Next-generation sequencing (NGS) is transitioning from being a research tool to being used in routine genetic diagnostics, where a major challenge is distinguishing which of many sequence variants in an individual are truly pathogenic. We describe some limitations of in silico analyses of NGS data that emphasize the need for experimental confirmation. Using NGS, we recently identified an apparently homozygous missense mutation in NUBPL in a patient with mitochondrial complex I deficiency. Causality was established via lentiviral correction studies with wild-type NUBPL cDNA. NGS data, however, provided an incomplete understanding of the genetic abnormality. We show that the maternal allele carries an unbalanced inversion, while the paternal allele carries a branch-site mutation in addition to the missense mutation. We demonstrate that the branch-site mutation, which is present in approximately one of 120 control chromosomes, likely contributes to pathogenicity and may be one of the most common autosomal mutations causing mitochondrial dysfunction. Had these analyses not been performed following NGS, the original missense mutation may be incorrectly annotated as pathogenic and a potentially common pathogenic variant not detected. It is important that locus-specific databases contain accurate information on pathogenic variation. NGS data, therefore, require rigorous experimental follow-up to confirm mutation pathogenicity. Hum Mutat 33:411–418, 2012. © 2011 Wiley Periodicals, Inc.