Daniel MacArthur and colleagues report a catalog of validated loss-of-function (LoF) variants predicted to completely disrupt human protein-coding genes (Science 335, 823–828, 2012 ). They identified 2,951 candidate LoF variants based on an analysis of whole-genome sequences of 185 samples from the pilot phase of the 1000 Genomes Project. They then computationally filtered this candidate list by taking into account local sequence context, gene annotation and read mapping and quality. Extensive experimental validation was performed, including genotyping with Illumina arrays and custom Sequenom assays. After manual reannotation, they report a list of 1,285 high-confidence LoF variants. Of this list, 32% were found to affect only a subset of the transcripts for the relevant gene, suggesting that functional proteins may still be produced. They estimate that there are ∼100 LoF variants per human genome, with ∼20 in a homozygous state. Genes with at least one LoF variant show less evolutionary conservation and have more paralogs, suggesting possible redundancy. The authors also imputed 417 of these LoF variants into Wellcome Trust Case Control Consortium (WTCCC) case-control datasets and found no excess of association signals compared to other protein-coding variants, suggesting that these variants have a minor role in common disease.