ExtendNMR Consortium report meeting
8-9/5 2007, Cambridge.

Present:
Ernest Laue, Wolfgang Rieping, Rasmus Fogh, Tim Stevens, Wayne Boucher, Kim
Henrick, Wim Vranken, Michael Nilges, Benjamin Bardieux, Alexandre Bonvin,
Martin Billeter, Peter-Rene Steiner, Hans-Robert Kalbitzer, Wolfram Gronwald,
Wilhelm Malloni, Silvia De Sanctis, Gert Vriend (8/5 only).

Workpackages are presented in the order they were discussed.



Workpackage 2:

Deliverables:


D4 'Improved structure calculation tools more suitable for large complexes and
proteins'. 30 Oct. 2007

Main responsibles are Paris and Utrecht groups. The following would count
towards the deliverable

- New methods in ARIA for symmetric multimers, where a single copy of the
molecule is used together with crystallographic symmetry operators. This could
also be used for solid state, as the symmetry operations include translation
symmetry.

- Some probabilistic aspects about to be added to ARIA (log-normal potentials).
This should be part of ARIA 2.3 ( expected release late 2007). This will also
contribute to validation (e.g. identification of problematic regions).

- The ARIA/SOLARIA/CCPN collaboration on labeling schemes is intended
for solid state, but the work on labeling schemes would be equally useful for
large proteins.

- There are various improvements in HADDOCK, which is of course a program
explicitly for determing structures of complexes. These should be coming out in
Haddock 2.0 this year.:

   - Solvated docking, which determines water in interfaces.
   - Improved docking to DNA, with generation of appropriate DNA conformations
   - Use of diffusion anisotropy.
   - Up to six molecules supprted.
   - Check for 180 deg rotated binding.
   - Interfacing of HADDOCK with CCPN data model and CcpNmr, including a
   HADDOCK-specific package and a package for molecular dynamics parameters.

- The concept of 'proxy assignments' developed in the Utrecht group in the
previous grant period would fit in here. These should be included in the CcpNmr
Analysis program (ACTION TJStevens) before the end of the grant period.

- The development of ISD (Rieping, now U Cambridge) woudl also fit under
'improved tools', especially with respect for generating structures from sparse
data.


D6 'Structure calculation tools that do not require prior spectral assignment'
30 April 2008

The main results would be expected from the U Cambridge group. The NEXUS/CLOUDS
protocol is well under way, but a working version must be ready by October.
Currently only really good clouds are amenable to this, but Tim Stevens will
check whether new dismbiguation software has improved the position.

The Regensburg group has worked on automatic structure calculation based on
limited shift assignment. Just 35% assignment (e.g. backbone) can be leveraged
to 100%.


D7 ' Tools to assign NMR data based on existing structural information.
30 April 2008

Work of this nature is being planned under CLOUDS/NEXUS at U Cambridge.

The Regensburg group has results on using X-ray data to improve NMR solution
structures.

The Billeter group is developing a program to assign NMR spectra on the basis of
a known structure and 15N labeled spectra HSQC, HNHA, HNHB, HSQC-NOESY).


D8 'Validation tools for the assignment and structure of complexes.'
31 Aug 2008

The main results to be expected here are in the area of integration of existing
tools. Geerten Vuister in Nijmegen is working on a new program, CING, to combine
WHATIF, WHATCHECK, chemical shift and violation analysis. Work proceeds slowly
because a suitable person could not be found to hire. It was proposed that Alan
de Souza might be employed to do this, and EDLaue will contact Geerten Vuister on
this. A start has already been made to link CING to the CCPN framework (Wim
Vranken).

To a large extent validation of complexes is no different from validation of
multi-domain proteins. One useful addition would be special validation reports
for inter-monomer interface. Gert Vriend will add this.

The work of the Regensburg group on R-factor determination, though not specific
to complexes, would be a natural addition to this deliverable.

The use of log-normal potentials and interfacing of WHATIF in ARIA will also
count towards this deliverable.


Discussion: data deposition standards. All agree that more detailed deposition
of data will improve validation and allow future recalculation. The PDB stands
by its 'accept-anything' policy, but will tighten rules as much as possible in
view of recent scandals. A wwPDB meeting in September would be a good forum for
new NMR standards, and ExtendNmr might form a good nucleus for creating these.
The general procedure should be to gather agreement from famous scientists in
anticipation of future detailed rules. The general feeling is to deposit as much
information as is practical. A first proposal for standard data deposition rules
should be sent to leading spectroscopists (ACTION ???).

Deliverable D4 seems solidly on schedule.
Deliverables 6-8 have a reasonable amount of software, but all are less
advanced. Especially D8 needs work on interfacing to the CCPN data model, and
additional staff might help on this task.


Work Package 5.

D1 'Initial workshop' has been held.

D2 'Collection and analysis of existing NMR data'
30 Oct. 2006

Data set collection (DEN, raw data) is currently behind schedule, with few sets
collected (currently 3 datasets for complexes from Utrecht thanks to Alexandre
Bonvin). Alexandre Bonvin promised three new data sets, and more data sets will
be provided by U Cambridge, which should take care of requirements.

Extension of RECOORD (recalculated structures). Should be done using CCPNGrid -
testing framework needs to be set up to fit.

Analysis tools are generally being worked on. A CING interface is in progress,
HTML output for violation analysis has been written, and work is in progress on
generating topology files automatically for ligands.

The Presentation of Peter-Rene Steiner (Bruker) was given under WP5. The
contents fit under various headings and willbe mentined there.

D9 'Workshop to present tools'
August 2008.

Points to watch:
Are enough data being gathered?
Will test framework be ready when tools come in?
Will tools and integration fit into high-throughput testing?

Discussion:
It was agreed that the CCPN data model should keep a record of which assignment
program made a given assignment (but not a history of previous assignments). A
link to a Method record was recommended for ease of future expansion.

We should aim to have a preliminary GUI to run tests and visualise results by
November. It is still agreed that we should try out the methods on some
particular sample. There are several ways we could find one - Christian Edlich
(U Cambridge) might be one source., Muriel Delepierre another. The sample
protein should form a complex or (probably better) bind a ligand. We should set
the tests up on CCPNGRID.

The special problem of complexes, transient species will need some attention, to
fit with the grant application. Analysis already has facilities for tracking
moving peaks. We can store whatever BMRB has slots for. Special visualisation
facilities might be needed. Pymol was mentioned as a possible program to use,
but CcpNmr Analysis will go through with its integration with YASARA - YASARA is
commercial but the conditions seem satisfactory for this particular
visualisation use. It might make better sense to use existing structureal
genomics software to test what binds and try to take it from there.



Work Package 1

Deliverable D3  'New tools to process NMR spectra and identify and quantify
signals. 30 Oct. 2007

The Orekhov group is developing multidimensional decomposition (MDD) to handle
non-linear sparse data collection and data compression. The Billeter group is
developing PRODECOMP, a decomposition method built on projection NMR. Bruker is
collaborating with Orekhov, enabling sparse data collection and agreeing on
storage formats, with the Billeter group to enable projection data acquisition,
and with both on integration with Brukers TOPSPIN software. The Billeter group
is also working on autimatic assignment tools based on PRODECOMP data, in part
using GARANT, in part with new routines. The Laue group would be interested in
integrating PRODECOMP output with its own automatic assignment routines. The
Laue group has produced a Bayesian peak picker currently awaiting integration
with CcpNmr Analysis. The same picker could also be interfaced with PRODECOMP
and/or MDD to help decide on decomposition intervals on the acquisition axis and
to peakpick the final shapes. Work on quantitative maximum entropy routines is
in progress in the Laue lab.

Integration is proceeding (WP3) and work package 1 seems on track




Work package 4

Kalbitzer, Nilges and Billeter groups have almost identical resources on this
work package.

Deliverable D5 - 'New algorithms for random sampling of conormational space and
nD spectral simulations'

Although this was not discussed at the meeting, the new algorithms for random
sampling of conformational space would correspond to parts of the work discussed
under WP2 (structure calculation tools), especially ISD and the improvements in
ARIA. The work of the Billeter and Orekhov groups does provide new algorithms,
but they do not exactly fit the wording of the deliverable.

The Kalbitzer groupo is mainly working on extending their program AUREMOL, which
is tightly integrated with Bruker. A main feature of the program is the matching
of back-calculated and experimental spectra with the calculation of an R-factor
from the match. AUREMOL contains a number of algorithms that fit well with this
and other work packages, such as back-calculation of 2D and 3D NOESY
experiments, Bayesian signal recognition, R factor validation. The KNOWNOE
program, now with network anchoring included, works by simulating a NOESY
experiment and iteratively improving the fit to the experimental spectrum. It
can provide a 30% assignment from an unassigned start, and close to 100%
assignment starting from a 30% asignment (e.g. backbone). Integration with
Brukers TOPSPIN is well advanced and has received much attention. Integration
with CCPN is still to do.

One objective of this work package, partly reflected in the wording of the
deliverable, was to provide a series of generic algorithms that could be used
aross the entire collaboration. So far we haev a central repository, but nothing
has been deposited in it - this clearly must change soon. The kinds of
algorithms that might usefully be deposited include multiway decomposition
(MDD/PRODCOMP), Bayesian peak pickers (Laue/Kalbitzer), Prallelised Markov Chain
Monte Carlo (ARIA/ISD), assignment threader (CLOUDS/NEXUS), a PCA module for NMR
structures (COCO/Vranken) and a number of algorithms from the Kalbitzer group.

There is much potential in the work package, but not much is available yet.
Points to watch:
- We must make those algorithms available in the repository.
- Can we claim to have developed faster or more efficient algorithms?
- It is not clear how much of the (impressive) work from Regensburg, such as the
nD spectrum simulations, is available for deposition/integration or can be
claimed in the deliverable.




Workpackage 3 'Integration of tools developed in WP1 and WP2 for studies of
biomolecular complexes and their interactions'

Deliverables:

D10 'Integrative software package'.

NOTE - the grant proposal has an error in it. There are ten deliverables in the
deliverables list, but the Gantt chart contains eleven including 'D5 - Analysis
of available structures and NMR data of protein complexes to identify the most
appropriate NMR parameters for determining interaction sites'. The Gantt chart
is incorrect, and we shall work from the deliverables list.

- The software from WP1 (mainly MDD and PRODECOMP) is being integrated. The
necessary data model changes have been done, the process is planned and is
proceeding on schedule.
- The software from WP2 (ARIA, HADDOCK, CLOUDS/NEXUS, ISD) is well integrated
already and work is proceeding.
- Integration with Bruker software is progressing.
- An integrated package will be based on the CcpNmr Analysis interface. Work is
in progress to set up an umbrella interface that can call any of several
applications (including CcpNmr Analysis).

On 'analysing of available structures to determine NMR parameters for
determining interaction sites' there is little progress to report. Data are
being provided to ARIA and RECOORD and DOCR/FRED projects.

Points to watch:
Joint software repository is created, but not yet used
Integration of validation software
Integration of Regensburg software
Common GUI to launch and visualize programs
'Comparison of data from different approaches'
'Facilitate studies of excited states, dynamics, and weak interactions'
'Examine binding of compounds from library as proof of concept'
 
 
 
 
 
 General agreements:
 
-  Each group should contribute to each workpackage coordinator what they have
 done towards the WP goals. Reporting deadline is 45 days from 30 June (end of
 period). Deadline for scientific part of the report is 45 days from 9 May (end
 of meeting) - admin will handle the budget details after this.

 - We should request an extension of the grant. Each group should communicate how
 long an extension, for wahta and why to the coordinator. NB no follow-up grant
 can be applied for till the extension is over.
 
- We need deliverable reports for finished deliverables.

- Work force statistics (Appendix 4) on the mid-term report should be filled in
by each group. U. Cambridge will handle the rest. U. Cambridge should send
directions to others about how to reach appropriate web site.

- We should use the ExtendNmr mailing list for communication.

- We all need to put a page on the web site showing what we are doing, linking
to relevant pages, etc. Something is needed both per participant and per
workpackage. Wim Vranken will put up a template that people can edit
and send out a mail (now done).

- We must add code for general algorithms to the code repository. Anything that
is reasonably generic and could in theory be used by other participants would
be fine. People should consider possible algorithms while they prepare the web
site.

- We need to organise the reporting meeting for the fall. There will be a 20-30
minute slot for each workpackage. Every group should prepare slides for their
work and pass them on to the relevant workpackage coordinators for incorporation
in a big powerpoint. Two competing dates were proposed. 12-13/11 Peter Rene
Steiner and all Bruker people are busy, or 19/11 when Prof Kalbitzer will no
longer be in Europe. Jane Stiles to send out a question schema.