Standards for hybrid and integrative methods Jill Trewhella, The - - PowerPoint PPT Presentation

Standards for hybrid and integrative methods

Jill Trewhella, The University of Sydney

Structural & biophysical methods for biological macromolecules in solution Sungkyunkwan University, Suwon Korea, June 19-26, 2016

• SAS as a constraint in hybrid modelling
• Challenge of validation of hybrid models
• SAXS/NMR and SAXS/docking –demonstrably

successful hybrid marriages

• The different regions of the SAS curve and

what we need to pay attention to

• Accurate molecular mass determination as an

important validation step; different methods

• Current state of draft publication guidelines
• The SAS Biological Data Base (SASBDB)

wwPDB SAStf Meeting report, Trewhella et al, Structure 21, 875, 2013 3D modeling from SAS (especially SAXS) data is becoming automated

• a global repository is needed that holds

standard format SAS data that is searchable and freely accessible for download;

• criteria need to be agreed upon for

assessment of the quality of deposited SAS data and the accuracy of SAS-derived models, as well as the extent to which a given model fits the SAS data;

• increasing diversity of structural biology

data and models calls for archiving

• ptions for models derived from diverse

data; and

• need for definition of what to archive in

the PDB and what complementary archives might be needed (taking into account both scientific needs and funding). wwPDB SAS task force

Outcome of the First wwPDB Hybrid / Integrative Methods Task Force Workshop

(at EBI UK, Oct. 2014)

• A. Sali, H.M. Berman, T. Schwede, J. Trewhella, G.

Kleywegt, S. K. Burley, J. Markley, H. Nakamura, P. Adams, A.M.J.J. Bonvin, W. Chiu, M. Dal Peraro, F. Di Maio, T.E. Ferrin, K. Grünewald, A. Gutmanas, R. Henderson, G. Hummer, K. Iwasaki, G. Johnson, C.L. Lawson, J. Meiler, M.A. Marti-Renom, G.T. Montelione,

• M. Nilges, R. Nussinov, A. Patwardhan, J. Rappsilber,

R.J. Read, H. Saibil, G.F. Schröder, C. Schwieters, C.A.M. Seidel, D. Svergun, M. Topf, E.L. Ulrich, S. Velankar, and J.D. Westbrook Structure 23, 1156-1167, 2015

• Models with all relevant experimental data and metadata, as well as experimental and

computational protocols should be archived; inclusivity is key.

• A flexible model representation needs to be developed, allowing for multi-scale

models, multi-state models, ensembles of models, and models related by time or

• ther order.
• Procedures for estimating the uncertainty of integrative models should be developed,

validated, and adopted.

• A federated system of model and data archives should be created.
• Publication standards for integrative models should be established.

Structural information Method Atomic structures of components X-ray/neutron Xtalography, NMR, 3DEM, comparative modeling, molecular docking 3D maps and 2D images Electron microscopy and tomography Atomic and protein distances NMR, FRET, DEER, EPR, other spectroscopic techniques; chemical crosslinks/mass spectrometry, disulfide bonds/ gel electrophoresis Binding site mapping NMR spectroscopy, mutagenesis, FRET Size, shape, P(r) SAS Shape and size Atomic force microscopy, ion mobility mass spec., fluorescence correlation spectroscopy and fluorescence anisotropy Component positions Super-resolution optical microscopy, FRET imaging Physical proximity Co-purification, native mass spectrometry, genetic methods, and gene/protein sequence covariance Solvent accessibility Footprinting methods, e.g. H/D exchange by mass spec. or NMR, functional consequences of point mutations Proximity of genome segments Chromosome Conformation Capture and other data Propensities for different interaction modes Molecular mechanics force fields, potentials of mean force, statistical potentials, and sequence co-variation

Comparison of structures for 82 kDa Malate Synthase G from NMR-only data and joint fit of SAXS-NMR data

• NMR/SAXS refinement improves backbone rmsd values with respect

to the crystal structure from 4.5 to 3.3 Å, largely due to more accurate translational positioning of domains

• The mid-q scattering range had most influence

Grishaev et al.J.Biomol. NMR 376, 95, 2008 NMR only SAXS-NMR

 3.05 NMR 1.01 Xtal 0.97 NMR/SAXS

• SAS/NMR co-refinement has been fully

implemented in Xplor-NIH

• Implemented for both X-ray and neutron

scattering

• Test examples for proteins, DNA and

flexible systems detailed in:

 Schwieters and Clore (2014) Using small-

angle solution scattering data in Xplor-NIH structure calacutions. Prog. Nucl. Magn.

• Reson. Spectrosc. 80, 1-11

Red lines depict backbone coordinates of the lowest energy 10 structures calculated omitting SAXS/WAXS data (Panel A) and including SAXS/WAXS data (panel B). Blue cartoon is representation of the X-ray structure from PDB ID 193L.

Lysozyme example

Figure 2: Schwieters and Clore (2014) Prog. Nucl. Magn. Reson. Spectrosc. 80, 1

Figure 3 from Schwieters and Clore (2014) Prog. Nucl. Magn. Reson.

• Spectrosc. 80, 1.

Comparison of SAXS/WAXS curves for lysozyme. Panels depicting the agreement to experiment of the SAXS/WAXS curves associated with the 10 lowest energy structures calculated without (panel A) and with (panel B) SAXS/WAXS data, respectively. The experimental data is shown in black with gray vertical bars equal to 1 SD; the curves calculated from the simulated annealing structures are shown in

• red. The residuals, given by, are

plotted above each panel.

Structure statistics for 10 lowest energy lysozyme structures w and w/o SAXS

all NMR data deposited NMR structures 1E8L (model 49) X-ray structure without SAXS with SAXS NOE violations 4.3 ±2.5 0.2 ±0.4 0.0±0.0 RDC R-factor, medium 1% 9.9 ±1.5 5.9 ±0.3 5.9±0.3 13.3 RDC R-factor, medium 2% 13.8 ±2.4 9.2 ±0.8 5.7±0.4 15.2 dihedral violations 4.4 ±1.2 0.1 ±0.3 0.0±0.2 SAXS χ2 2.3 ±1.4 0.4±0.1 1.7±0.6 0.86 HBDB energy (kcal/mol) 116.4 ±18.8 −160.7 ±13.3 −45.6±9.1 (−36.2) −255.12 torsion DB violations 3.8 ±2.6 1.4 ±1.5 1.7±0.9 (2) bond violations 6.5 ±5.2 0.4 ±0.8 0.0±0.0 angle violations 8.6 ±5.9 0.1 ±0.3 10.3±0.8 (10) 48 improper violations 3.1 ±2.7 0.0±0.0 0.0±0.0 24 bad non-bonded contacts 21.2 ±7.5 2.7 ±1.8 166.5±7.8 (176) 48 precision to mean (Å) 2.30±0.58 0.84±0.14 0.50±0.13 − precision to mean (Å) 2.78±0.56 1.50±0.14 0.52±0.20 1.46 C rmsd to X-ray struct. (Å) 2.82±0.59 1.32±0.19 1.48±0.10 (1.46) −

SAS improves the performance of docking methods, e.g.:

 pyDock, Jimenez-Garciaet al. (2015) Nucleic Acids Res 2015 43, W356.  FoXSDock, Schneidman-Duhovny et al. (2013) Biophys J 105, 962  HADDOCK, Karaca & Bonvin (2013) Acta Crystallogr D69, 683

Using a common set of 70 benchmark cases, success rates for combined docking and SAXS were:

 43% for pyDOCK,  48% for HADDOCK  63% for FoXSDock

compared with 23% when only SAXS data were used.

Schneidman-Duhovny et al. (2012) BMC Struct Biol 12, 17

q

Regions of the scattering curve

Size shape surface

LogI(q)

Guinier Plot  Rg and I(0) shape and size Kratky Plot  foldedness Porod Plot volume

q2 (Å-2) q (Å-1) q (Å-1) q2I(q) P(r) r (Å) q4I(q)

Lysozyme SAXS Data

Effects of more and less subtle inter- particle interference & aggregation

Jacques & Trewhella (2010) Protein Science 19, 642

Determining the size of the scattering particle: from I(0)

𝐽 𝑟 = 𝑂 Δ 𝜍𝑊 2𝑄 𝑟 𝑇(𝑟) For non-interacting particles and 𝑟 = 0 (i.e. 𝑇 𝑟 & 𝑄 0 are 1) 𝐽 0 = 𝑂 ∆ 𝜍𝑊 2 = 𝐷∆𝜍2 𝜑2𝑁 Where: C is the concentration in g/cm3 𝜑 is the partial specific volume in cm3/g M is the mass of the particle in g I(0) is in cm-1 Note: Doubling the concentration of the scattering particles will double the intensity, while doubling the size will quadruple the intensity.

Multiplication of the mass Mr by Avergadros number yields the molecular mass of the scattering particle:

𝑁𝑠 = 𝐽 0 𝑂𝐵 𝐷∆𝜍2 𝜑2

Data can be placed on an absolute scale using the well- characterised scattering of water (for both X-rays* and neutrons) This method requires an accurate concentration determination

*Orthaber et al. (2000) J. Appl. Cryst. 33, 218

→ NCVWeb Home → Contrast → Rg → Compost NCVWeb MULCh: ModULes for the analysis of Contrast variation data The purpose of this set of programs is to allow the extraction of structural parameters from Neutron Contrast Variation data for two component systems. There are three modules to facilitate this: 1.Contrast: This module determines the contrast ("scattering power" relative to the solvent) for each subunit in the complex for various proportions of D2O in the solvent. The contrast is important for the subsequent modules, but also helpful in planning and experiment. 2.Rg: Analyses the dependence of the radius of gyration upon contrast. From this analysis information can be extracted relating to the radii of gyration of each subunit and their separation. 3.Compost: Decomposes the scattering profiles down to scattering from each subunit, and a cross term, related to the disposition of each. Testing and basic description of the work has been published: A.E. Whitten, S. Cai and J. Trewhella, J. Appl. Crystallogr. If this page is used we ask that you cite that article. The MULCh manual can be found here The source code for the modules can be found here Test data and input files for MULCh can be found here

http://smb-research.smb.usyd.edu.au/NCVWeb/ Δ𝜍, 𝜑, and dry volume from sequence+ligands and solvent constituents

Hints for measuring protein concentration

• A280 measurement and calculated extinction coefficient

can be quite reliable if:

 Your spectrophotomoter is calibrated and you

measure within a concentration range that you have sufficient signal.

 You avoid agents in your buffer that absorb in the UV,

especially DTT – use TCEP instead (stable/non- absorbing).

 You check the absorbance for the folded and

unfolded form and adjust extinction coefficient If different.

 You perform multiple measurements.

Determining the size of your scattering particle; relative scaling

A popular alternative to placing data on an absolute scale is to use a known mono-disperse protein scatterer (such as lysozyme*, bovine serum albumin) as a standard with the relationship: 𝑁𝑠 𝑣𝑜𝑙𝑜𝑝𝑥𝑜 = 𝐽 0 𝑣𝑜𝑙𝑜𝑝𝑥𝑜𝐷𝑡𝑢𝑏𝑜𝑒𝑏𝑠𝑒 𝐽 0 𝑡𝑢𝑏𝑜𝑒𝑏𝑠𝑒𝐷𝑣𝑜𝑙𝑜𝑝𝑥𝑜 𝑁𝑠 𝑡𝑢𝑏𝑜𝑒𝑏𝑠𝑒 This method suffers from problems in finding a satisfactory standard as most readily available proteins aggregate with time and when subjected to intense radiation sources, and it assumes all proteins have the same 𝜑 and Δ 𝜍 which in fact are sequence dependent. Also requires accurate concentration determinations.

*Krigbaum and Kugler (1970) Biochemistry 9, 1216

Determining the size of your scattering particle; from Porod invariant

DATPOROD, implemented in AUTOGNOM (ATSAS package)

• Uses the relationship 𝑊 = 2𝜌2 𝐽(0)

𝑅

 Forces q-4 dependence for the high-q data  Uses the smoothed I(q) profile generated in GNOM to cope with the

noisy high-q data to determine the correction.

 Scattering data range set to 𝑟𝑛𝑏𝑦 =

8 𝑆𝑕

Petoukhov et al (2012) J. Appl. Cryst. 45, 342 See also: Fischer et al. (2010) J. Appl. Cryst. 43,101 (DATMOW) Rambo and Tainer (2013) Nature 496, 477 (DATVC) This method is subject significance uncertainties in the high-q dependence.

Porod law

• Is a description of the decay of the scattering profile

at high q

• For a folded particle with a sharp boundary (i.e. flat

surface), Porod's law predicts the scattering intensity: 𝐽(𝑟)~𝑇𝑟−4 where S is the surface area of the particle

• Note: biomolecules do not have generally sharp

boundaries and their state of “foldedness” varies.

Porod invariant, Q

𝑅 =

∞

𝑟2𝐽 𝑟 𝑒𝑟 = 2𝜌2

𝑊

𝑠

∆𝜍2 𝑠 𝑒𝑠

Requires an integration from 0 → ∞ , applies only to folded particles. Determining Q is experimentally fraught; we only measure from 𝑟𝑛𝑗𝑜 → 𝑟𝑛𝑏𝑦, high q decay may not obey Porod law, background subtraction errors can be confounding for low S/N data. Nevertheless, for a uniform scattering density folded particle:

𝑅 = 2𝜌2𝑊∆𝜍2 and as 𝐽 0 = ∆ 𝜍𝑊 2

𝑊 = 2𝜌2 𝐽(0) 𝑅

Kratky plot

𝑟2𝐽 𝑟 𝑤𝑓𝑠𝑡𝑓𝑡 𝑟 Used to assess degree of compactness or “foldedness” of the scattering particle. A folded protein decays as 𝑟−4, a Gaussian chain as 𝑟−2 folded globular unfolded partially folded

𝑟2𝐽 𝑟 𝑟2𝐽 𝑟 𝑟2𝐽 𝑟

𝑟 (Å-1) 𝑟 (Å-1) 𝑟 (Å-1)

Reliable scattering data are those you can demonstrate are from the particle you are interested and have been demonstrated to be free from instrumental and sample state biasing effects. Always measure a concentration series to assess potential concentration-dependent effects from inter-particle repulsive (non-unity S(q)) or weak attractive forces (mild aggregation) Always calculate the size of your scattering particle – preferably by multiple methods but best by placing data on an absolute scale and working to ensure you can accurately determine the concentration of your scattering particle.

Acta Crystallographica – notes for authors 11.3. Small-angle scattering data Guidelines for articles reporting structural modelling of small angle scattering may be found at http://journals.iucr.org/services/sas/. For articles that present experimental SAS data, the deposition of an ASCII file representing the background-corrected scattering profile(s) with errors is recommended.

draft requirements for presenting biological macromolecule small-angle scattering data It is not the intention of this document to define a quality requirement for SAS experiments that would be acceptable for publication. Rather, the purpose is to outline the way in which SAS experiments should be presented in order to enable the reader to independently assess the quality of any interpretations made by the authors.

Jacques et al. (2012) Acta Cryst. D68, 620

High sample quality is required for SAS experiments. As such the authors must demonstrate as clearly as possible that any sample is of sufficient quality to yield meaningful results in a SAS experiment.

Sample Quality

• A complete description of the sample; sequences (including

purification tags), modifications and cofactors.

• Sample purification procedure, along with estimate of final

purity and method for estimating.

• Solvent blank composition (including pH for aqueous

systems).

• Sample concentration(s) and method(s) of determination,

including extinction coefficients used.

• A SAS-independent assessment of monodispersity (DLS

and/or aggregate-free gel filtration and/or MALLS) where available.

• A statement describing how the solvent blank was obtained

(e.g. dialysis, column flow through).

• For SANS contrast variation experiments, a statement

describing the level of deuteration in biomolecules and their solvents and how it was determined.

Details on the execution of the SAS experiment

Data acquisition

• Instrument type (e.g. model or beamline) and configuration

(point or line source, collimation details, detector details). In the case of SANS there may be several configurations (e.g. detector positions, number of guides, apertures etc.).

• Wavelength (including Δλ/λ for neutrons) and q-range.
• Sample environment (including cell pathlengths, temperature)

and exposure times.

• Standards measured and controls [secondary scattering

standards for assessing I(0) data (e.g. lysozyme, water)].

• For X-rays, monitoring for radiation damage and means for

controlling (addition of scavengers, sample flow, analysis of time frames).

• Data reduction protocol and software.
• Where a line source is used, beam geometry must be provided

[either in terms of dimensions of a defined shape (e.g. parameters of a trapezoidal profile), or as a plot of the beam profile file].

In order for a reader to be able to assess the quality of a SAS experiment, it is necessary that the data be presented in a clear, well described manner.

Presentation of the scattering data

• Where possible, scattering profiles [I(q) versus q] and P(r)

profiles [P(r) versus r] should be reported in the bulk manuscript.

• Present either as linear q–log I(q) or log q–log I(q)
• I(q) on an absolute scale; multiple curves may be offset on

the same plot for clarity, provided that this is explained in the figure caption.

• For structural characterisation of isotropic samples, show

Guinier plots with q-range.

• Show Kratky plot, especially if using Porod volume

• Report if the presented data are desmeared (to correct for

beam geometry or polychromaticity), and desmearing method.

• Expected and scattering data derived molecular weight,

including all parameters (and their uncertainties) used.

• Record data at multiple solute concentrations to eliminate the

possibility of concentration-dependent oligomerisation or inter-particle interference.

• Report if there was no concentration dependence, otherwise

extrapolate to infinite dilution. Report all details for final data set(s) selection.

• For SAXS/SANS contrast variation, show
• Stuhrmann plots of Rg

2 versus the reciprocal of contrast and extracted

component scattering functions (including cross-term) where appropriate;

• nature and number of contrast points;
• plot of normalized ±√I(0) versus solvent density particle matching point.

Where the experimenter is looking to support a three- dimensional model, any modelling must be justified and described thoroughly.

Modelling

• Report all software used for modelling [including generating

P(r) profiles].

• Provide χ2 values and a plot of the model fit to the

experimental I(q) versus q.

• Present Analysis of the ambiguity of the reconstruction

(averaging or clustering).

• For rigid-body modelling, describe how the starting models

were obtained (e.g. crystal structure of a domain, homology model etc.) as well as any connectivity or distance constraints and how they were chosen.

Data collection parameters Structural parameters Instrument Rg (Å) from Guinier Beam geometry I(0) (cm-1) from Guinier Wavelength (Å) Molecular mass Mr (from I(0)) q range (Å-1) Rg (Å) from P(r) Exposure time (s) I(0) (cm-1) from P(r) Sample concentration range (mg mL-1) dmax (Å) Temperature (°C) Porod volume estimate (Å) Molecular mass determination Software employed Partial specific volume Primary data reduction Contrast (Δρx1010 cm-2) Data processing Mr from sequence (Da) ab initio analysis dry/hydrated volumes from sequence (Å3) Validation and averaging Computation of model intensities Rigid body modelling Generic Table

C5C6C7 C5C6*C7 tC5C6C7 tC5*C6C7 C6*C7 C7 Guinier results

• Expt. 1

Rg (Å) 37(1) 25(2) 15.3(1) qRg-range (Å-1) 0.65 -1.3 0.72-1.22 0.57-1.3

• Expt. 2

Rg (Å) 37(2) 38(2) 40(2) 44(2) qRg-range (Å-1) 0.54-1.3 0.54-1.3 0.56-1.16 0.56-1.16 P(r) results

• Expt. 1

Rg (Å) 39.38(9) 25.58(4) 15.53(3) dmax 145 90 60 q-range (Å-1) 0.017-1.189 0.028- 0.305 0.039- 0.251

• Expt. 2

Rg (Å) 39.38(7) 40.3(1) 42.6(5) 45.6(9) dmax 141 150 169 200 q-range (Å-1) 0.014- 0.193 0.015-0.193 0.018-0.193 0.02-0.193 Protein mg mL-1 range for extrapolation to infinite dilution

• Expt. 1

0.9-3.1 0.6-5.1 5.7-11.7

• Expt. 2

0.7-3.4 0.5-1.9 0.3-1.4 0.2-1.2 Molecular weight, MW from I(0) values*

• Expt. 1

37,427 23,249 11,803

• Expt. 2

38,204 35,416 39,708 37,055 Ratio of I(0) derived to expected MW 1.07 0.99 1.06 0.99 1.10 1.10

Nadvi et al. “Clinically-linked Mutations in the Central Domains of the Cardiac Myosin-binding Protein C Associated with Distinct Phenotypes show Differential Structural Effects,” Structure 24, 105-115, 2016

Data collection parameters

Instrumentation Australian Synchrotron SAXS beam-line Expt 1: 30 June 2014 Expt 2: 18 April, 2015 Beam geometry 250 µm × 150 µm 250 m x 150 m q-range meas (Å-1) 0.01-0.476 0.006-0.344 Exposure time (secs) 1 s x 24 1 s x 44 frames Temperature (oC) 22 22 Sample details C5C6*C7 C6*C7 C7 C5C6C7 C5C6*C7 tC5C6C7 tC5C6*C7 Partial specific volume (ν, cm3 g-1) 0.735 0.736 0.735 0.735 0.735 0.735 0.735 Contrast (Δρ, 1010 cm-2) 2.841 2.826 2.794 2.841 2.841 2.839 2.820 Molecular weight (MW) from seq. Da 35,709 21,498 10,734 35,738 35,709 37,576 37,519

Software employed for data reduction, analysis and interpretation

SAXS data reduction ScatterBrain Calculation of expected MW, Δρ and ν values MULCh SAXS data analysis ATSAS 2.6.0 SAS Data Analysis ab initio bead modelling DAMMIF (via ATSAS on-line) Atomic structure modelling SASREF (via ATSAS on-line) 3D graphic model reps PYMOL

Trewhella, J. “Small-angle scattering and 3D structure interpretation,”

• Curr. Opin. Struct. Biol. 40, 1–7, 2016.

http://www.sasbdb.org/

The purpose of a model is not to fit the data but to sharpen our questions Samuel Karlin

American mathematician who improved DNA sequencing analysis by applying mathematics to early efforts at genome sequencing Royal Society presentation, 20 April 1983

Remember that all models are wrong; the practical question is how wrong do they have to be to not be useful.

George E.P. Box Box and Draper, Empirical Model Building and Response Surfaces (1987), 74.

Application of a bad model tuned to limited data can yield dangerous results when applied outside the data bounds.

Ralph Nelson Multi-disciplinary scientist studying interactions in coupled nonlinear phenomena

A tricky example

Sda: small protein inhibitor of the bacterial histidine kinase, KinA Crystal structure indicates a stable monomer NMR data indicate a monomer SAXS data originally interpreted as indicating a dimer?

Whitten et al (2007) J. Mol. Biol. 368, 407

Crystal structure of a small histidine kinase inhibitor, Sda. Molecules A, B and C constitute the asymmetric unit. Molecules A*, B* and C* are related to the asymmetric unit by rotation about a 2-fold crystal symmetry axis (perpendicular to the middle of the image). Overlay of the 3 molecules in the asymmetric unit.

Jacques et al (2009) Acta D65, 574

CRYSOL fits to SAXS data with 2 values

Trimer from crystal structure is missing about 25% of the mass. Trimers have largest buried surface areas 1112 & 1081 Å2 for A-B-C and B-C-A* trimers, respectively.

Jacques et al (2009) Acta D65, 574

subunit A A-B-C-A*-B*-C* hexamer B-C dimer A-B* dimer A-B-C trimer B-C-A* trimer 0.85 2.0 19.3 1.1 5.6 2.2

CORAL 2 = 1.49 Crystal trimer + missing residues SASREF 2 = 0.88 (severe steric clashes) 3 NMR monomers SASREF 2 = 0.85 2 NMR monomers

Dimer pair 2 A-B 2.7 B-C 1.2 A-C 1.0 CRYSOL fits to SAXS data with dimer models created by aligning the NMR monomer with different Xtal dimers PISA stable assemblies: A-B, B-C, trimer, hexamer OLIGOMER fits to SAXS data with dimer and trimer models created by aligning the NMR monomer with Xtal structure 2 0.85 Mixture 2 Hex-Mon 1.3 Tri-Mon 1.0 A-B+B-C 1.0