Abstract
The identification of noncoding functional elements based on high-throughput genomic data remains an important open problem. Here we describe a novel computational approach for estimating the probability that a point mutation at each nucleotide position in a genome will influence organismal fitness. These fitness consequence (fitCons) scores can be interpreted as an evolution-based measure of potential genomic function. We first partition the genome into clusters of positions having distinct functional genomic “fingerprints,” based on cell-type-specific DNase-seq, RNA-seq, and histone modification data. Then we estimate the probability of fitness consequences for each cluster from associated patterns of genetic polymorphism and divergence using a recently developed probabilistic method called INSIGHT. We have generated fitCons scores for three human cell types based on publicly available genomic data and made them available as UCSC Genome Browser tracks. Like conventional evolutionary conservation scores, fitCons scores are clearly elevated in known coding and noncoding functional elements, but they show considerably better sensitivity than conservation scores for many noncoding elements. In addition, they perform exceptionally well in distinguishing ChIP-seq-supported transcription factor binding sites, expression quantitative trait loci, and predicted enhancers from putatively nonfunctional sequences. The fitCons scores indicate that 4.2–7.5% of nucleotide positions in the human genome have influenced fitness since the human-chimpanzee divergence. In contrast to several recent studies, they suggest that recent evolutionary turnover has had a relatively modest impact on the functional content of the genome. Our approach provides a unique new measure of genomic function that complements measures based on evolutionary conservation or functional genomics alone and is particularly well suited for characterizing turnover and evolutionary novelty.
