MDSPACE: Extracting Continuous Conformational Landscapes from Cryo-EM Single Particle Datasets Using 3D-to-2D Flexible Fitting based on Molecular Dynamics Simulation

MDTOMO: Continuous conformational variability analysis in cryo electron subtomogram data using flexible fitting based on Molecular Dynamics simulations

“…1 and Eq. 4 in Vuillemot et al (2023) 13 ). The weight given to the potential defined by the correlation coefficient, with respect to the classical MD potential, is a free parameter (the so-called force constant; k in Eq.…”

“…However, considering a relatively small number of available subtomograms, discrete classification methods are usually used with a small number of classes, which makes them ill-suited in the context of highly heterogeneous datasets composed of a continuum of conformational states. This heterogeneity issue, known as continuous conformational variability, is well-known in single particle (SPA) cryo electron microscopy (cryo-EM) and is currently an active field of research [7][8][9][10][11][12][13][14][15][16]17 . The number of methods for analyzing continuous conformational variability in SPA datasets has been growing over the last few years, but only a few such methods have been proposed for cryo-ET so far 18,19 .…”

“…HEMNMA-3D is another method for analyzing continuous conformational variability in cryo-ET subtomograms, which is able to use either a reference-volume voxel displacement (as TomoFlow) or a reference-structure atomic displacement to match the conformational states in subtomograms 19 . The development of HEMNMA-3D was inspired by the SPA methods that describe the continuous conformational variability by displacing atoms of an available molecular structure (reference structure) to match the conformational states in the data 7,12,13,16 . Such methods restrain the conformational exploration to the energy region that preserves the structural elements of the complex, which reduces the risk of overfitting low-SNR data.…”

“…Such methods restrain the conformational exploration to the energy region that preserves the structural elements of the complex, which reduces the risk of overfitting low-SNR data. They displace atoms of a reference structure using different methods to model the deformation field, such as normal mode analysis 7,12 , MD simulations 13,15 , and Zernike polynomials 16 . HEMNMA 7 and its deep-learning-based extension DeepHEMNMA 12 perform a flexible alignment of the reference structure to single particle images along the direction in the energy landscape that is defined by a set of vectors of collective motions, called normal modes.…”

“…However, the methods based on normal modes and Zernike polynomials restrict the estimation of the dynamics to a predefined direction in the energy landscape (the direction defined by normal modes or Zernike polynomials), which may not be sufficient to fully describe the conformational heterogeneity in the data. Contrary to these methods, a recently proposed SPA method named MDSPACE uses molecular dynamics (MD) simulations, in combination with normal modes, which allows a better exploration of the energy landscape 13 . For atomic displacements using a combination of normal modes and MD simulations, MDSPACE employs the algorithm named Normal Mode Molecular Dynamics (NMMD) 20 , which was initially proposed for flexible fitting of atomic structures into cryo-EM maps in the context of structural modeling.…”