Structure-based mutation analysis (PKU)

From Bioinformatikpedia

To all who are engaged in Psyche's task, of sorting out the seeds of good from the seeds of evil, i dedicate this discourse. FRASER

Short task description

This week, we will introduce mutations in the known tertiary structure of our protein and calculate and compare the potential energy of mutant and wildtype protein with different methods. See the complete task description for details. Our journal might be found here.

Finding the right structure

As proposed we searched the UNIProt entry for our protein and then selected the entry with the highest resolution and the lowest r-Value. In our case this is 1J8U which is the protein in a complex with its cosubstrate BH4. IN the following we will only use this structure, but we also list the results we found. <figtable id="tab:uniprotresult">

Table with results from UNIProt with r-Value inserted
Entry Method Resolution (Å) r-Value Chain Positions PDBsum
1DMW X-ray 2.00 0.200 A 118-424 [»]
1J8T X-ray 1.70 0.197 A 103-427 [»]
1J8U X-ray 1.50 0.157 A 103-427 [»]
1KW0 X-ray 2.50 0.220 A 103-427 [»]
1LRM X-ray 2.10 0.211 A 103-427 [»]
1MMK X-ray 2.00 0.199 A 103-427 [»]
1MMT X-ray 2.00 0.213 A 103-427 [»]
1PAH X-ray 2.00 0.176 A 117-424 [»]
1TDW X-ray 2.10 0.206 A 117-424 [»]
1TG2 X-ray 2.20 0.213 A 117-424 [»]
2PAH X-ray 3.10 0.251 A/B 118-452 [»]
3PAH X-ray 2.00 0.175 A 117-424 [»]
4ANP X-ray 2.11 0.204 A 104-427 [»]
4PAH X-ray 2.00 0.169 A 117-424 [»]
5PAH X-ray 2.10 0.163 A 117-424 [»]
6PAH X-ray 2.15 0.171 A 117-424 [»]

</figtable> In <xr id="tab:uniprotresult"/> there are all results according to which we selected 1J8U to be our reference for this weeks task. The corresponding line is marked in yellow. Unfortunately, as we discovered in the preparation of SCWRL, the coverage in not from 103 to 427 for 18JU but only from 118 to 427.

1J8U

In order to know the structure of the protein and its important residues, we have a look at its structure with PyMol and visualize the BH4 and the Fe-ion with the most important residues.

<figure id="fig:1J8Uwhole">
Rendering of the overall structures of 1J8U using PyMol. The protein is colored cyan overall, whereas the Fe-atom is colored red and the important residues are shown as sticks and colored in the element-based fashion. Binding is shown with yellow strokes, if the distance is bigger than 1.5 Å
</figure>
<figure id="fig:1J8Uclose">
Rendering of a close-up of the structures of 1J8U using PyMol. The protein is colored cyan overall, whereas the Fe-atom is colored red and the important residues are shown as sticks and colored in the element-based fashion. Binding is shown with yellow strokes, if the distance is bigger than 1.5 Å
</figure>

<figure id="fig:structuremoving">

Structure of 1J8U with both ligands Iron (red) and BH4 (element coded color) with their binding sites orange and yellow respectively. BH4-binding sites are van-der-Waal's based and hydrogen bond based, whereas iron is covalently bound (orange)

</figure>

Mutations

As you probably know from last weeks dataset the first two mutations are located before residue 103 and therefore not contained in the structure. We changed them to mutations, which we think are interesting from a structural view. We propose the following dataset, chosen mostly from well known SNPs from OMIM. They include mutations causing no reported effect, the mild related hyperphenylalaninemia (reduced activity, but functional enzyme) and phenylketonuria. We will present a similar table for each method, in which we show the results we would suggest from this method alone in the end we present our overall results here, which will be a intuitive consensus for all the intermediate results.

SNP effect prediction validation
ARG158GLN disease causing effect
True.jpeg
GLN172HIS non-disease no effect
True.jpeg
ARG243GLN disease causing effect
True.jpeg
LEU255SER disease causing effect
True.jpeg
MET276VAL non-disease effect
True.jpeg
ARG297CYS disease causing no effect
Wrong.jpeg
New.jpeg
ALA322GLY hyperphenylalaninemia no effect
Wrong.jpeg
GLU330ASP disease causing effect
True.jpeg
New.jpeg
GLY337VAL disease causing effect
True.jpeg
ARG408TRP disease causing effect
True.jpeg

ARG158GLN

<figure id="fig:ARG158GLNmutation">

The bindingsites are colored according to the overview picture above the mutation (ARG158GLN)is colored in red, and the sticks for this residue as well as for the binding residues are shown as well.

</figure>This mutation, which is diseasecausing is in a moderate distance to the important binding sites (<xr id="fig:ARG158GLNmutation" />). A clear explanation, why this has an effect on the protein can not be found from the structure alone. But even with the results from last weeks Task, we just had to rely on the clashes which occurred when we estimated the structure.

GLN172HIS

<figure id="fig:ARG158GLNmutation">

The bindingsites are colored according to the overview picture above the mutation (GLN172HIS) is colored in red, and the sticks for this residue as well as for the binding residues are shown as well.

</figure>In <xr id="fig:ARG158GLNmutation" /> one can see the big distance between the binding sites and the mutation location. Since we know that this is harmless, one would tend to say, that its clear, due to the distance, but as we know, that we also have mutations in a big distance which cause the disease, we are bound to say we do not know the mechanism.

ARG243GLN

<figure id="fig:ARG243GLNmutation">

The binding sites are colored according to the overview picture above the mutation (ARG243GLN) is colored in red, and the sticks for this residue as well as for the binding residues are shown as well.

</figure>This is the second mutation of this kind, which means a change from arginine to glycine. As the other mutation, this one is disease causing. But in difference to the other case, this one is rather close to the binding areas (see <xr id="fig:ARG243GLNmutation" /> )

LEU255SER

<figure id="fig:LEU255SERmutation">

The binding sites are colored according to the overview picture above the mutation (LEU255SER)is colored in red, and the sticks for this residue as well as for the binding residues are shown as well.

</figure>When one compares this mutation ( in <xr id="fig:LEU255SERmutation" />) to the reference structure in <xr id="fig:structuremoving" /> one can see, that one of the yellow residues (BH4 binding site) is mutated. And as one would expect this mutation is harmful.

MET276VAL

<figure id="fig:MET276VALmutation">

The binding sites are colored according to the overview picture above the mutation (MET276VAL) is colored in red, and the sticks for this residue as well as for the binding residues are shown as well.

</figure>As the mutation shown in <xr id="fig:MET276VALmutation" /> is rather far from any important binding site, one would expect it to be of little effect for the protein, which is the case for this mutation.

ARG297CYS

<figure id="fig:ARG297CYSmutation">

The binding sites are colored according to the overview picture above the mutation (ARG297CYS) is colored in red, and the sticks for this residue as well as for the binding residues are shown as well.

</figure>As the last mutation, this one is located quite far from any annotated sites ( <xr id="fig:ARG297CYSmutation" />), however it is located in the kink of two helices, and therefore the expectation which this mutation might have on the protein tends towards disease causing, which is actually the right choice.

ALA322GLY

<figure id="fig:ALA322GLYmutation">

The binding sites are colored according to the overview picture above the mutation (ALA322GLY) is colored in red, and the sticks for this residue as well as for the binding residues are shown as well.

</figure> The mutation, which affects one of the amino acids responsible for the BH4-binding, is shown in <xr id="fig:ALA322GLYmutation" /> and as one would guess is disease causing.

GLU330ASP

<figure id="fig:GLU330ASPmutation">

The binding sites are colored according to the overview picture above the mutation (GLU330ASP) is colored in red, and the sticks for this residue as well as for the binding residues are shown as well.

</figure>As the last mutation in <xr id="fig:ALA322GLYmutation" /> this mutation affects a binding site (this time iron) and is disease causing.

GLY337VAL

<figure id="fig:GLY337VALmutation">

The binding sites are colored according to the overview picture above the mutation (GLY337VAL) is colored in red, and the sticks for this residue as well as for the binding residues are shown as well.

</figure>This mutation (shown in <xr id="fig:GLY337VALmutation" />) is far from any known binding site and located in a loop on the far outside of the protein, but it is disease causing.

ARG408TRP

<figure id="fig:ARG408TRPmutation">

The binding sites are colored according to the overview picture above the mutation (ARG408TRP) is colored in red, and the sticks for this residue as well as for the binding residues are shown as well.

</figure>In <xr id="fig:ARG408TRPmutation" /> one can again see a mutation which is pretty far from the annotated sites, but still is disease causing.


SCWRL

preparations

We use the repairPDB script to extract the sequence from our model file. Afterwards we do some bashing to get the sequence in small letters but for the mutation, which has to be in capitals. <source lang="bash"> $ repairPDB 1J8U.pdb -seq > 1J8U.seq $ tr '[:upper:]' '[:lower:]' < 1J8U.seq > lower18JU.seq $ sed 's/^\(.\{<pos-118>\}\)<original in small letter>\(.*\)/\1<target in capital letter>\2/' lower18JU.seq > lower18JUmut<pos>.seq </source>

usage

Afterwards, we used a little bashscript which just applies all SCWRL-calls after another. In the following you see one example: <source lang="bash"> $ /opt/SS12-Practical/scwrl4/Scwrl4 -i ../1J8U.pdb -s lower18JUmut158.seq -o mut158.pdb > mut158.log </source>

results

The following pictures show the structure optimized by SCWRL. The mutant is colored in red for this residue as the wildtype is colored in green, as the structure apart from this residue is unchanged, the color is chosen automatically by PyMol. Polar bonds are colored respectively to mutant and wildtype, but note, that there might be overlaps, where only one color is shown.


<figtable id="tab:scwrlenergies">

Comparison of all the energies for the mutations and wildtype for scwrl
Mutation energy energy mut/energy wt prediction
wildtype 164.21 1 no effect
ARG158GLN 172.845 1.05 effect
GLN172HIS 169.699 1.03 no effect
ARG243GLN 152.987 0.93 effect
LEU255SER 166.123 1.01 effect
MET276VAL 164.472 1.00 effect
ARG297CYS 166.646 1.01 no effect
ALA322GLY 164.285 1.00 no effect
GLU330ASP 171.68 1.04 effect
GLY337VAL 164.304 1.00 no effect
ARG408TRP 393.804 2.39 effect

</figtable>

foldX

The other approach we used is foldX

preparation

We use the run.txt from the example at the test and adapt the list and individual list as its shown in the journal.

usage

<source lang="bash">

scwrl -runfile run.txt

</source>

results

As we already have results from the SCWRL run, we just compare those with the new results from foldX. The comparison with the sourounding structures is left for the reader as an exercise. The residue placed by SCWRL is colored in blue whereas the foldX residue is orange. THe surrounding structure is colored randomly by PyMol. <figure id="foldxstructcompare">

In this Figure, we show the differences in the structures which we got from both methods, where scwrl is colored in blue and foldX is colored in cyan. This is the only change and we expect it to have a very low effect on the activity of the protein, because no residues which are known for their importance are found in this region

</figure>


<figtable id="tab:foldxenergies">

Comparison of all the energies for the mutations and wildtype for foldX
Mutation energy energy mut/energy wt energy mut/energy relative wt prediction
wildtype 14.00 1 1 no effect
ARG158GLN 6.36 0.45 1.31 effect
GLN172HIS 16.69 1.16 0.99 no effect
ARG243GLN 6.15 0.43 0.72 effect
LEU255SER 23.57 1.68 1.21 effect
MET276VAL 21.99 1.57 1.06 effect
ARG297CYS 20.31 1.45 1.01 no effect
ALA322GLY 21.76 1.55 1.05 no effect
GLU330ASP 19.65 1.40 1.02 effect
GLY337VAL 24.01 1.71 1.10 effect
ARG408TRP 34.36 2.45 1.58 effect

</figtable>

minimise

preparation

To use minimise we have to clear the structure of hydrogen atoms and solvent which we do again using the repairPDB script. <source lang="bash"> $ repairPDB 1J8U.pdb -noh -nosol > 1J8U_pure.pdb </source>

usage

Afterwards a little script is used to apply the minimization five times and each time using the output as input.

  • usage:

<source lang="bash"> $ runAllMinimiser.sh <filestem> </source>

results

Just to see the differences between the results of one run and several using the output again as input, we produce this little animated .gif, which is basicly all the results from the 18JU.pdb on till the fifth minimise result. <figure id="fig:helixminimise">

Superimposition of all structures derived from minimise for 1J8U which means the wildtype. For a better comparison first the wildtype is shown in blue. Then it will get transparent and the first result of minimize is shown in teal. Then the transparent wildtype disappears and the first result will get transparent and the second result will show. This continues until the final result which is red. Then the animation will start again.

</figure> In <xr id="fig:helixminimise" /> one can see that minimise shifts the start of the helix further down in the picture, and once it reaches a certain point, the start of the helix is converted to a coiled region instead. A similar effect can be found if one compares sheets in the output of the different minimise runs. Here a twisting of the sheets takes place, which can bee seen in <xr id="fig:sheetminimise" /> <figure id="fig:sheetminimise">

Structure of 1J8U before (cyan) and after (red) minimise. The red structure is the result of five consecutive runs of minmise, and shows a twist in the orientation of the sheet.

</figure> <figure id="fig:timecourseminimise">

This picture shows the time course of the energy derived from minimise for 14 steps for the 1J8U-structure we got from the pdb and excluded the solvent

</figure> As one can see in <xr id="fig:timecourseminimise" /> consecutive multiple minimise runs do not decrease the energy but increase it drastically. But the first two minimise steps reduce the energy, and therefore we want to compare the energyoutput from minimise to the two others found in <xr id="tab:foldxenergies" /> and <xr id="tab:scwrlenergies" />. <figtable id="tab:minimiseenergies">

Comparison of all mutations with energies derived from minimise
mutation energy foldX energy foldX/energy wt energy scwwrl energy scwrl/energy wt prediction
wildtype -7516.267118 no effect
ARG158GLN -7564.267932 1.00 -7400.825142 0.98
GLN172HIS -7513.718344 0.99 -7375.087666 0.98
ARG243GLN -7522.768865 1.00 -7476.255952 0.99
LEU255SER -7522.768865 1.00 -7476.255952 0.99
MET276VAL -7502.369873 0.99 -7374.091542 0.98
ARG297CYS -7616.286033 1.01 -7484.442763 0.99
ALA322GLY -7509.307761 0.99 -7389.927555 0.98
GLU330ASP -7494.936652 0.99 -7367.788205 0.98
GLY337VAL -7504.533976 0.99 -7380.216180 0.98
ARG408TRP -5289.743707 0.70 -5438.301688 0.72 effect

</figtable> TODO: this and maybe compare with the other weird energy outputs from SCWRL and foldX und dann halt irgendwie daraus die prediction für die einzelnen Mutationen ableiten

surface considerations

As we only can get very little information from the minimise-results, we try to include the overall surface of the protein as well as the important sites into the prediction of the effectivity. In the following, mutants are colored red and wildtype residues green. The wildtype surface is shown in green and grey (for non mutant areas). In contrast the mutant surface will be shown as a mesh and in red color.

<figtable id="tab:scwrlenergies">

Comparison of the surface changes with respect to known binding sites and location
Mutation surface changes bindingsite prediction
ARG158GLN mild to big no effect
GLN172HIS mild no no effect
ARG243GLN big no effect
LEU255SER mild yes effect
MET276VAL big no no effect
ARG297CYS big no effect
ALA322GLY mild yes effect
GLU330ASP mild yes effect
GLY337VAL big no effect
ARG408TRP huge no effect

</figtable>

Gromacs

We choose the scwrl structures as input, since there is little difference to observe so far, but scwrl will be used in next week's task and results will be more comparable that way. The provided fetchPDB-script just wgets the gzipped PDBfile for a given PDB identifier, but was not used, since we already had our input structures. To extract the protein alone from the PDB-file, we used repairPDB with the -jprot option. repairPDB provides a number of usefull options, e.g. it extracts or removes DNA, crystal water, selects individual chains or renumbers the sequence.

The minimization preparation and runs were performed with this script. There is also a short script to run g_energy from the .edr-files here. We initially choose the forcefields TIP3P for water, AMBER03 (option 1 in the interactive menu), CHARMM27 (option 8) and GROMOS96 53a6 (option 13) for the protein, since they are amongst the most popular, but run into troubles with GROMOS and replaced it with OPLS-AA (option 14). All of them work with our chosen water model TIP3P. We used the provided standard configuration file for gromacs, described elsewhere.


Results

Runtime

For most mutants and most forcefields the minimization converges between step 250 and 300 and calculation stops then, so there is little sense in starting with a value of more then say 1000 for nsteps. If the energy does not converge by then, we'd suspect some error in the input. We give a few runtimes for the wildtype and the AMBER03 forcefield in <xr id="fig:runtime"/>, performed on the lrz grid. The maximum runtime we observed was around 50 seconds (not shown), the average runtime for a minimization is around 30 seconds. Since the calculation of every step requires the same amount of work, the expected runtime behaviour until convergence is perfectly linear in nsteps plus a constant for initialization. The provided datapoints confirm this nicely. <figure id="fig:runtime">

Runtime for different values for nsteps of minimization with Gromacs, using the AMBER03 forcefield and our wildtype 1J8U.

</figure>

Energies

We show the curve of the angle, bond and potential energy during the wildtype minimization for all three forcefields in <xr id="fig:1J8U_energies"/>. In all cases, the energy goes down steeply at first. While the bond energy does not change much after that and potential energy slowly converges, the angle energy rises during optimization until it also converges at a level slightly higher than its initial lowest point. This is as expected, since the total energy is optimized by adjusting the conformation of the protein. The mutants behave qualitatively similar, plots will be available here, once they are uploaded.

The tables <xr id="tab:gromacs_energies_1J8U"/> to <xr id="tab:gromacs_energies_mut408"/> show the energies of angle, bond and the potential energy of the wildtype and the ten mutants. In brackets, the difference to the wildtype is given. The bond energy comprises a small part of the total energy and changes not much in the mutants (the largest change in ARG158GLN is 2.6%). Neither direction nor amplitude of the energy changes seem to correlate with malignance of the mutations directly.

For the AMBER force field the average change in potential energy for the mutants is +304 kJ/mol, the average change in benign or hyperphenylalanemia causing mutations is only +139 kJ/mol. On the other hand, the smallest change in energy is observed in a disease causing, the GLY337VAL mutation. All but one disease causing mutation increase the potential energy, but two of the three benign ones do the same.

So far, we observe that smaller changes in energy indicate a less severe impact on the structure, but the prediction value appears limited, since the false positives number (benign mutations thought to be malign) is high. Setting a cut-off of the potential energy change to 100 kJ/mol would correctly classify five disease causing mutations above and two benign mutations below this threshold, giving an accuracy of 70%.

For the CHARMM force field the average change in potential energy for the mutants is +712 kJ/mol, the average change in benign or hyperphenylalanemia causing mutations is only +101 kJ/mol. The largest change in a benign mutation is only +160.53 kJ/mol and there are four extreme changes above 900 kJ/mol in the malign mutations. Setting a cut-off of the potential energy change to 100 kJ/mol would correctly classify six disease causing mutations above and two benign mutations below this threshold, giving an accuracy of 80%.

For the OPLS-AA force field the average change in potential energy for the mutants is +151 kJ/mol, the average change in benign or hyperphenylalanemia causing mutations is only +57 kJ/mol. The largest change in a benign mutation is +147.33 kJ/mol and there are three larger changes above 100 kJ/mol in the malign mutations, but also one change of +147 kJ/mol in an benign mutation. Setting a cut-off of the potential energy change to 50 kJ/mol would correctly classify five disease causing mutations above and two benign mutations below this threshold, giving an accuracy of 70%.

In all force fields, the largest change is caused by the malign ARG408TRP mutation, a visibly extreme disruption of the structure. Rather surprising is the large change in the benign GLN172HIS, that is located in a coiled region at the outside of PheOh and was one of the easier cases in last week's prediction task.

<figtable id="fig:1J8U_energies"> Course of the energies during minimization of 1J8U with gromacs.

Energie of the angles in 1J8U in the AMBER03 forcefield.
Energie of the bonds in 1J8U in the AMBER03 forcefield.
Potential energy in 1J8U in the AMBER03 forcefield.
Energie of the bonds in 1J8U in the CHARMM27 forcefield.
Potential energy in 1J8U in the CHARMM27 forcefield.
Energie of the angles in 1J8U in the OPLS-AA forcefield.
Energie of the bonds in 1J8U in the OPLS-AA forcefield.
Potential energy in 1J8U in the OPLS-AA forcefield.

</figtable>


<figtable id="tab:gromacs_energies_1J8U"> Energies after Gromacs minimization of the wildtype.

Type Amber03 Charmm27 Oplsaa
Angle (kJ/mol) 2403.47 (Δ 0.00) 2092.75 (Δ 0.00)
Bonds (kJ/mol) 522.31 (Δ 0.00) 1162.70(Δ 0.00) 545.17 (Δ 0.00)
Potential (kJ/mol) -37536.96 (Δ 0.00) -40392.46(Δ 0.00) -65323.58 (Δ 0.00)
</figtable>


<figtable id="tab:gromacs_energies_mut158"> Energies after Gromacs minimization of the disease causing ARG158GLN mutation.

Type Amber03 Charmm27 Oplsaa
Angle (kJ/mol) 2387.31 (Δ -16.17) 2087.86 (Δ -4.90)
Bonds (kJ/mol) 509.25 (Δ -13.06) 1138.40(Δ -24.30) 543.58 (Δ -1.59)
Potential (kJ/mol) -37171.73 (Δ 365.23) -39342.72(Δ 1049.75) -65223.57 (Δ 100.01)
</figtable>


<figtable id="tab:gromacs_energies_mut172"> Energies after Gromacs minimization of the benign GLN172HIS mutation.

Type Amber03 Charmm27 Oplsaa
Angle (kJ/mol) 2479.53 (Δ 76.06) 2106.61 (Δ 13.86)
Bonds (kJ/mol) 522.68 (Δ 0.37) 1154.75(Δ -7.95) 546.84 (Δ 1.67)
Potential (kJ/mol) -37163.15 (Δ 373.82) -40231.94(Δ 160.53) -65176.25 (Δ 147.33)
</figtable>


<figtable id="tab:gromacs_energies_mut243"> Energies after Gromacs minimization of the disease causing ARG243GLN mutation.

Type Amber03 Charmm27 Oplsaa
Angle (kJ/mol) 2398.46 (Δ -5.01) 2086.91 (Δ -5.84)
Bonds (kJ/mol) 517.89 (Δ -4.43) 1144.20(Δ -18.50) 542.42 (Δ -2.76)
Potential (kJ/mol) -37333.11 (Δ 203.86) -39440.42(Δ 952.05) -65231.15 (Δ 92.43)
</figtable>


<figtable id="tab:gromacs_energies_mut255"> Energies after Gromacs minimization of the disease causing LEU255SER mutation.

Type Amber03 Charmm27 Oplsaa
Angle (kJ/mol) 2415.15 (Δ 11.68) 2092.62 (Δ -0.13)
Bonds (kJ/mol) 530.22 (Δ 7.91) 1150.95(Δ -11.75) 546.59 (Δ 1.41)
Potential (kJ/mol) -37613.71 (Δ -76.75) -40227.79(Δ 164.67) -65323.13 (Δ 0.45)
</figtable>


<figtable id="tab:gromacs_energies_mut276"> Energies after Gromacs minimization of the benign MET276VAL mutation.

Type Amber03 Charmm27 Oplsaa
Angle (kJ/mol) 2402.78 (Δ -0.70) 2094.12 (Δ 1.37)
Bonds (kJ/mol) 522.62 (Δ 0.31) 1153.45(Δ -9.25) 548.90 (Δ 3.72)
Potential (kJ/mol) -37460.79 (Δ 76.18) -40308.56(Δ 83.90) -65297.64 (Δ 25.95)
</figtable>


<figtable id="tab:gromacs_energies_mut297"> Energies after Gromacs minimization of the disease causing ARG297CYS mutation.

Type Amber03 Charmm27 Oplsaa
Angle (kJ/mol) 2404.12 (Δ 0.65) 2088.02 (Δ -4.73)
Bonds (kJ/mol) 521.46 (Δ -0.86) 1140.64(Δ -22.06) 547.49 (Δ 2.32)
Potential (kJ/mol) -36943.84 (Δ 593.13) -39121.65(Δ 1270.81) -65074.67 (Δ 248.91)
</figtable>


<figtable id="tab:gromacs_energies_mut322"> Energies after Gromacs minimization of the hyperphenylalaninemia causing ALA322GLY mutation.

Type Amber03 Charmm27 Oplsaa
Angle (kJ/mol) 2409.61 (Δ 6.13) 2091.98 (Δ -0.77)
Bonds (kJ/mol) 528.05 (Δ 5.74) 1147.60(Δ -15.10) 547.21 (Δ 2.03)
Potential (kJ/mol) -37567.33 (Δ -30.36) -40334.32(Δ 58.14) -65325.79 (Δ -2.21)
</figtable>


<figtable id="tab:gromacs_energies_mut330"> Energies after Gromacs minimization of the disease causing GLU330ASP mutation.

Type Amber03 Charmm27 Oplsaa
Angle (kJ/mol) 2408.24 (Δ 4.77) 2097.44 (Δ 4.69)
Bonds (kJ/mol) 522.71 (Δ 0.39) 1162.17(Δ -0.53) 544.42 (Δ -0.75)
Potential (kJ/mol) -37373.77 (Δ 163.19) -40487.35(Δ -94.89) -65280.02 (Δ 43.57)
</figtable>


<figtable id="tab:gromacs_energies_mut337"> Energies after Gromacs minimization of the disease causing GLY337VAL mutation.

Type Amber03 Charmm27 Oplsaa
Angle (kJ/mol) 2415.06 (Δ 11.58) 2105.72 (Δ 12.97)
Bonds (kJ/mol) 526.54 (Δ 4.23) 1156.77(Δ -5.93) 548.96 (Δ 3.79)
Potential (kJ/mol) -37518.37 (Δ 18.60) -40272.05(Δ 120.41) -65273.33 (Δ 50.25)
</figtable>


<figtable id="tab:gromacs_energies_mut408"> Energies after Gromacs minimization of the disease causing ARG408TRP mutation.

Type Amber03 Charmm27 Oplsaa
Angle (kJ/mol) 2430.32 (Δ 26.85) 2110.42 (Δ 17.67)
Bonds (kJ/mol) 524.16 (Δ 1.84) 1150.43(Δ -12.26) 546.33 (Δ 1.15)
Potential (kJ/mol) -36674.55 (Δ 862.41) -38873.69(Δ 1518.78) -64802.38 (Δ 521.21)
</figtable>