Exploring protein functions from structural flexibility using CABS-flex modeling

Abstract

Understanding protein function often necessitates characterizing the flexibility of protein structures. However, simulating protein flexibility poses significant challenges due to the complex dynamics of protein systems, requiring extensive computational resources and accurate modeling techniques. In response to these challenges, the CABS-flex method has been developed as an efficient modeling tool that combines coarse-grained simulations with all-atom detail. Available both as a web server and a standalone package, CABS-flex is dedicated to a wide range of users. The web server version offers an accessible interface for straightforward tasks, while the standalone command-line program is designed for advanced users, providing additional features, analytical tools, and support for handling large systems. This paper examines the application of CABS-flex across various structure-function studies, facilitating investigations into the interplay among protein structure, dynamics, and function in diverse research fields. We present an overview of the current status of the CABS-flex methodology, highlighting its recent advancements, practical applications, and forthcoming challenges.

Keywords: Monte Carlo simulation; coarse‐grained model; integrative modeling; molecular modeling; multiscale simulation; protein aggregation; protein flexibility.

FIGURE 1

CABS‐flex pipeline in default mode. The input protein structure serves as the starting point for the simulation and defines distance restraints based on either a default or a selected scheme. Optional inputs may include user‐defined distance restraints. The simulation employs a coarse‐grained CABS model. The protein dynamics trajectory is analyzed and clustered into 10 representative models, each capturing the essential variability of its cluster. These representative models are then reconstructed into all‐atom models. The outputs of this process include PDB structures of the representative models, fluctuation profiles (illustrating RMSF vs. amino acid residue number), and contact maps derived from the trajectory. RMSF, root mean square fluctuation.

FIGURE 2

Distance restraints imposed on the protein (PDB ID: 2gb1) in the CABS‐flex default modes SS1 (upper panel) and SS2 (lower panel). The left panels show the complete network of restraints shown as green and orange sticks, respectively for SS1 and SS2. In the middle, the restraints for a single residue are shown. The right panels show 10 protein models representing clusters of structures collected during one CABS‐flex run with given default parameters. One can observe here a difference in the movement of the part depicted in the middle panel resulting from a different number of restraints.

FIGURE 3

Comparison of flexibility profiles computed from the RCI and obtained from CABS‐flex simulations. Top: PDB ID 2ct5 NMR model 3 and AF2 model (UniProt: O96006). Bottom: PDB ID 2y4q NMR model 3 and AF2 model (UniProt: Q13563). In the right panels, the color and the thickness of the tube representation are proportional to the flexibility of each residue. RCI, random coil index.

FIGURE 4

The SARS‐CoV‐2 spike (S) protein in (a) “up” conformation (PDB ID: 7a98) bound to angiotensin‐converting enzyme 2 (ACE2) and in (b) “down” conformation (PDB ID: 7a98) which is without binding to the receptor. The N‐terminal S1 subunit consists of NTD (residues 14–306) in blue, RBD (residues 331–528) in tv_red, CTD1 (residues 529–591) in yellow, and CTD2 (residues 592–686) in cyan. The remaining residues form the S2 subunit and are shown in raspberry. (c) The interaction between S and ACE2 happens through the residues shown in light magenta on the RBD and wheat on the ACE2 receptor. The residues identified from CABS‐flex trajectories as (d) highly flexible and mutations could have significant effects on the evolution of the virus (Sanyal et al., 2022) and (e) have a significant stabilization effect on S‐ACE2 complex (Mishra et al., 2022) are highlighted in sky blue. SARS‐CoV‐2, severe acute respiratory syndrome coronavirus 2.

FIGURE 5

Aggrescan3D “dynamic mode.” (a) A3D pipeline: the input structure's flexibility is simulated using CABS‐flex to generate a set of predicted models that reflect the varying conformations of the structure; these models are then analyzed and scored based on their aggregation propensity, and the model with the highest aggregation propensity is presented as the final prediction. (b) Human p53 DNA‐binding domain (PDB ID: 2fej) aggregation propensity predictions using A3D “static mode” (left side) and “dynamic mode” (right side). Residues are colored in A3D scale (blue—solubility; red—aggregation). Aggregation prone regions are encircled. A3D, Aggrescan3D.

FIGURE 6

CABS‐flex ensembles provided valuable insights into the structural dynamics of S‐nitrosylated proteins. (a) The workflow utilizes SNOfinder to identify proteins susceptible to S‐nitrosylation, generates a structural ensemble of 20 models per protein with CABS‐flex to highlight their flexibility, and analyzes these models with SNOmodel to calculate key structural and biochemical parameters, thereby enhancing understanding of S‐nitrosylation mechanisms. (b) The cartoon representations show the models from the CABS‐flex ensemble (white) of the mitochondrial chaperone TRAP1 (deep teal). The stick and ball representations highlight the conformational dynamics of SNO cysteine (SNO, red), proximal cysteine (proximal, marine) of the ensembles, and the Sγ‐Sγ distances between them.

FIGURE 7

(a) Example CABS‐flex predictions (best prediction out of 10,000 simulated structures—blue; best prediction out of top 10 structures ranked by CABS‐flex—yellow) superimposed on the experimental PDB structures (red). (b) Comparison between CABS‐flex predictions (yellow) and AlphaFold2 predictions (cyan) superimposed on the experimental PDB structures (red).

FIGURE 8

CABS‐flex output structures rebuilt by two methods: (a) Modeller in cyan and (b) cg2all⁷² in dark green, superimposed on the experimental PDB structure in orange (PDB ID: 3oou). We highlight the accuracy of side‐chain reconstruction (left panels) and helical hydrogen bond networks (right panels) in the rebuilt structures.

FIGURE 9

Examples of protein structure fluctuations from CABS‐flex and molecular dynamics (MD) runs, and pLDDT confidence scores for proteins: (a) 6o2v_A and (b) 1d3y_B. MD runs were retrieved from the ATLAS database (Vander Meersche et al., 2024). On the right, the figure shows pLDDT confidence scores visualized on the protein structures. pLDDT, predicted local distance difference test.