Chromosomes are masters of organization. These long strings of DNA fold down into an ensemble of compact structures that keep needed parts of the genome accessible while tucking away those that aren't used as often. Understanding the complexity of these structures has been challenging; chromosomes are large systems, and deciphering the structure and dynamics requires a combination of experimental data and theoretical approaches. The FI-Chrom method, shared in a recent PNAS publication by Rice's José Onuchic and Vinícius Contessoto , is a new and effective approach for creating 3D maps of chromosomes from real-world data.
FI-Chrom uses data from chromosome Hi-C maps. These maps break out the chromosome into units of length called beads — about 50,0000 linear DNA bases each — and show how frequently each bead is close to other beads. This information shows only probabilities of any two beads being neighbors and no direct three-dimensional information. Imagine it as a logic puzzle where the rules, or parameters, read something like this: Bead A is 99% likely to be close to Bead B, 36% likely to be close to Bead C and 62% likely to be close to Bead D. A 3D model, the researchers knew, could be built by placing every bead in a space that didn't violate any of the Hi-C map's parameters. The only problem is that in Hi-C maps, there are hundreds of thousands of beads and tens of millions of mapped interactions showing bead closeness.
"We had chromosome maps that gave us, theoretically, 3D information, but we were really reading them in 2D space," explains Onuchic, the Harry C. and Olga K. Wiess Chair of Physics and a corresponding author of the study. "Now, we have created FI-Chrom, an open-access program that can turn these Hi-C maps into 3D models of chromosomes."
Antonio Oliveira Jr, a postdoctoral associate working with Onuchic and Contessoto and first author of the paper, turned to inverse statistical mechanics to solve this complex problem, building the FI-Chrom program. "FI-Chrom, which is available for any researcher to use, will provide us incredible insight into chromosome structure in any organism, from humans to yeast to bacteria," said Contessoto, assistant research professor of physics and astronomy.
While Oliveira used a framework known as maximum entropy, he didn't give FI-Chrom any prior instructions on what chromosomes should look like. Instead, he trained it on experimental data until FI-Chrom was able to produce models that matched the known Hi-C contact map.
"I have not, for example, told the program that human chromosomes generally separate into two compartments and minimize knots," Oliveira said. "Yet after multiple rounds of training, FI-Chrom began producing 3D models with these features. These physical features cannot be observed directly from the Hi-C maps but become clear from the 3D predicted ensemble of structures."
FI-Chrom also allows researchers to translate the Hi-C datasets into dynamic mechanisms. Although only structural information is included in Hi-C data, the researchers could model the different positions to infer movement because the 2D data show how often beads are near or far from each other.
"It is important to notice that the 2D Hi-C maps are obtained in a population of cells and do not encode a single chromosome structure but an ensemble of them," Contessoto said.
"We were able to use this dynamic Hi-C information to model chromatin loops," Onuchic said. "While we previously knew the loops existed, through the FI-Chrom model we were able to demonstrate that the loops form transiently rather than being static features of the chromosomes."
This research was supported by the Center for Theoretical Biological Physics and sponsored by the National Science Foundation (grants PHY-2019745, PHY-2014141 and PHY-2210291) and the Welch Foundation (grant C-1792).
