A Statistical Approach to Infer 3d Chromatin Structure

We propose a new algorithm to estimate the 3d configuration of a chromatin chain from the contact frequency data provided by hi-c experiments. Since the data originate from a population of cells, we rather aim at obtaining a set of structures that are compatible with both the data and our prior knowledge. Our method overcomes some drawbacks presented by other state-of-the-art methods, including the problems related to the translation of contact frequencies into Euclidean distances. Indeed, such a translation always produces a geometrically inconsistent distance set. Our multiscale chromatin model and our probabilistic solution approach allow us to partition the problem, thus speeding up the solution, to include suitable constraints, and to get multiple feasible structures. Moreover, the density function we use to sample the solution space does not require any translation from contact frequencies into distances.