Transcript Development of a Ligand Knowledge Base

Development of a
Ligand Knowledge Base
Natalie Fey
Crystal Grid Workshop
Southampton, 17th September 2004
Overview


Ligand Knowledge Base
Synergy of Database Mining and
Computational Chemistry:


Part 1: How computational chemistry can add
value to database mining results.
Part 2: How database mining can inform a ligand
knowledge base of calculated descriptors.
Ligand Knowledge Base

Aims:

Collect information about ligands and their (TM)
complexes:




Database mining.
Computational chemistry
Exploit networked computing and data storage
resources – e-Science.
Use data:


Interpretation of observations.
Predictions for new ligands.
Ligand Knowledge Base
Mine Structural
Databases
(e.g. CSD)
Compile systematic
structural information
about TM complexes
Computational Chemistry
(e.g. DFT)
Calculate structural
Ligand
Knowledge and electronic parameters
Base
for known and unknown
TM complexes
Part 1: “Unusual” Geometries
Query CSD for
structural pattern
Automatic
statistical analysis
of results
Main Geometry / Trends
apply outlier
criteria
Outliers
DFT geometry
optimisation
Optimised Geometries
compare with
crystal structures
Crystal Structure
and DFT agree
Crystal Structure
and DFT disagree
Part 1: “Unusual” Geometries
Value Added
Crystal Structure
and DFT agree
Why outlier?
Structure Report
Comment about
structure?
Yes
Note in database,
may confirm by DFT
No
Flag for detailed
investigation
Further
calculations
Additional results,
add to database
Part 1: “Unusual” Geometries
Crystal Structure
and DFT disagree
Value Added
Why?
Structure Report
Comment about
structure?
Revised
Calculations
Crystal Structure
and DFT agree
Note in
database
Problem with
Calculation
Yes
No
Problem with
Structure
Crystal Structure
and DFT disagree
Additional results,
add to database
Flag for detailed
investigation
Example – 4-coordinate Ruthenium



Main geometry: tetrahedral (14 structures)
2 square-planar cases: YIMLEL, QOZMEX
YIMLEL: cis-[RuCl2(2,6-(CH3)2C6H3NC)2]
N
N
Ru
Cl
Cl
4-coordinate Ruthenium

DFT result:



Use as CSD query, any TM…
SIVGAV – Pd
Supported by structural
arguments:



short Ru(II)-Cl, Ru-CNR.
correct range and geometry for Pd.
Run DFT with Pd:
Part 2: P-donor LKB

Range of DFT-calculated descriptors for
monodentate P(III) ligands and TM complexes.


Capture steric and /-electronic properties.
Identification of suitable statistical analysis
approaches:


Interpretation.
Prediction.
Part 2: P-donor LKB

Role of database mining:

Stage 1: Database generation.



Inform input geometries (conformational freedom).
Verification of chosen theoretical approach.
Stage 2: Database utilisation.


Supply experimental data for regression models.
Confirmation of calculated trends.
Examples
Stage 1

Conformers:
e.g. P(o-tolyl)3

Method verification:
tBu3
1.96
av. P-R, calculated

1.94
1.92
iPr3
Cy3
1.90
Et3 Pr3
Bu3
1.88 Me3
1.86
1.83
1.85
1.87
av. P-R, CSD
P
P
1.89
Examples

Stage 2:
Solid State Rh-P Distance (Rh(I), CN=4)
.003
Residual
predicted
2.425
2.375
2.325
0.000
-.003
2.28
2.32
2.36
Predicted Value
2.275
2.275
2.325
2.375
experimental
2.425
2.40
2.44
Conclusions




Synergy of approaches allows to add value to
structural databases.
Computational chemistry can be used to verify solid
state geometries.
Can exploit e-Science resources to add value on a
large scale.
Utility of large databases for structural chemistry of
transition metal complexes.


Computational requirements.
Statistical analysis.
Acknowledgements




Guy Orpen, Jeremy Harvey
Athanassios Tsipis, Stephanie Harris
Ralph Mansson (Southampton)
Funding: