Using the CCHMC BMI cluster for NMR structure and dynamics calculations

1
Using the CCHMC BMI cluster forNMR structure and
dynamics calculations

• Douglas Kojetin, Ph.D.
• UC College of Medicine

2
Biomacromolecular Structure

• Two principal methods (high-resolution)
• X-ray crystallography
• Solid-state
• NMR spectroscopy
• Solid-state
• Solution-state
• Different from MRI both use magnets
• MRI bulk density
• NMR atomic-level applications
• Others (Small Angle X-ray Scattering solution)

3
NMR Studies of

• 3D structure determination (structure ensemble
  vs. static crystal representative)
• NMR is limited (size, solubility)
• Dynamics (time-scale movements)
• Certain time-scale motions are important for
  protein function (interactions, catalysis, etc.)
• Sample quality (folded? domains?)
• Titrations/conditions (ligand, pH, temp)

4
Protein NMR Spectroscopy
ubiquitin
5
Protein NMR Spectroscopy
2D 1H-15N HSQC
Each point is sensitiveto chemical
environment Ligands, temp, pH, etc.
6
Data Analysis Step 1

• Unambiguously assign all atoms in the protein
• Backbone (HN, N, Ca, Cb, C)
• Side-chain

HNCACB, (2) HNCO/HN(CO)CA, (3) HNCA/HN(CO)CA
7
Structure determination

• Start from an extended polypeptide chain
• Constraints for structure calculation
• through-space 1H-1H connectivities (NOEs)
• Dihedral angles (predicted chemical shifts vs.
  random coil database)
• Hydrogen bonds
• Residual dipolar couplings (orientation)

8
Structure determination

• Automated software iteratively assign and remove
  bad assignments (CNS ARIA)
• Iteration 0-7 20 structures per iteration
• Last iteration 100-200 structures
• Entire process is iterated (nested) 1 day over
  20 desktop CPUs ethernet, NFS farm out calcs

9
Structure determination

• Solvent refinement
• Fine tune NOE automated assignments
• Better Ramachandran (stereochemical) quality
  fewer atom clashes
• AMBER MPI capable, but calculations take 3 hrs
  per structure 100 separate, independent
  calculations
• shell scripts to setup files, submit 100 jobs to
  queue write working files to /tmp/XXX on node
  copy important files to home dir delete /tmp/XXX
  when finished)
• Future MD calculations will take advantage of MPI

10
Dynamics calculations (time-scale motions)

• Measure 15N backbone relaxation rates (measures
  motions on ps-ns time-scale)
• Fit rates to motional models use statistical
  model selection to choose best fit
• Motional model type of tensor describing system
• Important motional parameters S2 (order
  parameter) and Rex (ms time-scale motions data
  cannot be described by ps-ns motions only)

11
Dynamics calculations (time-scale motions)

• Nested, single calculations

1
1. Initial calculation
2. Analyze data vs. 4 tensor types sphere,
prolate, oblate, ellipsoid
2
2
2
2
3. Final selection of tensor monte carlo
error estimation
3
4. Extract statistical information into csv
table
Repeated using different initial parameter
estimates describing system (NH bond length,
corr. Timr tensor params)
4
12
Dynamics calculations (time-scale motions)
1
PBS -S /bin/csh PBS -l walltime240000 PBS
-l nodes1ppn1opteron ...
2
2
2
2
PBS -S /bin/csh PBS -l walltime240000 PBS
-l nodes1ppn1opteron ... PBS -W
dependafterokCLUSTER_JOB
x4
3
4
13
Dynamics calculations (time-scale motions)
!/bin/sh 1 QUEUE INITIAL RUN local_tmqsub
JOB_DIR/local_tm local_tmidecho local_tm
gawk -F '.' 'print 1' sleep 2 2a SETUP
TENSOR RUNS sed "s/CLUSTER_JOB/local_tmid/g"
JOB_DIR/sphere \ gt JOB_DIR/sphere.mod ...
prolate, oblate, ellipsoid (4 total) 2b QURUE
TENSOR RUNS sphereqsub JOB_DIR/sphere.mod sphe
reidecho sphere awk -F '.' 'print 1 ...
prolate, oblate, ellipsoid (4 total) 3 QUEUE
FINAL RUN sed "s/FINAL_JOB/sphereidprolateido
blateidellipsoidid/g \ JOB_DIR/final gt
JOB_DIR/final.mod finalqsub JOB_DIR/final.mod
finalidecho final awk -F '.' 'print
1' 4 QUEUE FINAL STATS EXTRACTION sed
"s/STATS_JOB/finalid/g" JOB_DIR/stats \ gt
JOB_DIR/stats.mod statsqsub JOB_DIR/stats.mod
PBS -S /bin/csh PBS -l walltime240000 PBS
-l nodes1ppn1opteron ... PBS -W
dependafterokCLUSTER_JOB
1
2
2
2
2
3
4
14
Example Map S2 on structure
15
Thanks!

• CCHMC BMI
• Prakash
• Those who provide
• Good Samaritan ER