A new algorithm called EpiTensor can infer 3D genomic contacts from 1D maps of epigenomic data. The team analysed 16 histone modifications, DNase I sequencing and RNA sequencing data in five cell types to identify spatial patterns within topologically associating domains (TADs) at a resolution of 200 bp. Associations between promoters and enhancers, promoters and promoters, and between enhancers and enhancers that were identified by EpiTensor mirrored those determined using alternative approaches, such as Hi-C, ChIA-PET (chromatin interaction analysis by paired-end tag sequencing) or expression quantitative trait loci analyses. EpiTensor also detected interaction hotspots, which exhibited increased chromatin and transcriptional activity and were enriched for transcription factor and long non-coding RNA binding across cell types.