Structure Validation Tutorial

Sbcc

Informations

• Publié par: Nichor
• Nombre de lectures: 73
• Langue: English

Extrait

Updated 4/26/06 HLS

Structure Validation Tutorial

Brian Kelly

November 12, 2004

Now that you have solved a structure, how do you know that it is correct (explains your

data as well as possible, without bias)? You need to validate your work. This can seem

overwhelming at times, but it is absolutely necessary. You don’t need many materials to

validate your structure, basically a pdb, your scalepack file (.hkl or .sca), and an mtz. I

will use examples from structures solved from different people in the lab to help illustrate

how structures are validated.

You will be required to run a program called ADIT at the PDB before depositing your

data (both pdb and structure factor files). ADIT basically runs the same validation

programs as PROCHECK. You can do a trial run of ADIT before actually depositing

your structural information.

Many tools are available- you can access most of them from our lab links webpage, but

here is a list as well:

Biotech validation server

(EBI) - performs any or all of the quality checks provided by

the programs WHAT IF and PROCHECK

Structure validation central

(JCSG) - runs PROCHECK, SFCHECK, PROVE, ERRAT,

WASP, DDQ, WHATCHECK and PSQS

ADIT server

(RCSB) - runs PROCHECK, NUCheck and (if structure factors in mmCIF

format are provided) SFCHECK

MolProbity server

(various checks; Duke)

STAN (STructure ANalysis) server

(Uppsala) - produces Ramachandran and "CA-

Ramachandran" plots for proteins and Duarte-Pyle plots for nucleic acids. Moreover, it

runs the program WASP to check if any water molecules could be small cations instead

(Na+, Li+, Mg+2, Ca+2). It also runs the program CISPEP to investigate if any non-

proline peptide bond that has been modelled as

trans

could in reality be

cis

.

Verify3D server

(fold; UCLA)

ERRAT server

(contacts; UCLA)

Ramachandran server

(Bangalore) Slow, maybe not worth it.

RamPage server

(Ramachandran server; Cambridge)

VADAR server

(Alberta, Canada)

PARVATI server

(anisotropic temperature factors; Washington)

ANOLEA server

(Atomic Non-Local Environment Assessment; Namur, Belgium)

TB Consortium Bias Removal Server

(Texas A&M University) - generates maps that are

as little biased as possible and calculates real-space fit values

Electron Density Server

(Uppsala University) A very useful site with data on all pdb

entries. Good place to start with a molecular replacement model.

SAVS

(UCLA) Runs many different programs, including ERRAT and Verify3D. Good

place to check many features of model in one place.

1

Updated 4/26/06 HLS

To go over examples of each of the above programs would take many lengthy pages, but

I will go over the two suites I prefer because they are comprehensive and relatively easy

to use:

1.

Structure Analysis and Verification Server

The first suite is called

Structure Analysis and Verification Server

.

It is run by a

research group at UCLA, and the web address is:

http://shannon.mbi.ucla.edu/DOE/Services/SV/

This website will run programs called ERRAT, Verify-3D, Prove, Procheck,

What_Check (aka Whatif), and SFCheck.

You will need your pdb file and your mtz file.

On the home page, upload your pdb and mtz. (you can use your gun4.pdb and gun4.mtz

from the refinement tutorial)

Click Upload files.

You will have to choose which lines in the mtz correspond to F (amplitudes of the

observed data), SigF (estimated standard deviation on F), and Free (free R flag). The

SAVS will attempt to highlight these for you, depending on the labels it culls from the

mtzdump operation. Select the correct radio buttons and click “run all programs”. It will

take awhile for these programs to run (roughly 4 minutes for a smaller structure, and up

to 30 minutes for a larger structure).

Look at the output files. As you can see, there are different colors, which make

interpreting the results a little easier. The following table describes each section.

SFCHECK (Vaguine, Richelle et al. 1999)

Amino acid positions in the structure get a red flag if

1) shift of backbone is greater than 1.5

2) density correlation of backbone is less than 0.90

3) density index of backbone is less than 0.5

4) B-factor of backbone is greater than 60 Angstroms squared

5) connectivity is less than 65%

For each test, a green rating is given if > 95% of the positions

pass the test.

GOOD:

SATISFACTORY

Yellow rating if 90 - 95% of the positions pass the test.

WARNING:

INSPECTION

SUGGESTED

Red rating if fewer than 90% of the positions pass the test.

ERROR:

INSPECTION

RECOMMENDED

PROCHECK (Laskowski, Moss et al. 1993)

One of the many files produced from running

PROCHECK

is

GOOD:

2

Updated 4/26/06 HLS

the summary file: file.sum. The first column in this file for

each line indicates 1 of 3 things (as described in the file itself):

Nothing or Blank

SATISFACTORY

+

WARNING:

INSPECTION

SUGGESTED

*

ERROR:

INSPECTION

RECOMMENDED

WHAT_CHECK (Hooft, Vriend et al. 1996)

The output file produced from running

WHAT_CHECK

is

divided up into sections. Each section is identified by 1 of 3

words: NOTE

GOOD:

SATISFACTORY

WARNING

WARNING:

INSPECTION

SUGGESTED

ERROR

ERROR:

INSPECTION

RECOMMENDED

ERRAT (Colovos and Yeates 1993)

The data plot produced from running

ERRAT

contains a

percentage of the sequence that is below the 95% limit for each

chain in the input structure. This percentage is reported and no

color evaluation is made by SAVS. A large

ERRAT

score is a

good thing.

VERIFY_3D (Luthy, Bowie et al. 1992)

Using the averaged data points produced for each amino acid

in the sequence, the number of times the value is greater than

0.2 is converted into the percentage of the sequence that has

positions with values > 0.2. If this percentage is < 20%

GOOD:

SATISFACTORY

20 - 45%

WARNING:

INSPECTION

SUGGESTED

>= 45%

ERROR:

INSPECTION

RECOMMENDED

PROVE (Pontius, Richelle et al. 1996)

In the section delimiter identified with "**OUTLIERS", the

percent of the number of buried outliers is given. If this

number is < 1.0%

GOOD:

SATISFACTORY

1-5%

WARNING:

INSPECTION

SUGGESTED

3

Updated 4/26/06 HLS

> 5%

ERROR:

INSPECTION

RECOMMENDED

Each section will teach you something useful. The take-home message would be that if

you see red on your output, click on it, read the error message, and look at the

corresponding area of your structure. You will need to make sure you either have a valid

reason for your model having the features that caused the error message, or fix the region

that needs fixing.

For example, Verify 3D will provide you with a nice graph (3D-1D) profile of your

structure. This program analyzes the compatibility of an atomic model (3D) with its own

amino acid sequence (1D). Each residue is assigned a structural class based on its

location and environment (alpha, beta, loop, polar, apolar etc). Then a database

generated from vetted good structures is used to obtain a score for each of the 20 amino

acids in this structural class. For each residue, the scores of a sliding 21-residue window

(from -10 to +10) are added and plotted. The returned 3D-1D profile should for the most

part stay above 0.2 and never really dip below 0. Verify 3D has been useful in

identifying register errors (shifts) in models.

ERRAT will give you an “overall quality factor” and you should be seeing numbers in

the high 90% range if your structure is good. Below the plot, you will receive a small

text discussion regarding your results and how they compare to other models solved to

similar resolutions. On the contrary, if you receive a high value for error, you should go

look at these residues and make sure nothing is wrong, such as a buried charged residue,

or the like.

2. MolProbity

The next suite is located on the 3D Macromolecule Analysis and Kinemage Home Page.

The specific program you will want to use is called MolProbity. A research group at

Duke University runs it and its current web address is:

http://molprobity.biochem.duke.edu/

This site offers a few cool features including the ability to “flip” Asn, Gln, and His

residues for better hydrogen bonding throughout your structure. Looking at the density

alone, this can be challenging, but the Richardson’s have created an algorithm to look at

neighboring residues as well as solvent and scoring which decides how the greatest

amount of bonding can occur in a structure. Additionally, you can visualize clashes

between atoms after running the tools available on the site.

MolProbity

also offers Ramachandran plots, rotamer analysis and C

β

deviations.

There is a tutorial for MolProbity if you would like to run through a trial before using

your own data (also located in this directory as a PDF (

MolPobity_tutorial.pdf

)

4

Updated 4/26/06 HLS

MolProbity allows you to load electron density maps as well, currently in the “O” format

as well as X-PLOR ASCII and CCP4 (mode 2 only). This is a nice feature so you can

visualize the results and your data to determine if the program has given you a reasonable

error message. The only caveat is that since it is a java program, and runs in your web

browser, it can be a little on the slow side.

Interpreting the results from the all-atom contact test can be a little challenging. I have

taken the following text directly from the paper used to describe MolProbity:

With all hydrogens present, all-atom contacts are then calculated by P

ROBE

, which uses

traditional van der Waals radii for most atoms and 1.0 Å for polar H, in a rolling-probe

algorithm that leaves a dot when the 0.25 Å-radius probe intersects another not-

covalently-bonded atom. The results include a clustered list of disallowed atom pair

overlaps 0.4 Å, an overall clash score (number of bad overlaps per 1000 atoms) and two

kinemage graphics displays of contacts for the whole structure. Van der Waals contacts

are shown as back-to-back patches of green or blue dots on the surfaces of non-covalent

atom pairs within 0.5 Å of touching; hydrogen bonds are shown as lenses of pale green

dots outlining the interpenetrating surfaces of a donor and acceptor. Steric clash overlaps

are emphasized with spikes rather than dots, progressing from yellow to hot pink as the

clash becomes truly serious beyond 0.4 Å overlap. The reference for this paper is (Davis,

Murray et al. 2004).

After running these two suites of programs and investigating an errors or suggestions

with a critical eye, fixing what is necessary, you can rest well knowing that your structure

is generally good.

References

Colovos, C. and T. O. Yeates (1993). "Verification of protein structures: patterns of

nonbonded atomic interactions." Protein Sci

2

(9): 1511-9.

Davis, I. W., L. W. Murray, et al. (2004). "MOLPROBITY: structure validation and all-

atom contact analysis for nucleic acids and their complexes." Nucl. Acids Res.

32

(suppl_2): W615-619.

Hooft, R. W., G. Vriend, et al. (1996). "Errors in protein structures." Nature

381

(6580):

272.

Laskowski, R. A., D. S. Moss, et al. (1993). "Main-chain bond lengths and bond angles in

protein structures." J Mol Biol

231

(4): 1049-67.

Luthy, R., J. U. Bowie, et al. (1992). "Assessment of protein models with three-

dimensional profiles." Nature

356

(6364): 83-5.

Pontius, J., J. Richelle, et al. (1996). "Deviations from standard atomic volumes as a

quality measure for protein crystal structures." J Mol Biol

264

(1): 121-36.

Vaguine, A. A., J. Richelle, et al. (1999). "SFCHECK: a unified set of procedures for

evaluating the quality of macromolecular structure-factor data and their

agreement with the atomic model." Acta Crystallogr D Biol Crystallogr

55 ( Pt

1)

: 191-205.

5