There are several methods for predicting the pathogenic effect of single-nucleotide variants (SNVs).1,2,3,4,5,6indels7 and other genomic changes, including epigenetic features. Predicting the pathogenic effect of variants in coding regions of the genome tends to be more accurate than predicting the effect of variants in non-coding regions, partly because more useful sources of data are available for the former. Available data may include measures that indicate a potential effect of a variant on protein structure or function, or a functional effect of a variant. For example, a variant may be classified as non-synonymous (where one amino acid is substituted) or synonymous (where the amino acid is not modified), or it may form a stop codon. . These types of variants would be expected to have significantly different effects from each other, making accurate prediction of pathogenic effects more accessible. Non-coding regions of the genome also contain a plurality of functional elements, such as enhancers, promoters, untranslated regions, splice sites and non-coding genes that express microRNAs, in addition to other sites that may be functional, such as pseudogenes. . Predictions for non-coding regions are often optimized for these elements—particularly regulatory elements that mediate transcription of non-coding RNAs (ncRNAs). writing now Nature Biomedical EngineeringChikashi Terao and co.8 Report a machine-learning model for cell-type-specific prediction of the effects of genomic mutations on the expression of ncRNAs.