Nanoparticles: unusual QSAR for unusual structure

1
Nanoparticles unusual QSAR for unusual structure

• Novoselska Natalia
• Bakhtiyor Rasulev, Agnieszka Gajewicz, Tomasz
  Puzyn, Jerzy Leszczynski, Kuzmin Viktor

2
Recent Nano-QSAR studies

1. H. Tzoupis et. al, Binding of novel fullerene
   inhibitors to HIV-1 protease. J. Comput. Aided
   Mol. Des., 2011, 25, 959976
2. A. Toropova et. al. CORAL QSPR models for
   solubility of C60 and C70 fullerene
   derivatives. Molecular Diversity, 2011, 5,
   249-256
3. T. Puzyn, et. al. Using nano-QSAR to predict the
   cytotoxicity of metal oxide. Nature
   Nanotechnology, 2011, 6, 175-178
4. A. Toropov et. al, InChI-based optimal
   descriptors QSAR analysis of fullereneC60-based
   HIV-1 PR inhibitors by correlation balance. Eur.
   J. of Med. Chem., 2010, 45, 13871394
5. K. Muzino et. al, Antimicrobial Photodynamic
   Therapy with Functionalized FullerenesQuantitativ
   e Structure-activity Relationships. J Nanomedic
   Nanotechnol., 2011, 2, 175-17
6. ..
7. N.Novoselska et. al, 2D nanoQSAR models for
   predict the cytotoxicity of metal oxides
   nanoparticles. NanoScale, not yet issued

3
Is the SiRMS approach applicapable for
nanoparticles description?

• 2D-simplexes descriptors

Differentiation by type, charge, refraction,
donor/acceptor of hydrogen bond,
lipophilicity Lipophilicity was calculated by
additive scheme (XLogP) Renxiao Wang, Ying Fu,
Luhua Lai, J.Chem. Inf. Comput. Sci., 37 (1997)
Integral characteristics XLogP, Rf, AW, En
Kuzmin V.E. et al. Virtual screening and
molecular design based on hierarchical QSAR
technology. // Challenges and Advances in
Computational Chemistry and Physics, 2010, 8,
127-176
4
1. Analysis of efficiency SiRMS solubility of
C60 and C70 derivatives in chlorobenzene
P. Troshin et al. Material Solubility-Photovoltai
c Performance Relationship in the Design of Novel
Fullerene Derivatives for Bulk Heterojunction
Solar Cells Advanced Functional Materials,
2009 19, 5, 779788
5
1. Analysis of efficiency SiRMS solubility of
C60 and C70 derivatives in chlorobenzene
A. Toropov et. al CORAL QSPR models for
solubility of C60 and C70 fullerene
derivatives Molecular Diversity, 2011, 5, 249-256
Our results
R2 0.90 S 12.5 (mg/mL)
R2 (consensus) 0.98 S 2.5 (mg/mL)
6
2. Analysis of efficiency SiRMS
fullerene-based HIV-1 PR inhibitors
H. Tzoupis et. al, Binding of novel fullerene
inhibitors to HIV-1 protease J. Comput. Aided
Mol. Des., 2011, 25, 959976
CoMFA R2 0.98 Q2 0.61 S 0.154
CoMSIA R2 0.99 Q2 0.79 S 0.137
R2(consensus) 0.98 S 0.14
A. Toropov et. al, SMILES-Based Optimal
Descriptors QSAR Analysis of Fullerene-Based
HIV-1 PR Inhibitors by Means of Balance of
Correlations J. Comp. Chem, 2010, 31, 381392
A. Toropov et. al, InChI-based optimal
descriptors QSAR analysis of fullereneC60-based
HIV-1 PR inhibitors by correlation balance Eur.
J. of Med. Chem., 2010, 45, 13871394
R2 0.5-0.99 S 0.127-0.352
R2 0.76-0.97 S 0.271-0.681
7
Unusual QSAR Oh, really?
8
LDM Liquid Drop Model
In a liquid drop model, nanoparticle is
represented as the spherical drop, which
elementary particles are densely packed, and
density of cluster is equal to mass density. In
this model the minimum radius of interactions
between elementary particles in cluster is
described by Wigner-Seitz radius
Smirnov B M. Processes involving clusters and
small particles in a buffer gas. Phys. Usp.
2011, 54, 691721
9
3. Superconductivity critical temperatures of
inorganic nanoparticles
 
Compound Tc
ZnS 195
ZnSe 75
ZnTe 52
CdS 200
CdSe 80
CdTe 60
GaN 415
GaP 95
GaAs 130
GaSb 60
InN 315
InP 65
InAs 44
R2 (consensus) 0.83 S 0.3
Diagram of relative influence () on critical
temperatures
10
4. Comparative QSAR analysis of toxic effects of
metal oxide nanoparticles
Compound HaCaT cells, log(1/EC50) E. Coli, log(1/EC50) Size, nm Aggregation size, nm
Al2O3 2.49 1.85 44 372
Bi2O3 2.82 2.5 90 2029
CoO 3.51 2.83 100 257
Cr2O3 2.51 2.3 60 617
Fe2O3 2.29 2.05 32 298
In2O3 2.81 2.92 30 224
La2O3 2.87 2.87 46 673
NiO 3.45 2.49 30 291
Sb2O3 2.64 2.31 20 223
SiO2 2.2 2.12 150 640
SnO2 2.01 2.67 15 810
TiO2 1.74 1.76 46 265
V2O3 3.14 2.24 15 1307
WO3 - 2.56 50 180
Y2O3 2.87 2.21 38 1223
ZnO 3.45 3.32 71 189
ZrO2 2.15 2.02 47 661
Compound HaCaT cells, log(1/EC50) E. Coli, log(1/EC50)
Al2O3 2.49 1.85
Bi2O3 2.82 2.5
CoO 3.51 2.83
Cr2O3 2.51 2.3
Fe2O3 2.29 2.05
In2O3 2.81 2.92
La2O3 2.87 2.87
NiO 3.45 2.49
Sb2O3 2.64 2.31
SiO2 2.2 2.12
SnO2 2.01 2.67
TiO2 1.74 1.76
V2O3 3.14 2.24
WO3 - 2.56
Y2O3 2.87 2.21
ZnO 3.45 3.32
ZrO2 2.15 2.02
11
LDM Liquid Drop Model
12
Metal-ligand Binding Characteristics
(CI) - reflects the energy of the metal ion
during electrostatic interactions with a ligand
(CPP) - reflects the relative importance of
covalent interactions relative to ionic during
metal-ligand binding
M.C. Newman, et al . Using metalligand binding
characteristics to predict metal toxicity
quantitative ion characteractivity relationships
(QICARs). Environ. Health Persp., 1998, 106,
14191425
13
4. Comparative QSAR analysis of toxic effects of
metal oxide nanoparticles
  HaCaT cells (17 compounds) E.Coli (16 compounds)
R2 (training set) 0.96 0.93
S (training set) 0.10 0.13
R2 (test set) 0.92 0.78
S (test set) 0.12 0.32
14
4. Comparative QSAR analysis of toxic effects of
metal oxide nanoparticles
Diagram of relative influence () on toxicity to
HaCaT cells
Diagram of relative influence () on toxicity to
E.Coli
15
It was shown that SiRMS descriptors (in case of
fullerenes) and combination of LDM-based
descriptors with SiRMS (in case of inorganic
nanoparticles) can be helpful for QSAR
investigation of different properties of
nanomaterials.
16
Thank you for your attention!
AcknowledgementsA.V.Bogatski Physico-Chemical
Institute NAS of UkraineKuzmin
ViktorInterdisciplinary Center for
NanotoxicityBakhtiyor Rasulev, Jerzy
LeszczynskiUniversity of GdanskAgnieszka
Gajewicz, Tomasz Puzyn
17
LDM Liquid Drop Model
SiRMS LDM
Simple combination
Recalculation
18
Classification of nanoparticles
19
2. Analysis of efficiency SiRMS fullerene-based
HIV-1 PR inhibitors
20
Simplex Representation of Molecular Structure
Electrostatic Steric Informational
Charges Lipophilicity Polarizability Volume etc
Molecular Field
Physical-Chemical
Descriptoral
21
Random Forest method implemented in CF program
(http//qsar4u.com) was used for the development
of QSPR models at the 2D level of representation
of molecular structure.
Forest is a set of classification or regression
trees (T).
The major criterion for estimation of the
predictive ability of the RF models and model
selection is the value of R2OOB. Coefficient of
determination for OOB set
,
Determination coefficient for test set (R2test),
standard error (SE) and mean absolute error (MAE)
are also characteristics of the models. R2test
for test set is calculated similar to R2OOB.
22
Consensus