Recent technological advances have allowed us to study individual genomes at a base-pair resolution and have demonstrated that the average exome harbors more than 15,000 genetic variants. However, our ability to understand the biological significance of the identified variants and to connect these observed variants with phenotypes is limited. The first step in this process is to identify genetic variation that is likely to result in changes to protein structure and function, because detailed studies, either population based or functional, for each of the identified variants are not practicable. Therefore algorithms that yield valid predictions of a variant's functional significance are needed. Over the past decade, several programs have been developed to predict the probability that an observed sequence variant will have a deleterious effect on protein function. These algorithms range from empirical programs that classify using known biochemical properties to statistical algorithms trained using a variety of data sources, including sequence conservation data, biochemical properties, and functional data. Using data from the pilot3 study of the 1000 Genomes Project available through Genetic Analysis Workshop 17, we compared the results of four programs (SIFT, PolyPhen, MAPP, and VarioWatch) used to predict the functional relevance of variants in 101 genes. Analysis was conducted without knowledge of the simulation model. Agreement between programs was modest ranging from 59.4% to 71.4% and only 3.5% of variants were classified as deleterious and 10.9% as tolerated across all four programs.