Fig. 1

a Schematic description of Hi-M microscopy: Chromatin is imaged through multiple acquisition cycles, each targeting a specific genomic locus using a set of unique DNA-FISH oligonucleotides targeted by a complementary, fluorescently labeled oligonucleotide. A fiducial marker is simultaneously imaged to allow for registration and drift correction during post-processing. Using pyHiM, the 3D conformation of the target locus is reconstructed for each individual cell. b pyHiM is an open-source project hosted on GitHub. Extensive documentation and Jupyter notebooks are available for users and developers. c pyHiM is developed in Python and runs indifferently on Linux, Windows and macOS. d Input data: 3D images are organized by imaging channel (DAPI, fiducial, DNA-FISH spots, etc.) and FOV. A single json file combines all parameters needed to run the analysis pipeline. e 3D images are pre-processed by calculating the maximum intensity projection and applying 2D registration based on the fiducial images. f Masks for nuclei, oligopaint libraries, and DNA-FISH spots are computed using pre-trained deep learning models. Individual DNA-FISH spots are localized with sub-pixel accuracy using apiFISH (fork of big-FISH). g Individual traces are built by combining the localizations of all DNA-FISH spots detected within the same mask. Results are saved in ECSV format. f Post-processing analyses are performed to obtain pairwise distance and proximity frequency matrices for each combination of DNA loci and for different spatial regions of the sample containing different cell types