Reconstructing the three-dimensional architecture of extrachromosomal DNA with ec3D

Extrachromosomal DNAs (ecDNAs) are large, acentric, circular DNA molecules that occur pervasively across many human cancers. EcDNA can drive tumor formation and evolution, contribute to drug resistance, and associate with poor patient survival outcomes. Beyond mediating high copy numbers, the circular topology and dynamic conformational changes of ecDNA disrupt topological domains and rewire regulatory networks, thereby conferring an important role in the transcriptional regulation of oncogenes. Here, we develop ec3D, a computational method for reconstructing the three-dimensional structures of ecDNA and analyzing significant interactions from high-throughput chromatin capture (Hi-C) data. Given a candidate ecDNA sequence and the corresponding whole-genome Hi-C as input, ec3D reconstructs the spatial structure of ecDNA by maximizing the Poisson likelihood of observed interactions. Ec3D's performance was validated using both simulated ecDNA structures with varying conformations, and Hi-C data from previously-characterized cancer cell lines. Our reconstructions reveal that ecDNAs occupy spherical configurations and mediate unique long-range interactions involved in gene regulation. Through algorithmic innovations, ec3D can resolve complex ecDNA structures with duplicated copies of large genomic segments, identify multi-way interactions, distinguish between interactions arising from direct spatial proximity and secondary interactions resulting from alternative folding patterns or intermolecular (trans) contacts of ecDNA molecules. Our findings provide insights into how the spatial organization of ecDNA may influence gene regulation and contribute to increased oncogene expression.