Machine Learning Puts a Bow on Enhancer-Promoter Predictions

A new computational method can read genomic signatures on ribbons of DNA that loop, bow-like, to facilitate interactions between distantly situated enhancers and promoters. The method, called TargetFinder, uses machine learning to distinguish between interacting and noninteracting enhancer–promoter pairs, and it has generated accurate predictions 85% of the time, suggesting that it can identify subtle gene regulation mechanisms—and thereby reveal new therapeutic targets for genetic disorders. TargetFinder, which was developed by scientists at the Gladstone Institutes, reconstructs regulatory landscapes from diverse features along the genome. According to a study that appeared April 4 in the journal Nature Genetics (“Enhancer–Promoter Interactions Are Encoded by Complex Genomic Signatures on Looping Chromatin”), TargetFinder was used to analyze hundreds of existing datasets from six different cell types to look for patterns in the genome…


Link to Full Article: Machine Learning Puts a Bow on Enhancer-Promoter Predictions