ENMs at Multiple Scales: From Proteins to Chromosomal Dynamics, to AIpowered Pathogenicity Predictions

The past 30 years have shown the utility of Elastic network models (ENMs) in improving our

understanding of the coupled dynamics of biomolecules, from individual proteins to

supramolecular systems. ENMs also proved useful in hybrid models that permit us to visualize

the large-scale cooperative events at atomic detail. The global motions predicted by ENMs

have proven in numerous applications to provide a good description of molecular machinery

and allosteric behavior, opening the way to designing allosteric modulators of protein function.

Application to supramolecular structures, including cryo-EM structures, has been a major

utility. A major advantage of ENMs is their simplicity and computational efficiency, which

enables proteome-scale analyses, applications to large systems such as the entire chromatin,

and/or combination with machine learning (ML) algorithms. Such a recent ML approach that

incorporates ENM predictions in addition to sequence and structure data proved to yield an

accurate assessment of the effect of mutations on function, compared to those based on

sequence and structure exclusively. The use of AlphaFold database further improved the

accuracy of the predictions by providing access to a large structural dataset. The developed

tool, Rhapsody-2, generates in silico saturation mutagenesis maps for any query structure for

which structures (or structural models are available. Another recent adaptation to modeling

human chromosomal 3D dynamics showed the close correspondence between the spatial

mobilities of gene loci and the expression levels of the corresponding genes. These recent

developments and future directions will be discussed.

This work was supported by NIH award R01 GM139297