Enzymes are the true players of life, they catalyze chemical reactions inside cells, resulting in an intricate net of chain reactions able to create signals, movement, changes in color, growth, and thoughts.

But as we have already seen, enzymes can be useful in our everyday life and can be exploited for the production of chemicals, for environmental remediation, or the development of therapies. For some of these applications naturally occurring enzymes are not enough, they often have rates limitations, problems in working conditions, or cannot catalyze the reaction at all. For these reasons, redesigning natural enzymes, or creating completely new ones from scratch, has been a growing field for decades.

In nature, enzymes are subjected to evolution, and random mutations are selected, resulting in advanced and more specific catalysts. Directed evolution is a laboratory technic that aims to mimic the natural process, by creating large collections of randomly mutated enzymes, and screening them for better performances.

Even if laboratory evolution can achieve good results in terms of catalysis, it is a long and wasteful process, a possible solution could be the rational design of enzymes. The basic concept is to perform an in silico screening of possible enzyme variants before the actual protein is even produced. This requires the creation of a computational model of the protein (sequence, 3D structure, volume) and of the environment (solvent, ionic force, temperature...), to evaluate which protein can bind better the substrate, or which variant is more stable.

Computational enzyme engineering pipelines. Source Scherer M et al, 2021. 

The model

The starting input is a 3D model of the wild-type enzyme, this can be created by X-ray crystallography, or by predictive measures. The design of new catalysts requires high-resolution information on the residues in the active site so usually, information from both experimental and predictive analysis are combined together.

One of the easiest ways to represent large molecules is a sphere representation, this way atoms are modeled as spheres that overlap where bonds are present. Each atom has its characteristic properties like mass, charge, and radius that depend on its element. The properties of the model, like charge distribution, surface and hydrophobicity are necessary to have a representative picture of the real protein and of the possible effects of mutations. 

In catalysis, the interaction between substrate and enzyme is crucial, in most cases, the smaller molecule has to enter into the active site, which is buried inside the protein. The substrate is recognized at the entrance of the active site, once inside the internal pocket, it is stabilized in the right position by stacking forces or by H bonds so that it is optimally aligned with the active residues.



It is important to note that each model requires approximations that are usually a trade-off between accuracy and computational power.  

Positioning of the gluten peptide inside a gluten-binding protein in the active site.


The dynamics

Proteins, like other big biomolecules, are dynamic structures, they exist in solution and interact with other solutes, salts, and solvent molecules. Each protein contains also the internal degree of freedom given by the elasticity of bonds and their movement of torsion and rotation.

These movements arise from the forces that act on each atom, like the electrostatic potential between polar and charged groups, weak interactions like Van del Walls forces, and their momentum (controlled by temperature and pressure). All these parameters need to be taken into consideration when creating reliable models for molecular dynamics (MD) simulations. A simulation can take from 1 ns to 1 Î¼s, depending on the time-scale of the phenomenon that the researcher is trying to observe, the program used for the MD (e.g. GROMACS) will calculate for each time-step (in the order of fs) the new atoms' position. For large proteins like enzymes, which undergo slow conformational modifications, the length as well as the resolution of the simulation, can be high, requiring a lot of computational power. For these reasons the simulations are often submitted to large cluster computers, that can perform these calculations in a few days. In the case of  (MD) simulations, the most important region of the protein, like an enzyme active site or the epitope-binding region or an antibody, is modeled with more precision, using for example quantum mechanics description of atoms, other regions can have a coarse representation, or be fixed as a rigid structure.

So the truth is that there is not just one protein configuration, in the real world the protein exists in different forms in equilibrium with each other. Different environmental conditions like pH, the presence of salts, or temperature, shift the equilibrium to different configurations and can be controlled, for example, to guide the formation of the active conformation of the protein.

Good recognition of the substrate happens when there's a good surface complementarity between the two molecules and the substrate is covered by the activity of the solvent. H-bonds and good alignment of the substrate with the catalytically active residues can enhance drastically the efficiency of the enzyme. Molecular docking is a tool used to search for stable conformation of protein-ligand, and take into consideration all of these parameters to score the fit between the two molecules. 

Scheme of MD perform on proteins in a cubic box filled with water molecules. The potential energy function (V) takes in consideration the weak (electrostatic, VdW) and strong (bonds) interactions between atoms. Source: Salmaso V et al (2018).

So, after an initial characterization of the mechanism behind an enzymatic reaction, it is possible to create variants of the enzyme to increase:

  • thermal stability: Increased rigidity, and stability of unstructured regions (mobile loops)
  • substrate specificity: by modeling the dimension, shape, and hydrophobicity of the AS and its entrance
  • increase the efficiency of conversion: change of active residues, and correct alignment with the substrate to minimize the energy barriers necessary for the reaction.
  • increase catalytic rate: by changing the structures that are limiting the reaction, promoting conformational changes, or by removing barriers
Designing completely new enzymes can be done by using an initial protein frame as a scaffold, and by introducing, removing, or changing residues, in order to obtain a particular activity.
When a potential candidate variant seems to have good performances, it can be actually expressed as a protein with recombinational tools, to test its actual activity.


The better our understanding of proteins and their mechanisms, the easier is to create drugs and enzymes with better performances (specificity and efficiency of reaction). Even though the modeled system is as small as a single protein-substrate complex, it requires time and a lot of computational resources to be correctly modeled. Increased availability of experimental and predictive 3D structures is leading to more reliable models, enhancing the fidelity of the computational (rational) design of enzymes.

The studies on the characterization of lab evolution mutants, which usually outperform the designed ones, reveal that is still very difficult to predict the effect of mutations, both in the active site and further away in the structure.