Multiscale Quantum Mechanics/Molecular Mechanics (QM/MM) simulations have become increasingly important for studying many biochemical processes that involve bond breaking and/or charge transfers among highly charged particles. Simulating these processes require algorithmic advances for calculations in both the MM and QM domains, together with an efficient and flexible interface between the two domains, which are typically treated using different software packages. An optimal strategy of algorithm design is that such interface must not hinder individual-domain calculations but help to speed up the overall simulation. The recently developed Multiscale Modeling in Computational Chemistry (MiMiC) software achieves this by providing a very flexible and computationally efficient framework for multiscale simulations. Using this multi-layered parallelization scheme, MiMiC has displayed efficient scalability over more than ten thousand cores in a single QM/MM simulation while maintaining an overall parallel efficiency above 75%, enabling nanoscale QM/MM molecular dynamics of complex biological systems. The code has been officially released. More information about it can be found on the official webpage.
J. M. H. Olsen, V. Bolnykh, S. Meloni, E. Ippoliti, M. P. Bircher, P. Carloni, U. Rothlisberger, MiMiC: A Novel Framework for Multiscale Modeling in Computational Chemistry, J. Chem. Theory Comput. 15, 3810–3823 (2019).
V. Bolnykh, J. M. H. Olsen, S. Meloni, M. P. Bircher, E. Ippoliti, P. Carloni, U. Rothlisberger, Extreme Scalability of DFT-Based QM/MM MD Simulations Using MiMiC, J. Chem. Theory Comput. 15, 5601–5613 (2019).
MoNvIso, the Modeling eNvironment for Isoforms, is a homology modeling software developed using Python. Its main purpose is to identify the isoform of the protein most likely related to a specific disease, based on the mutations provided by the user.
It performs an evaluation on which isoform can map the highest number of mutations, then evaluates the “modellability” of all the isoforms to decide which one has the highest amount of protein surface covered by templates. It automatically searches for homologues (using BLAST API), aligns them (with COBALT), builds the Hidden Markov Model and uses it to search for templates (with HMMER API), aligns them and builds the model of the wild type and mutants (using MODELLER).
MoNvIso can be downloaded from this link.
Determining the complete set of ligands' binding-unbinding pathways is important for drug discovery and for rational interpretation of mutation data. Here we have developed a metadynamics-based technique that addresses this issue and allows estimating affinities in the presence of multiple escape pathways. The calculations require a relatively small computational cost, making this approach valuable for practical applications, such as screening of small compound libraries. This approach has been tested and applied by using the PLUMED package. More information about its implementation cab be found in the PLUMED NEST repository at this link.
- R. Capelli, P. Carloni, M. Parrinello, Exhaustive Search of Ligand Binding Pathways via Volume-Based Metadynamics. J Phys Chem Lett 10(12), 3495-3499 (2019).
- The PLUMED consortium. Promoting transparency and reproducibility in enhanced molecular simulations. Nat Methods 16, 670–673 (2019).
Localized Volume-based metadynamics
Enhanced sampling methods can predict free-energy landscapes associated with protein/ligand binding, characterizing the involved intermolecular interactions in a precise way. However, these in silico approaches can be challenged by induced-fit effects. We have developed a variant of volume-based metadynamics tailored to tackle this problem in a general and efficient way.
- Q. Zhao et al. Enhanced Sampling Approach to the Induced-Fit Docking Problem in Protein–Ligand Binding: The Case of Mono-ADP-Ribosylation Hydrolase Inhibitors. J Chem Theory Comput 17(12), 7899-7911 (2021).
The intrinsic plasticity of protein residues, along with the occurrence of transitions between distinct residue conformations, plays a pivotal role in a variety of molecular recognition events in the cell. Analysis aimed at identifying both of these features has been limited so far to protein-complex structures. We have developed a computationally efficient tool (T-pad), which quantitatively analyzes protein residues' flexibility and detects backbone conformational transitions. The code can be downloaded here.
R. Caliandro, G. Rossetti, and P. Carloni. Local Fluctuations and Conformational Transitions in Proteins. J Chem Theory Comput 8(11), 4775–4785 (2012).
Bridge and the graphical interface Bridge2
Bridge is a graph-based algorithm coded in Python. Bridge computes graphs that consist of nodes, which for a protein are protein H-bonding groups, and edges, which are H-bonds between these groups. The edges, or H-bonds between two graph nodes, can be direct H-bonds or water-mediated bridges, or water wires. Both the H-bond criteria and the length of the water wire can be chosen by the user. Once computed with Bridge, the H-bond graph can be queried to identify, for example, H-bonds that are sampled persistently, shortest-distance paths between protein groups of interest, or the H-bond cluster of a group of interest.
Bridge2 is the graphical user interface of Bridge. A major development contributed by Bridge2 is that nodes of the graph can be arranged interactively for optimal view of the H-bond network. Bridge2 contains tools for the analysis of simulation trajectories, including the identification of H-bond motifs commonly found at proton-binding sites or sites otherwise important for protein function.
Bridge2 can also be used to compute graphs of hydrophobic interaction networks.
Bridge2 can be downloaded here.
M. Siemers, A-N. Bondar. Interactive interface for graph-based analyses of dynamic hydrogen-bond networks: application to spike protein S. J Chem Inf Model 61, 2998–3014 (2021).
M. Siemers, M. Lazaratos, K. Karathanou, F. Guerra, L.S. Brown, A-N Bondar. Bridge: a graph-based algorithm to analyze dynamic H-bond networks in membrane proteins. J Chem Theory Comput 15, 6781-6798 (2019).
C-Graphs (Conserved graphs)
C-graphs and its graphical user interface enable the computation of conserved and comparison H-bond graphs. Given a set of at two static protein structures or two MD simulation trajectories, the conserved H-bond graph is defined as the graph composed of nodes and edges present in both structures/MD trajectories. The set of static protein structures or MD simulation trajectories may contain more than two structures/trajectories. For two protein structures, a comparison H-bond graph color codes the nodes and edges according to their presence in either of the structures. By projecting H-bond graphs onto the z-axis, which for a membrane protein is typically the membrane normal, C-Graphs allows the user to estimate the linear length of the H-bond networks. Given a set of static protein structures, C-Graphs uses a clustering algorithm to identify sites in which water molecules are conserved, and maps these waters onto the protein-water H-bond graph.
Static protein structures subjected to analyses with C-Graphs must not necessarily belong to proteins with the same amino acid residue sequence. C-Graphs contains a protocol that can be used to compute conserved H-bond graphs for static structures of distinct G Protein Coupled Receptors (GPCRs).
C-Graphs and the User’s Manual for C-Graphs can be downloaded here.
- É. Bertalan, E. Lesca, G.F.X. Schertler, A-N. Bondar A-N. C-Graphs tool with graphical user interface to dissect conserved hydrogen-bond networks: applications to visual rhodopsins. J Chem Inf Model 61, 3692-5707 (2021).
The Depth-First-Search (DFS) algorithm for topologies of dynamic lipid H-bond clusters visits the nodes (lipid headgroups) of the H-bond graph to identify four types of topologies: linear clusters, star and linear, circular, and combined star, circular and linear. The size of a lipid H-bond cluster is given by the number of nodes in the cluster.
You can download the DFS algorithm here.
- K. Karathanou, A-N. Bondar. Algorithm to catalogue topologies of dynamic lipid hydrogen-bond networks. BBA-Biomembranes 1864, 183859 (2022).