Home
This Title All WIREs
WIREs RSS Feed
How to cite this WIREs title:
WIREs Comput Mol Sci
Impact Factor: 25.113

Computational studies of protein–drug binding affinity changes upon mutations in the drug target

Full article on Wiley Online Library:   HTML PDF

Can't access this content? Tell your librarian.

Abstract Mutations that lead to drug resistance limit the efficacy of antibiotics, antiviral drugs, targeted cancer therapies, and other treatments. Accurately calculating protein–drug binding affinity changes upon mutations in the drug target is of high interest as this can yield a better understanding into how such mutations drive drug‐resistance, especially when the mutation in question does not directly interfere with binding of the drug. The main aim of this article is to provide an up‐to‐date reference on the computational tools that are available for the calculation of Gibbs energy (free energy) changes upon mutation, their strengths, and limitations. The methods that are discussed include free energy calculations (free energy perturbation, thermodynamic integration, multistate Bennett acceptance ratio), analysis of molecular dynamics simulations (linear interaction energy, molecular mechanics [MM]/Poisson–Boltzmann solvated area, and MM/generalized Born solvated area), and methods that involve quantum mechanical calculations (including QM/MM). The possibility to use machine learning is also introduced. Given that the benefit of accurately calculating binding affinity changes upon mutation depends on comparing calculated values with experimental measurements, a brief survey on experimental methods and observables is provided. Examples of computational studies that go beyond calculating the Gibbs energy changes are given. Factors that need to be addressed by the computational chemist and potential pitfalls are discussed at length. This article is categorized under: Structure and Mechanism > Computational Biochemistry and Biophysics Molecular and Statistical Mechanics > Free Energy Methods Molecular and Statistical Mechanics > Molecular Interactions
Molecular dynamics simulations explain resistance to vandetanib. (a) Superposition of the structures of wt (cyan) and mutant (green) RET, shown with a close‐up into the site of S904F mutation. (b) Cα RMSF of two protein regions of the inhibited (top) and ATP‐bound protein, calculated from MD simulations. (c) Snapshots from MD simulations showing the hydrogen‐bond interactions between Glu734 and Arg912 in the wt and S904F variants. Reproduced from Reference 119, CC‐BY 4.0 license
[ Normal View | Magnified View ]
Resistance mutations might affect non‐binding conformations. FLT3, an important molecular target in acute myeloid leukemia, binds quizartinib, a highly specific inhibitor, when the protein is in its inactive state. MD simulations however show no difference in protein–drug interactions when considering only the inactive state and comparing the wt protein with two mutants, D835F and Y842H, that lead to resistance. Instead, it was suggested that the mutations affect the active state and make it less probable for the protein to adopt a conformation that is prone to bind quizartinib. Calculations of that are based on the inactive state will not correctly account for the full conformational dynamics of the enzyme. Reproduced from Reference 118 with permission from Wiley Periodicals, Inc
[ Normal View | Magnified View ]
Drugs that bind to multi‐domain proteins. The viral NS3/4A protease is a drug target for treatment of hepatitis C. Inhibition is aimed at the protease domain, and many structures with inhibitors include only this domain. However, as shown in the figure, inhibitors (center, stick representation) may also interact with the helicase domain and domain‐domain interactions might be important for drug binding. In such cases, it is important to consider both domains when studying resistance mutations
[ Normal View | Magnified View ]
DCHF—an example for a subsetting schemes for the interaction between polyglycine (Gly)n (upper) and polyalanine in an α‐helical structure (bottom). Reprinted (adapted) with permission from Reference 97. Copyright (2010) American Chemical Society
[ Normal View | Magnified View ]
Building a model for QM calculation of protein–drug binding energies. Three models are shown for Abl1 binding to the drug dasatinib (Figure 2), with (a) three (b) five, and (c) nine interacting residues. In this case, the binding energy was within ~1 kcal/mol of the experimental value with the first two models, whereas the third resulted in an over‐stabilization of the complex, which may be explained by the apo‐structure adopting another (more stable) conformation. Adapted from Reference 31 with permission from The Royal Society of Chemistry
[ Normal View | Magnified View ]
Boundary conditions commonly used in molecular dynamics simulations.65 (a) Periodic boundary conditions (PBCs). Replicas are made in all directions to mimic an infinite system. (b) Stochastic boundary conditions (SBCs). Atoms are restrained within a certain distance from the boundary (dashed line). Reprinted from Reference 65, with permission from Elsevier
[ Normal View | Magnified View ]
Direct calculation of the binding energy change from simulations. With methods such as FEP and TI, it is possible to estimate the free energy change when gradually mutating a residue from its initial state to a final state. In the example shown, the structure of the Abl1 protein is presented where a single mutation, Phe317→Val317 leads to drug resistance. In the simulations, the side chain of Phe317 (violet) will be gradually mutated into Val317 (pink). The simulations should be run in the presence and absence of the drug (Equation (9)). The drug (dasatinib) is shown here in light gray
[ Normal View | Magnified View ]
Direct calculation of Kd from MD simulations. Multiple binding and unbinding events must be evident. To study mutations, the simulations should be repeated for the wild type and mutant by modifying the residue that is mutated (cyan)
[ Normal View | Magnified View ]
A secondary NS3/4A mutation restores the protein's activity introduced by a resistant mutation through modifying the active site hydrogen‐bonding network. The average number of hydrogen bonds between active site residues (including those of the substrate) is shown on the figure for prevalent interactions and demonstrated with lines, whose width is proportional to the average number of h‐bonds in the simulations. Substrate residues are shown in green on the right‐hand side of the figure, key residues are indicated on the left. The single mutant is shown on the top, the double mutant on the bottom, with a cation‐π interaction indicated (shown as a double line on the right). Note the strengthening of the charge–charge interaction involving Arg155 of the protease and Glu4 of the substrate. Adapted from Reference 120, CC‐BY 4.0 license
[ Normal View | Magnified View ]
The fitness landscape is dramatically influenced by drug treatment. Three clones (x, y, z) are equally fit in the absence of drugs. Drug 1 favors clones y and z, drug 2 favors clone z, and drug 3 favors clone y. Adapted from Reference 6, CC‐BY 4.0 license
[ Normal View | Magnified View ]
Molecular dynamics simulations and free energy landscape analysis reveals why the K510E mutation in the fungal protein Candida albicans LeuRS leads to drug resistance. (a) In the wt protein, the side chain of residue Tyr487 can adopt primarily two conformations, one that is similar to the inhibitor‐bound conformation, I, and another where it is hydrogen bonded to Lys510, K. A third conformation, S is less common in the simulations and is not shown. The simulations were run in the presence of a substrate. (b) Free energy landscapes, extracted from metadynamics simulations of the wt and the mutant show that the I conformation is higher in energy with respect to the K one, especially in the mutant, suggesting why the inhibitor binding energy is affected despite the large distance between residue 510 and the substrate (or inhibitor). Adapted with permission from Reference 121. Copyright (2015) American Chemical Society
[ Normal View | Magnified View ]
The dynamics of T315I and E255V/T315I mutants of Abl1. RMSF, calculated for simulations of Abl1, reveals that there is a decrease in the plasticity of the activation loop (A‐loop) residues upon this mutation. In the E255V/T315I double mutant, the flexibility is enhanced (see the highest RMSF for A‐loop residues). Since A‐loop dynamics is important for activation, this likely stabilizes the active conformation that does not bind ponatinib, which decreases the binding affinity and increases the catalytic efficiency. Reproduced from Reference 9 with permission from Elsevier
[ Normal View | Magnified View ]

Browse by Topic

Molecular and Statistical Mechanics > Molecular Interactions
Molecular and Statistical Mechanics > Free Energy Methods
Structure and Mechanism > Computational Biochemistry and Biophysics

Access to this WIREs title is by subscription only.

Recommend to Your
Librarian Now!

The latest WIREs articles in your inbox

Sign Up for Article Alerts