Team:HKUST/Model

iGEM HKUST

Modelling

Overview

Tensile properties of hagfish slime intermediate filament

Tensile properties like elasticity and stiffness are import criteria in the quality assessment of textile materials. It has been reported that native hagfish slime threads exhibit intriguing mechanical performance, which makes it appealing as a potential textile material. In the hydrated state, they exhibit an initial elastic modulus of ~6 MPa and a yield stress of ~3 MPa. After 70% extension, the threads experience a strain hardening process, resulting in an UTS of ~200 MPa and strain failure up to 220% [1]. Another notable feature of these threads is that their mechanical response can be tailored by a draw-processing mechanism as follows: The drawn threads are processed by subjecting hydrated native threads to a certain strain under wet conditions, and then dried afterwards [1]. The process results in a remarkable increase in both stiffness and ultimate tensile strength (UTS).


The molecular basis of the enhanced property during draw-processing

The molecular basis of the enhanced property during draw-processing is thought to be due to strain induced conformational transition from alpha-helical coiled-coil to aligned beta-sheet rich structures [2]. This process is also known as α→β transition.

The α→β transition is a universal deformation mechanism in alpha-helix rich protein materials such as wool, hair, hoof, and cellular proteins. This transition in hagfish slime threads has been confirmed by wide-angle X-ray scattering from previous research [2].


Figure 1: X-ray diffraction patterns for hagfish thread bundles strained in seawater. (A) Unstrained threads exhibited a typical “α-pattern,” whereas threads extended to a strain of 1.0 exhibited a typical “β-pattern” (C). Thread extended to a strain of 0.60 exhibited a mixed pattern, suggesting the presence of both α-helix and β-sheet structure (B).

However, the small number of strains examined do not allow previous researchers to provide a detailed account of the α→β transition from the x-ray data.

Therefore, we’d like to employ molecular modeling to carry out a real-time simulation of the α→β transition as a function of strain and estimate the tensile properties of our IF product simultaneously.

Molecular Modeling

Molecular modelling encompasses all theoretical methods and computational techniques used to mimic and study the structure and behaviour of molecules [3]. To explore the structural and functional properties of hagfish slime intermediate filament fibre, we started from studying its building block - the coiled coil IF protein dimer. First, we predicted the three-dimensional structure of IF dimer and then performed molecular dynamics (MD) simulations to reveal the real-time conformational changes during draw-processing and estimate the tensile properties at the same time.

Objectives

  1. To predict the molecular structure of hagfish slime IF subunits and dimer
  2. To demonstrate the real-time conformational changes during draw-processing at atomic level
  3. To estimate the tensile properties of our product as a versatile material

Figure 2: Flow chart of molecular modeling. We firstly predict the 3D protein structure based on sequence, and then perform molecular dynamics (MD) simulations on the predicted models to get results of interest.

Protein Structure Prediction

Predict protein structures based on sequences

In the iGEM part registry, there are only the amino acid sequences of alpha & gamma subunits but no 3D structures available. And from our interview with Dr. Douglas Fudge, we learned that the terminal regions of IF are inherently elastomeric domains, which may make it difficult to experimentally determine the structure at the atomic-resolution level. Therefore, we turned to the power of computation and algorithms to predict the tertiary structure of IF subunits and dimer.

There are many free software tools to choose from for creating a predicted model of a protein's tertiary structure. The methodologies employed by these software tools can be categorised as either template-based or ab initio protein structure prediction. Two web-servers were used in our prediction. SWISS-MODEL is a structural bioinformatics web-server dedicated to homology modeling, while I-TASSER (Iterative Threading ASSEmbly Refinement) server utilises both template-based threading and ab initio simulation methods to produce its final model [4,5].

Our original plan was to first predict the structure of individual subunits and the next step is to assemble two subunits into one dimer with protein docking. Herein, with the I-TASSER, we input two subunit sequences individually and created two individual jobs.<\p>

Figure 1: Alpha subunit model (yellow).
Figure 2: Gamma subunit model (red).

However, we later found that the SWISS-MODEL server allows us to input two query sequences together by adding hetero target, and the output is inherently a dimer model, which saves the effort of protein docking. So, we took advantage of this function and got the dimer model in only one step.


Figure 3: Dimer model including alpha & gamma subunits.

Quality assessment of predicted models

Without knowing the protein's native structure we cannot determine the accuracy of a predicted model. There are, however, tools available to assess the hypothetical quality of a protein model. First, we selected those top models scoring the highest in the in-built analysis values of two prediction web-servers. Then, we employed two external quality assessment tools which were ProSA and Molprobity, respectively. ProSA calculates an overall quality score for a specific input structure, and if this score is outside a range characteristic for native proteins, the structure probably contains errors [6]. Molprobity, developed by the Richardson group at Duke university, provides a summary of scoring statistics as shown below, and the green cell means that the model passes one test while the red cell means that it fails [7].

ProSA Evaluations

Figure 1: ProSA evaluation result for the dimer model. The z-score of the dimer structure is within the range of scores typically found for native proteins of similar size.
Figure 2: ProSA evaluation result for the alpha subunit model. The z-score of the alpha subunit is beyond the range, indicating this structure probably contains errors.
Figure 3: ProSA evaluation result for the gamma subunit model. The z-score of the gamma subunit is beyond the range, indicating this structure probably contains errors.

We can see from the figures above that the z-score of the dimer structure is within the range of scores typically found for native proteins of similar size, while the z-scores of two subunits are beyond the range, which indicates that the dimer model is a more ideal prediction than the other two subunits.

Molprobity Evaluations

Figure 4: Molprobity evaluation result for the dimer model. The dimer model passes five test statistics including Protein Geometry - Ramachandran outliers, Ramachandran favored, Rama distribution Z-score and Low-resolution Criteria - CaBLAM outliers, CA Geometry outliers.
Figure 5: Molprobity evaluation result for the alpha subunit model. The alpha subunit only performs well in one test statistics: All-Atom Contacts - Clashscore, all atoms.
Figure 6: Molprobity evaluation result for the gamma subunit model. The gamma subunit only performs not bad in one test statistics: All-Atom Contacts - Clashscore, all atoms.

The figures above obviously show that the dimer model satisfied much more test statistics than the other two subunits although it’s still not perfect owing to the limitations of in silico prediction.

Conclusion

According to the evaluation results of both tools, we may conclude that the dimer model possesses a better structural quality than the other two subunit models and thus we eventually chose the dimer structure as our prior model for the following molecular dynamics simulations, which also saved the effort of applying protein docking to two individual subunits.


Molecular Dynamics Simulations

Overview

Molecular Dynamics (MD) is a numerical method of simulation, in which the atoms and molecules are allowed to interact for a fixed period of time, giving a view of the dynamic "evolution" of the system. The trajectories of atoms and molecules are determined by numerically solving Newton's equations of motion for a system of interacting particles, where forces between the particles and their potential energies are often calculated using interatomic potentials or molecular mechanics force fields [8].

Figure 1: The bonded forces of an atom in a md simulation are springs that exist in three forms. They act according to hooke's law; energy and force increase proportionally to the compressing and stretching of the spring away from its equilibrium distance.

With the predicted dimer model, we planned to perform molecular dynamics (MD) simulations to reveal the real-time conformational changes during draw-processing and estimate the tensile properties at the same time. Before the actual simulation takes place, we need to first achieve the solvation, energy minimization, and equilibration in water for our dimer molecule. After getting our molecule prepared, steered molecular dynamics (SMD) is performed to simulate the draw processing. By analysing the trajectories, we are able to visualize the real-time conformational changes and extract the tensile properties of the dimer molecule as an ideal rigid rod model.

For molecular dynamics simulation, our team used the Visual Molecular Dynamics (VMD) software. VMD is designed for modeling, visualization, and analysis of biological systems such as proteins, nucleic acids, lipid bilayer assemblies, etc [9]. In particular, VMD provides embedded scripting languages (Python and Tcl) for the purpose of user extensibility. Here we have collated a detailed explanation of the methods our team used for MD simulation with VMD and uploaded all scripting files onto our GitHub repository. Our aim is that these instructions will provide a reference point, helping future iGEM teams which wish to use VMD software to do molecular dynamics simulation.


Solvation, energy minimization and equilibration in water

Objectives of solvation, minimization and equilibration before SMD

  1. Approach in vitro aqueous incubation environment
  2. Many protein structures come from x-ray crystallography at extremely low temperature, so the determined conformation may not be realistic at room temperature [10]. We’d better do the minimization under desired conditions like room temperature.
  3. Prepare a nice starting structure for the following simulations by releasing constraints

As suggested by VMD user guide, before the actual molecular dynamics simulation, we need to first perform the following three steps to reach the stable conformation of the molecule:

  1. Minimize the whole system for 6000 steps
  2. Equilibrate the system in the NPT ensemble (constant atom number N, constant pressure P, and constant temperature T) for 50000 steps (i.e. 100 ps) with protein fixed
  3. Free all atoms, minimize for another 2000 steps, and run NPT dynamics for 100 ps again
Figure 1: Suggestions form VMD user guide. It usually takes several minimization-equilibration cycles to obtain a stable starting structure for the following simulations.

VMD includes many analysis tools and we use these as criteria to determine whether the molecule has reached the equilibrium. One of these, the NAMD Plot plugin, may be used to plot the TOTAL v.s. Timestep curve to show the variation of total free energy of the system during the minimization. Besides, the NAMD log files during minimization record a parameter called gradient tolerance, and normally when the gradient value drops to 1, it indicates the convergence to the minimum [11]. You can also analyze the extent to which your system has equilibrated using what is known as the Root Mean Square Deviation, or RMSD. The RMSD characterizes the amount by which a given selection of your molecule deviates from a defined position in space, and a flattening RMSD curve usually means the system is equilibrated [12].


Step 1: Minimize the whole system for 6000 steps


Figure 2: TOTAL v.s. Timestep plot during step 1
Figure 3: Value of gradient tolerance at timestep 5997 (6000 ts in total)

From the plot (fig.2) we can see that the total energy dropped to the negative range and stayed at the stable negative level afterwards. The gradient value at ts 5997 (fig.3) was close to 1, which also indicates the minimization was done.

Step 2: Equilibrate the system in the NPT ensemble for 50000 steps (i.e. 100 ps) with protein fixed


Figure 4: TOTAL v.s. Timestep plot during step 2
Figure 5: Trajectory of the solvated molecules during step 2 (click to animate)

From the plot (fig.4) we can see that the total energy raised a little bit due to the movement of water molecules and soon reached a stable negative level again.

Step 3: Free all atoms, minimize for another 2000 steps, and run NPT dynamics for 100 ps again

Figure 6: TOTAL v.s. Timestep plot during step 3
Figure 7: Root Mean Square Deviation (RMSD) curve during step 3
Figure 8: Trajectory of the solvated molecules during step 3 (click to animate)

From two plots above (fig.6 & 7), we can see that the total energy of the system reached a stable negative level and the RMSD curve also became flattening at ~1.7 +/-0.2 Angstroms.

Conclusion

We may conclude that the molecular system has reached equilibrium after these three steps and are ready for the following steered molecular dynamics simulation.


Steered Molecular Dynamics

Objectives

  1. To demonstrate the real-time conformational changes during draw-processing
  2. To estimate the tensile properties of our product

A tensile test, also known as a tension test, is one of the most fundamental and common types of mechanical testing. A tensile test applies tensile (pulling) force to a material and measures the specimen's response to the stress. By doing this, tensile tests are able to determine many tensile properties such as ultimate tensile strength, Young's modulus, strain-hardening characteristics, etc [13].

However, the tensile test can not account for the conformational changes like the α→β transition of the molecules from a micro point of view. Therefore, we introduced steered molecular dynamics (SMD) simulations into our model. SMD simulations apply forces to a protein in order to manipulate its structure by pulling it along the desired direction [14]. These in silico experiments can not only be used to reveal the real-time structural changes in a protein at the atomic level but also mimic the tensile test to estimate the tensile properties.

Figure 1: A tensile testing equipment in the laboratory

Real-time conformational changes at atomic level

In our simulation, we employed the SMD simulations with constant velocity pulling on the dimer molecule. Upon fixing the Cαs of the N-terminus amino acid, the SMD simulation pulled the Cαs of the C-terminus amino acid of two subunits at a constant velocity, whose direction is determined by the mass centers of fixed and pulled atoms, for 80000 timesteps (i.e. 160 ps). So, the dimer molecule would undergo structural deformation, and the VMD would create the trajectories recording the states of all atoms at each timestep.

Figure 2: SMD simulations with the dimer molecule. The pulling direction is determined by the mass centers of fixed and pulled atoms.

Results

Figure 3: Animation of the SMD simulation with constant velocity pulling.

The animation above shows the elongation and deformation of the dimer molecule. Unfortunately, the VMD software cannot update secondary structure automatically in the animation, so we have to use the console to recalculate the secondary structure manually at each frames.

Figure 4: Secondary structures of the dimer molecule at Frame 1, 57, 105, and 130 (the trajectory contains 160 frame in total)

According to the simulation outcome, we may demonstrate the conformational changes during the draw-processing from a micro point of view:

In the native state, two α-helical proteins wind into coiled-coils. Under stress, the coiled-coils start to unravel into random coils at specific stress values. When loading increases, the random coils are extended further and begin to form extended β-sheet structures with inter-chain hydrogen bonds. Eventually, the majority of α-helices unravel, and β-sheets extend and align, and this process exactly elucidates the α→β transition.

However, due to the limitations of the quality of our dimer model and the in silico force field parameters, we didn't see as much beta-sheet alignment as expected. We will try to refine the simulation quality in the future and give a more accurate and detailed molecular basis of the strain-hardening during draw-processing.


Tensile properties of hagfish slime IF fibres

Apart from explaining the α→β transition at atomic level, we can also regard the dimer molecule as an ideal rigid rod model to estimate the tensile properties of our product.

Hypothesis

The following model is given under the assumption that coiled coil IF protein dimers are the functional units of IFs and possess the same properties as all higher order structures up to the IFs themselves. By regarding the dimer molecule as a rigid rod, the diameter of the cross-section area of the molecule can be approximated to the distance between two fixed atoms while the length of the molecule can be approximated to the distance between centers of mass of fixed and pulled atoms, respectively.

Figure 6: The dimer molecule is regarded as an ideal rigid rod model (blue cylinder).
Figure 7: The diameter of the cross-section area of the molecule is approximated by the distance between two fixed atoms (red balls).

Formulas and Parameters

Figure 5: Constant velocity pulling in a one-dimensional case. The dummy atom is colored red, and the SMD atom blue. As the dummy atom moves at constant velocity the SMD atom experiences a force that depends linearly on the distance between both atoms.

In this type of simulation, we define the position of the fixed atom and the SMD (pulled) atom, and the pulled atom is attached to a virtual atom via a virtual spring. This virtual atom is moved at constant velocity and then the force between both is measured using:

\begin{equation}U = \frac{1}{2}k\left[vt - \left(\vec{r} - \vec{r_0}\right) * \vec{n}\right]^2\end{equation}

Formula 1: Calculate the elastic potential energy of the virtual spring, where U is the potential energy of the spring, k is the spring constant, v is the pulling velocity, t is the simulation time, r is the actual position of the pulled atom, \( r_0 \) is the initial position of the pulled atom, n is the normalized direction of pulling.

\begin{equation}\vec{f} = -\nabla U = \left( -\frac{\partial U}{\partial x}, -\frac{\partial U}{\partial y}, -\frac{\partial U}{\partial z} \right)\end{equation}

Formula 2: Calculate the force applied on the pulled atom, where f is the force applied on the pulled atom, U is the potential energy of the spring.

The stress-strain curve is universally used to give the relationship between stress and strain and demonstrate the tensile properties of textile materials. Therefore, we applied the formulas below to extract useful information from the SMD simulations for the stress-strain curve.

\begin{equation}\vec{F} = \vec{f} * \vec{n} = f_x * n_x + f_y * n_y + f_z * n_z\end{equation}

Formula 3: Calculate the force in the direction of pulling, where F is the force in the direction of pulling, f = (fx, fy, fz) is the force applied on the pulled atom, n = (nx, ny, nz) is the normalized direction of pulling

\begin{equation}strain = \frac{\Delta L}{L_0} = \frac{L_i - L_0}{L_0} = \frac{\sqrt{(x_i - x_0)^2 + (y_i - y_0)^2 + (z_i - z_0)^2}}{\sqrt{(x_1 - x_0)^2 + (y_1 - y_0)^2 + (z_1 - z_0)^2}}\end{equation}

Formula 4: Calculate the strain at each timestep, where L0 is the original length of molecule, (x0, y0, z0) is the position vector of center of mass of fixed atoms, (xi, yi, zi) is the position vector of center of mass of pulled atoms at timestep i = 1, 2, …, 80000

\begin{equation}\sigma = \frac{F}{A} = \frac{F}{\pi * \left(\frac{d}{2}\right)^2}\end{equation}

Formula 5: Calculate the stress at each timestep, where σ is the stress throughout the molecule, F is the magnitude of force in the pulling direction, A is the cross sectional area of the molecule, d is the distance between two fixed atoms

Results

With the data we obtained, we could plot the unique stress-strain curve of the dimer molecule (figure 8).

Figure 8: The stress-strain curve of the IF dimer molecule

We can see that the curve is approximately J-shape, which means that the molecule exhibits the low initial stiffness and later extreme strain hardening. Region I is a low-stiffness initial region that terminates at a strain of ɛ ≈ 0.7. In region II, stiffness rises dramatically until a strain of ɛ ≈ 2.0.


Figure 9: Stress-strain curves of the IF dimer molecule and multiple textile materials such as nylon and spider silk

In comparison with the stress-strain curves of other textile materials (figure 9), we may conclude that the model of our IF fibres initially possess remarkable elasticity as nylon and also exhibit high stiffness similar to spider silk after the draw-processing, which supports the outstanding tensile properties of the hagfish slime intermediate filament as a versatile material.


Figure 10: Stress-frame curve of the IF dimer molecule combined with snapshots of the secondary structure of dimer molecule at Frame 1, 57, 105, and 130

In addition, if we check the stress-frame curve combined with secondary structures at different frames (figure 10), it could be found that the rise of stiffness is accompanied by the appearance of β-sheet, which is consistent with previous research result that the α→β transition harden the IF fibres during draw-processing.


Colour Mixing

Objectives

  1. To predict the percentage of chromoproteins required to make a certain colour: so that we can control the resulting colour
  2. To generate a color chart, so that we can predict the theoretical range of colours possible

Colour Mixing Models

Eyes perceive green light and red light reflected from a yellow cat. Note that only red, blue, and green light are considered in this diagram. However, white light from the sun or from lamps contains many more colors of light.

The colour that we see is a result of a combination of different wavelengths of light with different power levels. When a light source shines on an non-luminous object (an object that does not glow), certain wavelengths of the light will be absorbed and the remaining wavelengths of light will be reflected and enter our eyes, which will be interpreted as colours in our brain. Because our eyes only contain three different types of colour-sensitive cells for red, green, and blue light respectively, we can use the three primary colours of light (red, green, and blue light) to produce a wide range of colours. This is the RGB colour model, a subset of additive colour mixing models, which produces colour using a mixture of different wavelengths of light.

The Blue Yellow Red Color Wheel. Image created by Ray Trygstad on 10/11/04. Cleaned up version created in Adobe Photoshop by stib 11:41, 7 Apr 2005 (UTC). Licensed under the Creative Commons Attribution-Share Alike 3.0 Unported license.

In contrast to luminous objects, non-luminous objects do not produce light directly but instead only reflect them. Thus, we apply the principles of subtractive colour mixing to predict the colour of this mixture of reflected light. In our project, we will apply one of the models of subtractive colour mixing : the RYB colour model. In this model, red, yellow and blue are the three primary colours. The RYB model was traditionally used by artists, due to the limitation of the colours of pigments. By combining different amounts of these three primary pigments, we can reproduce a wide range of colours. For example, combining Red and Blue will create Purple. (Figure. 1)

Secondary colors are the resulting colours when the primary colours are mixed in equal proportions.

Hypothesis

The chromoproteins used in our project are non-fluorescent, which means it does not emit light. Thus, we try to apply the RYB color model to predict and reproduce different colors by mixing different amounts of chromoproteins with different colors.

To simplify the color mixing model, we have made some assumptions. The chromoproteins are treated as pigments that attach to our IF at one terminus. And IF threads are assumed to be white. In total, we have the three primary colors plus white. We also assume that chromoproteins of different colors will evenly distribute within the hagfish fiber after the polymerization of the fiber and the interference among the chromoproteins are negligible.

Assumptions

  1. IF thread is treated as white
  2. Chromoproteins of different colors will evenly distribute within the hagfish fiber after the polymerization

Our Assumptions in our project

Assumptions

Descriptions

 Illustration

1. Our colour mixing system is based on the dyeing method of linking chromoprotein at the N- termini of 2 hagfish IF subunits

Each subunit (Alpha & Gamma) is connected with chromoproteins at N-terminal and forms dimer. Here Blue chromoproteins is linked on gamma subunits, Red chromoproteins is linked on alpha subunits

2. IF thread appears as White

our assumptions of IF without chromoproteins will appear as white, the red line means alpha subunits while blue line means gamma subunits.

3. Our 3 primary colours is Red, Yellow, Blue

We will be using 2 different chromoproteins for making each primary colour

  • Red : eforRed (BBa_K592012), mRFP1 (BBa_E1010)

  • Blue : aeBlue (BBa_K864401) , amilCP (BBa_K592009)

  • Yellow : amilGFP (BBa_K592010) , fwYellow (BBa_K1943001)

4. chromoproteins of different colors will evenly distribute within the hagfish fiber after the polymerization 

The right illustration shows how mixing of colours produces different colours of Intermediate filaments theoretically. Each subunit (Alpha & Gamma) is connected with chromoproteins at N-terminal and forms dimer, dimer will form tetramers and finally bound together as an entire intermediate filaments polymers. Here Blue chromoproteins is linked on gamma subunits, Red chromoproteins is linked on alpha subunits, the entire intermediate filaments is hoped to be purple in colour. (CP= Chromoproteins)




Major Targets in our color mixing system

This table demonstrates how we planned to do colour mixing of Intermediate filaments.

 

Targets

Descriptions

illustrations

Target 1 : mix only 2 different colours of chromoproteins to generate secondary colours: 

  • Creating Orange, Green, Purple IF Fibres

3 primary colours (Red, Blue, Yellow) are being used to link to each subunit. By linking different colour of chromoprotein to each subunits and mixing colour with 1:1 ratio, it is hope to generate 3 secondary colour (Green,Purple,Orange)

We will be comparing colour mixed fibre with fibre created by pure colour chromoproteins. Polymerization,colour intensity ,colour stability will be compared

Target 2 : mixing 3 different colours of chromoprotein-linked fibres

  • Creating colour palette with different level of Lightness & hue

After experimental trials, we will proceed to our second target, which is by mixing 3 different colours of chromoproteins to generate a colour palette.As mentioned in previous assumption, our IF without chromoproteins will be white. This will be our changing variable for Lightness of a colour. It is hoped that when addition of IF without chromoproteins to IF with chromoproteins, the entire filaments will be paler, thus increasing lightness.

Factors Affecting our results

There are three factors that may influence the color of the fiber:

  1. The type of chromoprotein attached
  2. Relative percentage of each chromoprotein attached
  3. Absolute percentage of each chromoprotein in the fibre

The relative percentage here means the relative ratio of different chromoproteins in the fiber. The absolute percentage of a chromoprotein means the percentage of IF that has been attached with a specific type of chromoprotein.

Credit: SharkD, CC BY-SA 4.0 , via Wikimedia Commons

To help illustrate, the HSL (hue, saturation, lightness) color system is used. In terms of the three parameters in the HSL system, we can say that the relative percentage determines the hue while the absolute percentage determines the color intensity or saturation. Hue here means a general color like red while the color intensity or saturation here means how intense or saturated the color is. For example, red is more saturated than pink. The hue will change when the relative percentage of chromoprotein changes. The color intensity will increase when the absolute percentage of each chromoprotein increases but the relative percentage remains unchanged. Lightness depends on the amount of light in the environment so it is not modeled.


Refining Our Model

To determine the color of an object precisely, we can use an instrument called a spectrophotometer to measure the absorption and emission spectra of the object. And using specific computer programs, we can render the color that humans can understand from this spectra. These experimental data will help us refine the color mixing model, and better estimate the color of the fiber. However, we do not have that data now. So we only have a rough prediction for the color based on subtractive color mixing.


References

[1] D. S. Fudge, K. H. Gardner, V. T. Forsyth, C. Riekel, and J. M. Gosline, “The Mechanical Properties of Hydrated Intermediate Filaments: Insights from Hagfish Slime Threads,” Biophysical Journal, vol. 85, no. 3, pp. 2015–2027, 2003.

[2] J. Fu, P. A. Guerette, A. Pavesi, N. Horbelt, C. T. Lim, M. J. Harrington, and A. Miserez, “Artificial hagfish protein fibers with ultra-high and tunable stiffness,” Nanoscale, vol. 9, no. 35, pp. 12908–12915, 2017.

[3] “Molecular modelling,” Wikipedia, 20-Oct-2020. [Online]. Available: https://en.wikipedia.org/wiki/Molecular_modelling. [Accessed: 24-Oct-2020].

[4] A. Waterhouse, M. Bertoni, S. Bienert, G. Studer, et al., “SWISS-MODEL: homology modelling of protein structures and complexes,” Nucleic Acids Research, vol. 46, no. W1, 2018.

[5] J. Yang, R. Yan, A. Roy, D. Xu, J. Poisson, and Y. Zhang, “The I-TASSER Suite: protein structure and function prediction,” Nature Methods, vol. 12, no. 1, pp. 7–8, 2014.

[6] M. Wiederstein and M. J. Sippl, “ProSA-web: interactive web service for the recognition of errors in three-dimensional structures of proteins,” Nucleic Acids Research, vol. 35, no. Web Server, 2007.

[7] V. B. Chen, W. B. Arendall, J. J. Headd,et al., “MolProbity: all-atom structure validation for macromolecular crystallography,” International Tables for Crystallography, pp. 694–701, 2012.

[8] J. Gelpi, A. Hospital, R. Goñi, and M. Orozco, “Molecular dynamics simulations: advances and applications,” Advances and Applications in Bioinformatics and Chemistry, p. 37, 2015.

[9] W. Humphrey, A. Dalke, and K. Schulten, “VMD: Visual molecular dynamics,” Journal of Molecular Graphics, vol. 14, no. 1, pp. 33–38, 1996.

[10] B. Halle, “Biomolecular cryocrystallography: Structural changes during flash-cooling,” Proceedings of the National Academy of Sciences, vol. 101, no. 14, pp. 4793–4798, 2004.

[11] “Gradient descent,” Wikipedia, 16-Oct-2020. [Online]. Available: https://en.wikipedia.org/wiki/Gradient_descent. [Accessed: 24-Oct-2020].

[12] M. I. Sadowski, “Protein Structure Comparison Methods,” Encyclopedia of Biophysics, pp. 2055–2060, 2013.

[13] “Tensile testing,” Wikipedia, 10-Oct-2020. [Online]. Available: https://en.wikipedia.org/wiki/Tensile_testing. [Accessed: 24-Oct-2020].

[14] D. Ma, “A Steered Molecular Dynamics Study On Elastic Behavior Of Polyethylene Chains,” Acta Polymerica Sinica, vol. 008, no. 5, pp. 448–453, 2008.

Bibliography

I-TASSER

  1. Roy, A., Kucukural, A., and Zhang, Y. (2010). I-TASSER: a unified platform for automated protein structure and function prediction. Nature Protocols, 5, pp.725-738
  2. Yang, J. Yan, R., Roy, A., Xu, D., Poisson, J., and Zhang Y. (2015). The I-TASSER Suite: Protein structure and function prediction. Nature Methods, 12, pp.7-8
  3. Yang, J., and Zhang, Y. (2015). I-TASSER server: new development for protein structure and function predictions. Nucleic Acids Research, 43, pp.174-181

SWISS-MODEL

  1. Arnold, K., Bordoli, L., Kopp, J. and Schwede, T. (2005). The SWISS-MODEL workspace: a web-based environment for protein structure homology modelling. Bioinformatics, 22(2), pp.195-201.
  2. Benkert, P., Biasini, M. and Schwede, T. (2011). Toward the estimation of the absolute quality of individual protein structure models. Bioinformatics, 27(3), pp.343-350.
  3. Bienert, S., Waterhouse, A., de Beer, T., Tauriello, G., Studer, G., Bordoli, L. and Schwede, T. (2016). The SWISS-MODEL Repository—new features and functionality. Nucleic Acids Research, 45(D1), pp.D313-D319.
  4. Waterhouse, A., Bertoni, M., Bienert, S.,et al. (2018). SWISS-MODEL: homology modelling of protein structures and complexes. Nucleic Acids Research.

ProSA

  1. Wiederstein & Sippl (2007) ProSA-web: interactive web service for the recognition of errors in three-dimensional structures of proteins. Nucleic Acids Research 35, W407-W410.
  2. Sippl, M.J. (1993) Recognition of Errors in Three-Dimensional Structures of Proteins.Proteins 17, 355-362.

Molprobity

  1. Christopher J. Williams, Jeffrey J. Headd, Nigel W. Moriarty, et al.(2018) MolProbity: More and better reference data for improved all-atom structure validation. Protein Science 27: 293-315.