Rare disorder analysis has been facilitated by the declining cost of sequencing. Despite numerous success stories the underlying cause remains unexplained in >50% cases, emphasizing the need for new discovery tools interrogating both coding and non-coding regions. We introduce Phen-Gen, which combines patients disease symptoms and sequencing data with prior domain knowledge in a holistic prediction to identify the causative gene(s). Phen-Gen works with both exome and whole genome sequencing data and has been evaluated in extensive simulations as well as real datasets.