In this paper, we describe a simple yet generic computational framework based on protein interaction networks to perform and evaluate disease gene-hunting, using colorectal cancer as a case study. We applied statistical measurements including specificity, sensitivity and Positive Predictive Value (PPV) to evaluate the performance of disease gene ranking methods, which we broke down into seed gene selection, protein interaction data quality and coverage, and network-based gene-ranking strategies. We discovered that best results may be obtained by using curated gene sets as seeds, applying protein interaction data set with high data coverage and decent quality, and adopting variants of local degree methods. ©2009 IEEE.