A method that analyzes quality control metrics from multi-sample genomic sequencing studies and nominates poor quality samples for exclusion. Per sample quality control data are transformed into z-scores and aggregated. The distribution of aggregated z-scores are modelled using parametric distributions. The parameters of the optimal model, selected either by goodness-of-fit statistics or user-designation, are used for outlier nomination. Two implementations of the Cosine Similarity Outlier Detection algorithm are provided with flexible parameters for dataset customization.