Dissecting the co-segregation probability from genome architecture mapping

Abstract

The genome architecture mapping (GAM) is a recently developed methodology that offers the co-segregation probability of two genomic segments from an ensemble of thinly sliced nuclear profiles, enabling to probe and decipher the 3D chromatin organization. The co-segregation probability from GAM, which typically probes the length scale associated with the genomic separation greater than 1 MB, is, however, not identical to the contact probability obtained in Hi-C, and its correlation with inter-locus distance measured with FISH is not so good as the contact probability. In this study, by using a polymer-based model of chromatins, we derive a theoretical expression of the co-segregation probability as well as that of the contact probability, and carry out quantitative analyses of how they differ from each other. The results from our study, validated with in-silico GAM analysis on 3D genome structures from FISH, suggest that to attain strong correlation with the inter-locus distance, a properly normalized version of co-segregation probability needs to be calculated based on a large number of nuclear slices (n > 103). SIGNIFICANCEBy leveraging a polymer model of chromatin, we critically assess the utility of co-segregation probability captured from GAM analysis. Our polymer model, which offers analytical expressions for the co-segregation probability as well as for the contact probability and inter-locus distance, enables quantitative comparison between the data from GAM, Hi-C, and FISH. Although the plain co-segregation probabilities from GAM are not well correlated with inter-locus distances measured from FISH, properly normalized versions of the probability calculated from a large number of nuclear profiles can still reasonably represent the inter-locus distance. Our study offers instructions of how to take full advantage of GAM analysis in deciphering 3D genome organization.