Recent efforts to characterize noncoding sequences using high-throughput biochemical assays have produced a wealth of data

Many of these nucleotides appear to fall in cis-regulatory elements

In conceptual terms, the new LINSIGHT method is closely related to our previous fitCons method

Manygeneticvariantsthatinfluencephenotypesofinterestarelocatedoutsideofprotein-codinggenes,yetexistingmethodsforidentifyingsuchvariantshavepoorpredictivepower.Hereweintroduceanewcomputationalmethod,calledLINSIGHT,thatsubstantiallyimprovesthepredictionofnoncodingnucleotidesitesatwhichmutationsarelikelytohavedeleteriousfitnessconsequences,andwhich,therefore,arelikelytobephenotypicallyimportant.LINSIGHTcombinesageneralizedlinearmodelforfunctionalgenomicdatawithaprobabilisticmodelofmolecularevolution.Themethodisfastandhighlyscalable,enablingittoexploitthe‘bigdata’availableinmoderngenomics.WeshowthatLINSIGHToutperformsthebestavailablemethodsinidentifyinghumannoncodingvariantsassociatedwithinheriteddiseases.Inaddition,weapplyLINSIGHTtoanatlasofhumanenhancersandshowthatthefitnessconsequencesatenhancersdependoncelltype,tissuespecificity,andconstraintsatassociatedpromoters.