Chromatin folding exhibits extensive cell-to-cell variability, yet whether this variability is dominated by continuously varying conformations or by recurrent organizational patterns remains unclear. Here, we combined high-resolution chromatin tracing with a novel machine learning algorithm to show that chromatin organization can be compactly represented by a limited repertoire of recurrent folding motifs. Using polymer simulations and cohesin depletion experiments, we further show that these motifs arise primarily from loop extrusion. Analysis of distinct tissues reveals that a shared motif repertoire is sufficient to describe diverse structural ensembles, with biological differences arising primarily through changes in motif occupancy and combinatorial usage. More generally, rather than generating entirely new conformations, transcriptional and disease-associated changes, including those occurring during disease onset, act primarily by redistributing motif occupancy within this common repertoire. In particular, gene activation is associated with a shift toward decompacted states and reduced contact frequencies, revealing a quantitative link between transcription and chromatin organization at the single-allele level. Together, these findings reveal that single-allele chromatin variability is structured by a constrained repertoire of recurrent folding motifs whose usage is modulated by regulatory context, providing a new framework to interpret how chromatin organization relates to gene activity.
