A myriad of noncoding genetic association signals are now awaiting the identification of causal alleles and their functional interpretation. We introduce the novel computational framework PINES (Phenotype-Informed Noncoding Element Scoring), which evaluates the functional impact of noncoding variants by integrating diverse epigenetic annotations. A unique feature of PINES is that it directs the analysis towards genomic annotations most relevant to phenotypes of interest. We show that PINES identifies functional noncoding variation more accurately than methods that do not use phenotype-specific knowledge. We apply PINES to fine map noncoding alleles at GWAS loci across a range of diseases, and predict new causal risk alleles for Parkinson's disease and inflammatory bowel disease. We also use PINES to confirm several high-penetrance variants implicated in Mendelian traits, as well as variants residing within known enhancer regions. PINES consistently identifies functional variants in fine mapping analyses, dissecting pathogenic loci while avoiding the resource-intensive traditional fine mapping studies. Due to its flexibility and ease of use through a dedicated web portal, PINES provides a powerful in silico method to prioritize and fine map functional noncoding variants.