Annotation of metabolites remains a major challenge in liquid chromatography-mass spectrometry (LC-MS) based untargeted metabolomics. The current gold standard for metabolite identification is to match the detected feature with an authentic standard analyzed on the same equipment and using the same method as the experimental samples. However, there are substantial practical challenges in applying this approach to large data sets. One widely used annotation approach is to search spectral libraries in reference databases for matching metabolites; however, this approach is limited by the incomplete coverage of these libraries. An alternative computational approach is to match the detected features to candidate chemical structures based on their mass and predicted fragmentation pattern. Unfortunately, both of these approaches can match multiple identities with a single feature. Another issue is that annotations from different tools often disagree. This paper presents a novel LC-MS data annotation method, termed Biologically Consistent Annotation (BioCAn), that combines the results from database searches and in silico fragmentation analyses and places these results into a relevant biological context for the sample as captured by a metabolic model. We demonstrate the utility of this approach through an analysis of CHO cell samples. The performance of BioCAn is evaluated against several currently available annotation tools, and the accuracy of BioCAn annotations is verified using high-purity analytical standards.