X-ray Crystallography Lecture 4 (SL22026) PDF

SL22026 - Protein Structure and Analysis Continuation of X-ray crystallography Lecture-3  Structure Refinement Building the model by hand (or by computer) gives a reasonable model but not one that is necessarily the best possible fit to all the available information To...

SL22026 - Protein Structure and Analysis Continuation of X-ray crystallography Lecture-3  Structure Refinement Building the model by hand (or by computer) gives a reasonable model but not one that is necessarily the best possible fit to all the available information To improve our built model, we “refine” it Refinement seeks to find the model that best predicts our original observations while simultaneously satisfying what we know about the chemical structure of proteins (e.g. phenyl rings are flat)  Refinement of the Structure Restraints (secondary structure, hydrogen bonding, multiple copies of molecules in the unit cell) can be employed to prevent protein from flying apart or reduce noise One needs to judiciously choose which restraints to employ Within the refinement programs are physics & chemistry libraries based on known properties of proteins (bond angles, lengths, etc.) to further restrain atom sampling  What does Refinement do? In essence, refinement will move positions of protein atoms to minimize the following equation R = (|Fo-Fc|)/(Fo) F0 is observed amplitude (from the X-ray experiment) and Fc is calculated amplitude from the model Data fit term- the R-factor (reliability) Measures how closely the diffraction amplitudes predicted by our model matches the diffraction amplitudes we actually observed in the experiment  Difference Maps Used throughout the refinement process after a model has been built A way to remove potential phase bias and errors Types of maps commonly used: F0-Fc map 2F0-Fc map Omit map  Difference F -F map Maps 0 c F0 (native/observed) map differs from Fc (calculated) map where there are missing or wrongly placed atoms Produces positive/negative peaks in areas where F0 differs from Fc The refinement process ends when the F0-Fc map is essentially featureless 2F0-Fc map F0 plus a difference map F0-Fc Map should look like the corrected model Usually one examines a F0-Fc map to identify errors in a model and a 2F0-Fc map to guide the construction of the new model  Examp The purple is a 2F0-Fc le map and the blue peak is a “positive” difference peak generated by an F0-Fc map Positive densities are caused by structural features not presented in the model Negative densities are caused by features in the model that are not real  Examp Model does not fit the le electron density - must rebuild! e.g. Refinement got stuck 2F0-Fc map can be used to rebuild the structure and to allow refinement to continue If this were a F0-Fc map the electron density would be only observed where the model did not match the electron density  Omit Map Used to reduce the bias introduced by phases calculated from the model If the structure is refined with the atoms from the questionable region omitted from the model, the model bias can be reduced and more details of the correct chain trace can be observed To examine troublesome regions If interpretation of part of an electron density map is somewhat doubtful Surface loop cannot be traced satisfactorily Electron density map for only part of the structure that is known correctly Example of an ligand placement in the electron density omit map SL22026 - Protein Structure and Analysis X-ray crystallography – Lecture 4 Structure Validation, Terminology, Protein Data Bank (PDB) and a few examples  Assessing Overall Quality of Structures Quality criteria Resolution R-factor & R-free Geometry B-factors Other experimental data Does the model agree with biochemical and other data (mutagenesis, kinetics, spectroscopy etc.) Who checks the X-ray crystallographer? The reviewers and researchers The protein data bank (PDB) Competing groups working on similar structures Understandi ng and Evaluating Crystallogra phic Data; Moss et al (2018) Journal of Structural Biology, 204, 19-25 Resolut ion  The R- factor One of the key statistics for judging a structure’s quality Does the model reflect the actual experimental data? The residual (or fraction) of the data that the model does not explain Low resolution structures can be as high as 30% For exceptional sub-atomic resolution structures as low as 10% R-factor usually around 20-25%  The R- factor For a typical protein R-factor in the range of 20% -25% The fundamental reason for the difference is the crystal quality (purity of the sample and conformational flexibility of the molecule) and accuracy of the model (phasing quality and resolution)  R- free Calculated the same way as R-factor but only looks at a fraction of the data that has never been used to the refine the structure 5-10% of reflections removed randomly from the data set prior to refinement Reflections for entire dataset called work set or used Reflections for removed reflections called test or free  R- free Unbiased measure of the success of structural refinement The refined model has never seen the omitted data so the comparisons report an unbiased evaluation of the accuracy of the model Indicator of incorrect modelling when >> R-factor For good models usually no more than 5% higher than R-factor Gives a more objective measure of the quality of the model Not biased towards these reflections Avoids model bias and overfitting of the data  Geome try Model must have reasonable bond lengths, bond angles and overall geometric agreement compared to other well-defined structures Deviations for bond length