Definition of loops for interaction contacts

• Loop definition presented by Suhas Rao
  • Loops in chromatin will show a peak in a contact map
  • However, other features (such as domain/compartment structures) may interfere and needs to be accounted for in the model
  • If loops can be removed by deleting one of the interacting elements, it can be called a “validated loop”
• Loops and local quantum domains
  • When one of the motifs of a loop is deleted, the peak will disappear, however, in some cases the domain boundary can still be maintained (see slide “validated loops”, right middle panel). So it might be possible to distinguish these two features.
• Dependency on the experimental methodology because of usage of interaction levels of local neighbors?
  • The definition heavily depends on local interactions around the loop point. However, such mathematical definitions should be technological independent, therefore, it may need to be applicable to other types of data as well.
  • For example, ChIA-PET, although calling enriched interactions, does not provide local interaction data as HiC do, so how loops can be defined from that? For FISH it may become more challenging to detect neighboring contact frequencies.
  • Maybe not every technology is suitable to detect any features, so some kind of limit may still be needed. HiC is suitable for loops, other technologies might be not suitable. Like enhancers, which cannot be detected by every technology.
• Alternative / competing definitions for loops?
  • There are also other types of relationships between chromatins, like domain (large continuous segments of the genome with higher contact frequencies), compartment (large discontinuous segments of the genome with higher contact frequencies), etc. and not everything should be a loop.
  • The lack of a definition of a loop is having negative impacts on the field and most of the field accepted that a loop is something with increased contact frequency, however, what is the baseline for this increase is not clearly defined. This model uses local contact levels to clarify that.
  • So the point is how to see the local contact frequency around the pairs of interest. Different technologies may have different takes for that.
• Local environment for the loops? Will the removal of local CTCF boundaries near the loop eliminate the loop as well?
  • The causal motif is important for the loop but all other factors affecting the loop may need to be investigated
• Resolution of “local neighborhood” of the loop, is it a consensus for how wide this “local neighborhood” should be defined?
  • In practice this width is flexible so there is not a fixed answer to that yet.
  • It might be dependent to the experiment technology as well
  • The current case is a demonstration and it should be further defined in the future that how large this neighborhood should be.
• Does every local maxima correspond to a loop in HiC?
  • There might be some local maximas from noises in the experiment so there should be other ways of verification (for example, if we go into extreme high resolution, every contact will become a “local maxima”)
  • Replicate experiments may help however a rigorous statistical model should also be good to distinguish.
• Higher dimensions for loops (3-point contact and more) or “hubs” can also be defined in a similar way.