Overview of First CAPRI - LORIA · P3 1.8GHz Xeon: ˘ 106=s @ N = 16; ˘ 105=s @ N = 30. Summary of...
Transcript of Overview of First CAPRI - LORIA · P3 1.8GHz Xeon: ˘ 106=s @ N = 16; ˘ 105=s @ N = 30. Summary of...
Overview of First CAPRIBlind Protein Docking Experiment
Dave Ritchie
Department of Computing ScienceKing’s College, University of Aberdeen
Aberdeen AB24 3UE, UK.
Overview of First CAPRIBlind Protein Docking Experiment
Contents
• CAPRI Background
• Hex Docking Algorithm
• CAPRI Target Structures
• CAPRI Participants & Algorithms
• CAPRI Results
• Conclusions, Future Directions
CAPRI“Critical Assessment of Predicted
Interactions”
• CASP did have a Docking Section
• Docking was a side-show ?
• Too few targets, too easy/hard ?
• Hence CAPRI
Janin, Wodak, Sternberg, Moult,Ten Eyck, Vajda, Henrick
• To assess the state of the art
• To promote developments in the field
• Hosted by EBI (Henrick, Tate)
• 2001/2: 7 targets made available
• ∼ 20 International groups participated...
Lactobacillus HPrK / Hpr (CAPRI Target 1)
Hex: Protein Shape Complementarityusing Density Representations
τσ(r)
(r)
Surface Density: σ(r) =
{
1; r ∈ surface skin volume
0; otherwise
Interior Density: τ (r) =
{
1; r ∈ VDW volume
0; otherwise
Favourable:
∫
(σA(rA)τB(rB) + τA(rA)σB(rB))dV
Unfavourable:
∫
τA(rA)τB(rB)dV
Penalty Factor: Q = 11
Score: SAB =
∫
(σAτB + τAσB − QτAτB)dV
Spherical Polar Shape Reconstruction
(Looking at MCV Antibody CDR Loops)
τ (r) =N
∑
nlm
aτnlmRnl(r)ylm(θ, φ)
Image Expansion Coefficients
A Atom Gaussians -
B N = 16 1,496
C N = 25 5,525
D N = 30 9,455
Correlations - Overlap as a Function ofCoordinate Operations
Basic Overlap:
∫
σA(r)τB(r)dV =
N∑
nlm
aσnlmbτ
nlm
Rotation: R(α, β, γ)σA(r) =N
∑
nlm
aσ′nlmRnl(r)ylm(θ, φ)
Translation: T (R)σA(r) =N
∑
nlm
aσ′′nlmRnl(r)ylm(θ, φ)
Rotated Coeffs: aσ′nlm =
l∑
m′=−l
R(l)mm′(α, β, γ)aσ
nlm′
Translated Coeffs: aσ′′nlm =
N∑
n′l′
T(|m|)nl,n′l′(R)aσ
n′l′m
Hence:
∫
σ′A(r)τ ′′
B(r)dV =N
∑
nlm
aσ′nlmbτ ′′
nlm etc.
Search: Calculate for ∼ 109 orientations
P3 1.8GHz Xeon: ∼ 106/s @ N = 16; ∼ 105/s @ N = 30
Summary of Hex Docking Strategy
• Contour VDW & SAS surfaces – “marching tetragons”
• Sample surface volumes onto (0.6A)3 grid
• Calculate shape & electrostatic coefficients – once
• Scan search space with N=16 correlations (shape only)
• (Optionally constrain search to one or both epitopes)
• Re-score top 10,000 with N=30 (+electrostatics)
• Refine top 500 by soft OPLS minimisation
• Cluster solutions
• Output best member of each cluster
• Pray!
Docking Very Large Molecules
The 7 CAPRI Docking Targets
Target Receptor Ligand Type Complex Lab
1 HPr Kinase HPr U/U Fieulaine et al. Janin
2 Rotavirus VP6 MCV U/B Vaney et al. Rey
3 Hemagglutinin HC63 U/B Barbey-Martin et al. Knossow
4 α -Amylase AMD10 U/B Desmyter et al. Cambillau
5 α -Amylase AMB7 U/B Desmyter et al. Cambillau
6 α -Amylase AMD9 U/B Desmyter et al. Cambillau
7 SpeA TCR 14.3.D U/U Sundberg et al. Mariuzza
• At least one docking partner presented in its unbound form
• Participants permitted 5 attempts for each target
• Any predictive approach allowed: homology/literature, etc.
Participants & Algorithms (Selected)
Predictor Affiliation Software Algorithm
Abagyan Scripps ICM Force Field
Camacho/Vajda Boston CHARMM Force Field Refinement
Gardiner Sheffield GAPDOCK Shape + Area GA
Sternberg/Smith Imperial FTDOCK FFT
Bates/Fitzjohn ICRF Guided Docking Force Field
Ten Eyck/Mitchell SDSC DOT FFT
Vakser/Tovchigrechko SUNY/MUSC GRAMM FFT
Olson Scripps Harmony Spherical Harmonics ?
Weng/Chen Boston ZDOCK FFT
Eisenstein Weizmann MolFit FFT
Wolfson/Nussinov ? BUDDA/PPD Geometric Hashing
Iwadate Kitasato TSCF Force Field + Solvent
Ritchie/Mustard Aberdeen Hex Spherical Polar Fourier
Palma Lisbon BIGGER Geometric + Electrostatic
Gray/Baker Washington ? Monte Carlo + FlexibilitySeveral other participants (not shown) attempted a small no. of targets
CAPRI Results (Unofficial)
Predictor Software Algorithm T1 T2 T3 T4 T5 T6 T7
Abagyan ICM FF + + +
Camacho CHARMM FF + +
Gardiner GAPDOCK GA +
Sternberg FTDOCK FFT +
Bates Guided Docking FF +
Ten Eyck DOT FFT +
Vakser GRAMM FFT
Olson Harmony ?
Weng ZDOCK FFT + + +
Eisenstein MolFit FFT + + +
Wolfson BUDDA/PPD GH +
Iwadate TSCF FF
Ritchie Hex SPF + +
Palma BIGGER GF +
Gray ? MC + +
NB. Some predictions were disallowed: “slamming” or > limit of 5 entries
Hex 3.1 Results for CAPRI Targets 1–7
Target Description Constraints Samples Rank Contacts RMS Time/h
1 HPrK/HPr β1 ≤ 45◦ 1.5 × 108 −/1 2/56 − 0.2
2 VP6/MCV β1, β2 ≤ 45◦ 5.6 × 108 −/5 0/52 − 5.3
3 HA/HC63 β1, β2 ≤ 45◦ 6.3 × 108 4/5 43/63 7.43 6.5
4 Amylase/AMD10 β1, β2 ≤ 45◦ 5.6 × 108 −/5 0/58 − 4.7
5 Amylase/AMB7 β1, β2 ≤ 45◦ 5.6 × 108 −/5 0/64 − 4.7
6 Amylase/AMD9 β1, β2 ≤ 45◦ 5.6 × 108 5/5 53/65 2.16 4.7
7 SpeA/14.3.D β1 ≤ 45◦ 1.5 × 108 −/5 2/37 − 0.25
Submitted Orientation for Target 1
Docked Orientation for Target 3
Docked Orientation for Target 6
Target 7 – What Went Wrong?
CAPRI Targets 1–7 Revisited
Shape Only +EL +MM +EL+MMTarget R/L RMS N Rank RMS Rank RMS Rank RMS Rank RMS
1 2.39/0.57 30 1 4.06 11 4.06 33 4.18 19 4.18
2 0.62/0 30 37 3.05 41 3.05 123 2.90 119 2.90
3-AC 1.13/0 30 112 9.99 18 2.75 3 6.95 3 6.95
3-CE 1.13/0 30 62 8.91 39 8.91 18 8.43 15 8.43
4 0.41/0 30 63 3.05 10 3.05 125 8.74 106 8.74
5 0.41/0 30 826 2.74 118 2.74 18 2.49 20 2.49
6 0.41/0 30 1 0.75 1 0.75 1 0.64 1 0.64
7 1.09/0.49 30 122 5.37 295 9.22 39 5.60 18 8.83
Average Rank: 175 76 51 43
Log-Average: 55 38 29 23
Conclusions & Future Directions
Hex
• Fast and reasonably accurate: 2 out of 7 ain’t bad !
• SH surface sampling works well for large molecules
• Electrostatics & Soft OPLS minimisation (usually) help
• Clustering certainly helps
• Need better model of flexibility than “soft docking”...
• Eigenvector analysis of Essential Dynamics ??
CAPRI
• Protein Docking Algorithms are Improving
• Special Issue of PSFG, 1st Quarter 2003
• Annual Meetings: Alternating Theoretical/Assessment
• Ongoing Call to Crystallographers for Targets...
Acknowledgements
DWR: BBSRC 1996 – 2000
Thanks To:
Russell HamiltonGuillaume Valadon
CAPRI Organisers & The EBI
Hex Software & Preprints
http://www.biochem.abdn.ac.uk/hex/
CAPRI Call For Targets
http://capri.ebi.ac.uk/call for targets.html