Background:Long-range chromatin interactions are spatially and temporally maintained and established through a set of protein complexes.Various efforts have been made to elucidate these proteins;however, many are still unknown.Here, we present 3CPET, a tool based on a nonparametric Bayesian approach, to infer the set of the most probable cofactor proteins involved in maintaining chromatin interactions,making it a valuable downstream analysis tool in chromatin conformation studies.Results:We developed 3CPET, a method based on a non-parametric hierarchical Dirichlet process, to infer the set of protein networks involved in maintaining chromatin interactions using ChIA-PET data,transcription factor binding sites, and protein interactions.When tested on ER-α ChIA-PET associated interactions, 3CPET was able to predict known ER-α co-factors and showed a significant overlap with previously-reported experimental results.Similarly, using the RNA polymerase-Ⅱ associated interactions, we were able to predict cofactors known to be part of known transcription-related complexes such as the Mediator.3CPET associated the predicted networks to different loci with different expression and functional enrichment patterns.Simulation studies showed that 3CPET has a true positive rate of 60%-85% when compared with simulated knockdown data.Conclusions:A better characterization of the proteins establishing the chromatin architecture can help shed some light on the way transcription is regulated and may help us distinguish new biomarkers.3CPET presents a valuable tool toward this step, as it can infer a list of candidate chromatin maintaining networks and the regions they may control.