Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2018 Feb 23.
Published in final edited form as: Chem Eng Sci. 2016 Sep 30;159:131–139. doi: 10.1016/j.ces.2016.09.033

Computer-aided Molecular Design of Water Compatible Visible Light Photosensitizers for Dental Adhesive

Farhana Abedin 1,2, Brock Roughton 2, Qiang Ye 1, Paulette Spencer 1,3, Kyle Camarda 4
PMCID: PMC5699481  NIHMSID: NIHMS823519  PMID: 29176909

Abstract

Dental adhesive resin undergoes phase separation during its infiltration through the wet demineralized dentin and it has been observed previously that the hydrophilic-rich phase is a vulnerable region for failure due to the lack of photo-polymerization and crosslinking density. The lack of photo-polymerization is mostly due to the partitioning of photo-initiators in low concentrations within this phase. Here, a computational approach has been employed to design candidate water compatible visible light photosensitizers which could improve the photo-polymerization of the hydrophilic-rich phase. This study is an extension of our previous work. QSPRs were developed for properties related to the photo-polymerization reaction of the adhesive monomers and hydrophilicity of the photosensitizer using connectivity indices as descriptors. QSPRs and structural constraints were formulated into an optimization problem which was solved stochastically via Tabu Search. Four candidate photosensitizer molecules have been proposed here which have the iminium ion as a common feature.

Keywords: photosensitizer, dental adhesive, QSPRs, optimization, MINLP, connectivity indices, Tabu Search

Introduction

Due to the aesthetic appeal and concern for mercury release, dentists are becoming more inclined towards using dental composite restoration than the conventional amalgam despite the shorter clinical lifetime of the former (Demarco et al., 2012). This has drawn attention in the past to improve the mean clinical lifetime of the composite restoration. It was observed in the past that the weak link to failure of the composite restoration was the dentin/adhesive interface, the adhesive being applied prior to the composite restoration (Spencer et al., 2010). Previous investigations indicated that incomplete infiltration of the dental adhesive resin through the demineralized dentin, poor polymerization, phase separation, enzymatic and hydrolytic degradation of the adhesive, degradation of the exposed collagen were some of the reasons for failure of the restoration (Cadenaro et al., 2005; Hashimoto et al., 2011; Pashley et al., 2004; Santini and Miletic, 2008; Shokati et al., 2010; Spencer and Wang, 2002). The major components of dental adhesives are dimethacrylate cross-linker, monomethacrylate monomer usually 2-hydroxyethyl methacrylate (HEMA) to facilitate infiltration, and a photo-initiator system. When exposed to the dental light curing unit (LCU), the photo-initiator system would trigger photo-polymerization reaction of the dental resin monomers. As the adhesive resin diffuses through the wet demineralized dentin, it undergoes phase separation into hydrophobic- and hydrophilic-rich phases (Spencer and Wang, 2002). It was found in a previous study that the cross-linker and the photo-initiator system being hydrophobic, were present in low concentrations within the hydrophilic-rich phase (Ye et al., 2012). This meant that the hydrophilic-rich phase would become a vulnerable region for failure. This phase would undergo poor photo-polymerization and cross-linking density due to the lack of the photo-initiator system and cross-linker respectively. In this study, several candidate hydrophilic visible light photosensitizer molecules were designed specifically for dental adhesive application. Photosensitizer is a component of the photo-initiator system and when it is irradiated with the dental LCU, photosensitizer molecules are promoted to the excited state which in combination with the co-initiator form free radicals for triggering the polymerization reaction (Cook and Chen, 2011). The approach used here for the design was computer-aided molecular design (CAMD).

The overall goal of this study is to improve the lifetime of the dental adhesive layer and hence the composite restoration. A hydrophilic photosensitizer was designed because improved water compatibility will ensure partition of the photosensitizer in higher concentration within the hydrophilic-rich phase during phase separation, and hence this would enhance the photo-polymerization of this phase. The newly designed photosensitizer should have absorption wavelength in the visible range because the dental light curing units also possess emission wavelength in the visible range.

Computer-aided molecular design (CAMD) is an efficient and inexpensive technique capable of predicting molecular structures or mixture or system with desired optimal properties. The design approach consists of two phases which are forward and inverse problems. The forward problem involves selecting appropriate compounds or components in a mixture which form the model building set, and establishing correlations between the properties of interest and the model building set. This form the quantitative structure property relationships (QSPRs). The inverse problem involves formulating an optimization problem using structural constraints and the correlations (QSPRs), and solving it to predict molecular structures or components with properties approaching the desired target values. CAMD has been used in the development of pharmaceutical products/biologics, catalysts, polymers, solvents or their mixtures, refrigerants etc. (Gani et al., 1991). Group contribution in CAMD is a method in which the functional groups representing the compounds or mixture, and the number of times they occur are used to determine the properties (Gani et al., 1991). Merrero et al. used group contribution method to predict properties such as critical pressure, critical volume, and normal boiling/melting temperatures of organic compounds (Marrero and Gani, 2001). In the study conducted by Merrero et al., in addition to simple groups higher order groups were used to capture more information such as isomers and complex structures such as polyfunctional acyclic or heterocyclic compounds (Marrero and Gani, 2001). Folic et al. designed solvents for reactions by CAMD where the models for solvent properties were obtained using solvatochromic equations and group contribution method (Folić et al., 2008). Solvents for enhanced reaction kinetics were also designed in the past using CAMD approach where the objective function was to maximize the reaction rate constant at 298 K (Struebing et al., 2013).

Apart from group contribution, connectivity indices which are topological descriptors have been employed in the past to obtain QSPRs. Connectivity indices describe the molecular structure of an entity in terms of connectivity as well as electronic configurations (Bicerano, 2002). As a result connectivity indices are thought to contain more structural information than the group contribution method (Camarda and Maranas, 1999). Gani et al. used zero and first order connectivity indices to predict the contribution of missing groups and they described the approach as group contribution+ since group contribution was used in combination with the connectivity indices for property prediction (Gani et al., 2005). Connectivity indices have been successfully used in the past to design pharmaceuticals, environmentally benign catalyst, cross-linked polymer etc. (Chavali et al., 2004; Eslick et al., 2009; Roughton et al., 2012b; Siddhaye et al., 2000). Roughton et al. employed connectivity indices to estimate properties such as glass transition temperature, Gordon-Taylor constant for carbohydrate excipient (Roughton et al., 2012b).

The inverse problem of CAMD could be solved either deterministically or by a stochastic method. Roughton et al. designed ionic liquids by a deterministic method using CPLEX solver in GAMS optimization software package (Roughton et al., 2012a). Genetic algorithm and Tabu Search are stochastic methods which have been employed in the past to solve the inverse problem in CAMD. Roughton et al. have used Tabu search to design carbohydrate excipient to minimize protein aggregation (Roughton et al., 2012b). Camarda et al. and Eslick et al. have also used Tabu Search for polymer design (Camarda and Maranas, 1999; Eslick et al., 2009). Venkatasubramanium et al. showed a CAMD framework using genetic algorithm for polymer design (Venkatasubramanian et al., 1994).

Although in the past, quantitative structure property relationships (QSPRs) of dyes for properties specifically related to photovoltaic applications have been developed but in this study QSPRs using model building set of mostly dyes have been developed for dental application. Vankatraman et al. successfully developed quantitative structure property relationships (QSPRs) of dyes representing coumarin derivatives for properties important to photovoltaic application using frequency based eigenvalue descriptors (Venkatraman et al., 2014). A recent study by Venkatraman et al. exhibited the QSPRs for phenothiazine dyes for application in photovoltaic cell using the eigenvalue descriptors (Venkatraman and Alsberg, 2015).

In this study, a model building set consisting of mostly dyes was used to develop the QSPRs for properties such as octanol/water partition coefficient, molar extinction coefficient, photon absorption efficiency (PAE), degree of conversion and rate of polymerization of dental monomers within the hydrophilic-rich phase. Finally the QSPRs are used in the inverse problem to design water compatible visible light photosensitizers with properties that will result in improved durability of the hydrophilic-rich phase. This is an extension of our previous study where only octanol/water partition coefficient and molar extinction coefficient were used as properties to design the photosensitizer molecules for dental application (Abedin et al., 2015a). This study represents a more complete step towards designing optimum photosensitizer for improved performance of the hydrophilic-rich phase.

Materials and Methods

Forward Problem: Properties and Model Building Set

The model building set was chosen rationally so that all the compounds have photosensitizing capability, possess maximum absorption wavelength in the visible range and are hydrophilic in nature. The properties of interest for the design are octanol/water partition coefficient, molar extinction coefficient, photon absorption efficiency (PAE), degree of conversion and rate of polymerization of dental monomers within the hydrophilic-rich phase. Octanol/water partition coefficient was chosen as a measure of the hydrophilicity of the molecules and the logarithm of the octanol/water partition coefficient (log P) for the molecules in the model building set were obtained from the literature (Pellosi et al., 2012; Sheikh, 1976; Wagner et al., 1998; Wainwright et al., 1999). The molecules in the model building set for the property, log P were: Bromophenol blue, Eosin Y, Erythrosin B, Fluorescein, Methylene blue chloride, Rose bengal, Victoria blue B and Victoria blue R. Table 1 shows the log P values for the molecules in the model building set. These eight molecules formed the model building set to develop the QSPR for the log P.

Table 1. logP values used to develop the QSPR.

Name of the molecule log P Reference
Eosin Y 0.18 [26]
Fluorescein -0.32 [26]
Rose bengal 0.59 [26]
Erythrosin B 0.46 [26]
Methylene blue -0.96 [28]
Bromophenol blue -2.35 [27]
Victoria blue R 1.59 [29]
Victoria blue B 2.8 [29]
Open in a new tab

The molecules in the model building set for the rest of the properties were: New Fuchsin, Victoria blue B, Methylene blue chloride, Eosin Y disodium salt, Bromophenol blue sodium salt, Erythrosin B, Fluorescein sodium salt, [3-(3,4-dimethyl-9-oxo-9H-thioxanthen-2-yloxy)-2- hydroxypropyl]trimethylammonium chloride (QTX) and Rose bengal sodium salt. There were nine candidate molecules in this model building set and most of them were dyes. Figure 1 shows the molecular structure of the compounds in this model building set.

Figure 1. Molecules in the model building set which were used to the collect experimental data.

Figure 1

Open in a new tab

Molar extinction coefficient (ξ) of a molecule at a given wavelength is a measure of the probability of a molecule to absorb energy when it is irradiated at that wavelength and be promoted to the excited state. This phenomenon would play an important role in the generation of reactive species for photo-polymerization reaction. Therefore, molar extinction coefficient (ξ) of a photosensitizer is an important property to determine the performance of the photoinitiator system. The absorbance of the photosensitizer and hence the photoinitiating performance is also dependent on the overlap of the emission spectrum of the light curing unit (LCU) and the absorption spectrum of the photosensitizer. This overlap between the two spectra is the photon absorption efficiency (PAE). These properties were experimentally determined, and the details for the experimental procedure along with the data were discussed by Abedin et al. (Abedin, 2016). Aqueous solution of each compound in de-ionized water was prepared and the absorption intensity was measured at 480 nm using a UV-vis spectrophotometer, Cytation 3 multimode microplate reader (BioTek, Winooski, VT). The molar extinction coefficient (ξ) was determined using the following equation and it was determined at 480 nm which was close to the maximum emission wavelength of the halogen LCU (Eacute et al., 2008).

Absorbance=ξ×[CmolL]×Xcm (1)

where ξ is the molar extinction coefficient in L/(mol cm) and X is the path length and [C] is the molar concentration.

The UV-vis spectrum of the aqueous solution of each compound was determined experimentally using the UV-vis spectrophotometer, Cytation 3 multimode microplate reader (BioTek, Winooski, VT). The PAE was obtained by converting the energy densities of the LCU at each wavelength to the number of photons per square centimeter and seconds (nphλ) by using equation 2 and multiplying this quantity with the absorption of the photosensitizer at the corresponding wavelength (Abedin, 2016). The resultant quantity was then plotted against wavelength which was integrated to give PAE. The PAE was normalized for concentration and the relative values of the normalized PAE was used for determining the QSPR. PAE had been determined similarly in the past (Neumann et al., 2005; Stahl et al., 2000) and the emission spectrum of halogen Spectrum 800 LCU was used for the calculations here (Eacute et al., 2008).

nphλ=Wλhc (2)

where W is the energy output of the LCU, λ is the wavelength, h is the Planck's constant and c is the speed of light.

Although the molar extinction coefficient and relative normalized PAE, representing the photochemistry of the photosensitizer play an important role in the generation of the reactive species but the effectiveness of these species in triggering the polymerization reaction is also an important parameter to consider for the efficiency of the reaction. Degree of conversion (DC) could indicate the effectiveness of the reactive species in driving the reaction and hence they had been considered as target properties. The rate of polymerization can be linked to the polymer shrinkage which is detrimental for the seal at the adhesive/dentin interface. The DC and rate of polymerization of only the hydrophilic-rich phase have been considered here. The procedure to obtain these data have been discussed by Abedin et al. (Abedin, 2016). Briefly, the photoinitiator system (0.5 wt% Camphorquinone, 0.25 wt% Photosensitizer in the model building set, 0.25 wt% Ethyl 4-(dimethylamino)benzoate, 0.5 wt% 2-(dimetylamino)ethyl methacrylate, 0.5 wt% Diphenyliodonium hexafluorophosphate) within the dental adhesive resin (45 wt% HEMA and 55 wt% BisGMA) was partially replaced with each photosensitizer molecule in the model building set, and then 33 wt% deuterium oxide was added to separate the adhesive into hydrophobic and hydrophilic-rich phases. Ye et al. and Abedin et al. have discussed this phase separation process (Abedin, 2016; Abedin et al., 2016; Ye et al., 2012). The polymerization kinetic study was performed in ATR sampling mode using the time resolved spectrum collector with a resolution of 4 cm-1. A PerkinElmer Spectrometer Frontier (PerkinElmer, Waltham, MA, USA) was employed to collect the data (Abedin, 2016; Abedin et al., 2015b).

Molecular Descriptors

In this study, connectivity indices have been used to predict the properties. Connectivity indices are topological descriptors (Randic, 1975) and to calculate the connectivity indices of various order it is first necessary to understand the simple connectivity index, δ and the valence connectivity index, δv. For a hydrogen suppressed graph of a molecule, δ for a non-hydrogen atom is the number of edges emanating from the atom that is the number of non-hydrogen atoms to which it is connected (Bicerano, 2002). δv of a non-hydrogen atom can be obtained from equation 3 and it stores information regarding the electronic configuration of the atom (Bicerano, 2002).

δv=Zv−NHZ−Zv−1 (3)

where Zv is the number of valence electrons of a non-hydrogen atom, Z is the sum of Zv and the number of inner shell electrons and NH is the number of hydrogen atoms bonded to the non-hydrogen atom of concern.

Zeroth order connectivity index of a molecule is the summation of δ over the all non-hydrogen atoms (vertices) which form the nodes of the hydrogen suppressed graph (Bicerano, 2002). Zeroth order valence connectivity index of a molecule is obtained similarly using δv. The zeroth order connectivity indices stores atomic information of the molecule. The first order connectivity indices were obtained by summing the product of simple connectivity indices of two connected atoms that is the summation is made over the bonds within the molecule (Bicerano, 2002). Higher order connectivity indices such as 5th order can be obtained by summing over a path length of 5. The extent to which structural and electronic information of a molecule is stored increases as the order of the connectivity becomes higher. In this study, connectivity indices of up to 5th order were used and for compounds which were salts in the model building set, the connectivity indices of the organic anion or cation had been calculated. Generalized equations that were used for the calculation of connectivity indices of molecules in the model building set and also during building novel molecules in the inverse problem had been discussed by Roughton et al. (Roughton et al., 2012b). Table 2 shows the connectivity indices for one of the dye molecule, Erythrosin B in the model building set.

Table 2. Connectivity indices for Erythrosin B.

Connectivity indices Valence connectivity indices
0χ: 20.6624 0χv: 22.7251
1χ: 13.7686 1χv: 12.5208
2χ: 13.4874 2χv: 10.9956
3χ: 12.9308 3χv: 8.6793
4χ: 10.9332 4χv: 7.3461
5χ: 9.8168 5χv: 4.6871
Open in a new tab

Quantitative Structure Property Relationships (QSPRs)

The mathematical correlations between molecular structures and target properties which represent QSPRs were obtained by linear regression using the statistical software package R (R-Development-Core-Team, 2010). The linear regression was carried out by the ordinary least squares (OLS) method. Based on the coefficient of correlation (R2), the number of descriptors were reduced from 10 to 5 in case of the rate of reaction, 6 for log P, and 7 for ln(ξ), normalized PAE, DC. Leaps package was used to select the descriptors (Lumely, 2009). The reduced number of descriptors were then used to develop linear models. The model with high R2 and lowest Mallow's Cp was selected. Mallow's Cp statistics is a measure of precision and bias in the model (Roughton et al., 2012b). As descriptors are added to the model, the Cp statistic is penalized and it increases after a certain number of descriptors (Roughton et al., 2012b). Figure 2 shows that the model with 6 descriptors for the relative normalized PAE exhibited minimum Cp, and hence this model was selected. A small Cp statistic means that the regression coefficients of the model are true. The predictive capability of the correlations were evaluated by the cross-validation method “Leave-one-out” (LOOCV). The DAAG package in R was used for carrying out the cross-validation. In LOOCV, each molecule in the model building set was left out and the correlation was carried out again using the selected descriptors. This correlation was then used to predict the property for the molecule left out. The predictive squared correlation, Q2 was then determined which gave a measure of the predictive capability of the model. Q2 close to R2 indicates a good predictive capability of the model.

Figure 2.

Figure 2

Open in a new tab

The model with 6 descriptors had the lowest Cp value. Addition of descriptor in excess of 6 led to an increase in the Cp value in case of the property, relative normalized PAE.

Inverse Problem: CAMD Formulation and Solution Technique

The molecules in the model building set were broken down into fragments which form the sub-groups for building new molecules. Figure 3 shows some example sub groups used in designing the new candidate molecules. A total of 42 sub-groups were created and 6 sub-groups at maximum were allowed to build the molecule. This meant that there were approximately 5489 million possibilities since the order in which the sub-groups were connected was important and repeated sub-groups were allowed. The inverse problem was formulated as a mixed integer non-linear problem (MINLP). The non-linearity arises due to the presence of higher order connectivity indices. The MINLP was solved via an optimization technique and the objective was to minimize the difference between the property value and the corresponding set target value. The problem formulation is given below (Abedin et al., 2015a; Roughton et al., 2012b):

Figure 3.

Figure 3

Open in a new tab

Example of some molecular sub-groups used in building novel photosensitizer molecule.

Minimizes=∑M1Pmscale|Pm−PmTarget|Pm=fm(χ)χ=g(ai,j,wi)hc(ai,j,wi)≥0 (4)

where Pm is the property m obtained from the QSPR designated as fm(X), X are the connectivity indices which are obtained from the function g, ai,j and wi are adjacency matrix and identity vector respectively, hc represents structural molecular constraints, PmTarget is the target value for the property m, Pmscale is the scaling factor.

In this study, the CAMD framework used previously by Roughton et al. to design carbohydrate excipient for preventing protein aggregation was employed (Roughton et al., 2012b). The difference between the property value and target value for each property was normalized. Here, the scaling factor for DC was 0.4 and for log P it was 0.3. The rest of the properties were given equal scaling factor of 0.1.

The optimization problem was then solved stochastically by “Tabu Search” which yields locally near optimal solutions. Tabu Search stores a set of previous solutions in the Tabu list to which it can compare the current solution, and if it is similar to those in the list, the solution is rejected. This allows new search space to be explored and prevents the solution from oscillating around a local optimum.

Results and Discussions

Quantitative Structure Property Relationships (QSPRs)

QSPRs link the molecular structures in the training set (model building set) with the target properties. The correlations with the combination of descriptors (connectivity indices) yielding lowest Cp and relatively high R2 were chosen to form the QSPRs for the properties. Cross-validation was also carried out to ensure that the models were able to accurately predict the properties of the newly built molecules. In case of the octanol/water partition coefficient, correlation between log P and descriptors was obtained and for the molar extinction coefficient (ξ), the correlation was between ln(ξ) and the descriptors. The QSPRs for these properties had been discussed in a previous study and hence would not be discussed here (Abedin et al., 2015a). The QSPRs for the other three properties which were PAE, degree of conversion and rate of polymerization of the hydrophilic-rich phase are discussed below. In case of all the properties R2 is close to Q2, indicating good predictive capability of the QSPRs. Table 3 summarizes the characteristics of the QSPR for each property.

Table 3. Summary of the models selected to represent QSPR for each property.

Property No. of descriptors used to develop QSPR Adjusted R2 Mallow's Cp statistic Q2
ln(ξ) 6 0.99999 6.0 0.99272
log P 5 0.99996 5.1 0.99813
Relative Normalized PAE 6 0.9996 6.0 0.9861
Degree of Conversion (DC) 6 0.9896 7.3 0.8633
Rate of Polymerization (RT) 5 0.9742 6.0 0.9169
Open in a new tab

Photon Absorption Efficiency (PAE)

As mentioned earlier PAE is a measure of the performance for photosensitizer/LCU combination. Maximizing PAE would enhance the absorption of the photosensitizer and hence could in turn improve the photo-polymerization reaction. Figure 4 is the parity plot for the relative normalized PAE. The QSPR for relative normalized PAE is given below:

Figure 4.

Figure 4

Open in a new tab

Parity plot showing that the predicted relative normalized PAE by QSPR is close to the experimental values.

Relative Normalized PAE=34.3226−(1.7485×χ0)−(8.0407×χ2)+(6.1803×χ3)+(5.108×χ5)+(9.9007×χv2)−(26.9838×χv5) (5)

Adjusted R2=0.9996, Q2=0.9861

Degree of Conversion (DC) for the Hydrophilic-rich Phase

The degree of conversion (DC) is a means to understand the effectiveness of the reactive species in driving the photo-polymerization reaction. For the design purpose, it is desired that the degree of conversion is high to enhance the durability of the dentin/adhesive bond within the hydrophilic-rich region. Figure 5 shows that the predicted and experimental DC are close. The QSPR for DC is given below:

Figure 5.

Figure 5

Open in a new tab

Parity plot for the property, degree of conversion (DC). The degree of conversion for the hydrophilic-rich phase was measured experimentally.

DC=−5.1037−(1.3282×χ1)+(2.1164×χ2)−(3.0557×χ4)+(2.9822×χ5)+(0.7338×χv2)−(1.1734×χv4) (6)

Adjusted R2=0.9896, Q2=0.8633

Rate of Polymerization for the Hydrophilic-rich Phase

High rate of polymerization indicates that the reactive species are effective in the polymerization reaction but for dental application very high rate will lead to high shrinkage which will be detrimental for the dentin/adhesive bond (Braga and Ferracane, 2002; Lim et al., 2002). On the contrary, if the rate is too slow it will take a long time for the reaction to complete. Figure 6 exhibits the parity plot for the rate of polymerization. The QSPR for the rate of polymerization (RT) is given below:

Figure 6.

Figure 6

Open in a new tab

Parity plot showing comparison between predicted rate of polymerization of the hydrophilic-rich phase to the experimental values.

RT=−0.0393+(0.0176×χ0)−(0.0559×χv1)+(0.0383×χv2)−(0.0286×χv3)+(0.0299×χv4) (7)

Adjusted R2=0.9742, Q2=0.9169

Complex properties such as the degree of conversion and rate of polymerization are difficult to predict, and factors such as steric hindrance and impact of experimental conditions/errors are not taken into account by the descriptors used here. Hence, the R2 and Q2 for these two properties were found to be inferior compared to the other three properties.

CAMD of Novel Hydrophilic Visible Light Photosensitizers for Dental Adhesive

The novel molecule was built by connecting the sub-groups with each other. A maximum of 6 subgroups could be connected to form a molecule. Solutions were then generated by making local moves such as swapping, inserting, deleting or replacing the sub-groups. The end groups that would be connected to the terminal connectors of the first and last sub-groups of the molecule were hydroxyl groups.

The target value for log P was -0.55 (Abedin et al., 2015a) and the basis for this was that this represents the log P value for HEMA which was found to partition at relatively higher concentration in the hydrophilic-rich phase compared to the other components (Fujisawa and Masuhara, 1981). The target values for the relative normalized PAE and ln(ξ) were chosen such that they were approximately close to the maximum values obtained using the known candidates. The reason for keeping these target values close to the maximum is because maximizing these properties may increase the formation of reactive species and hence result in improved photo-polymerization reaction of the monomers. In our previous studies, it was observed that DC of the hydrophilic-rich phase mimics was in the range of 63% to 97% given that the photo-initiator concentration was sufficient (Abedin et al., 2014; Abedin et al., 2015c). In another study, it was observed that the DC of the hydrophilic-rich phase for physically separated model dental adhesive system was approximately 20% to 32% in presence of a reaction accelerator, iodonium salt (Abedin et al., 2016). The low DC for this case was most likely due to the lack of either or both photosensitizer/co-initiator or low efficiency of the co-initiator within the hydrophilic-rich phase (Abedin et al., 2016). Based on our earlier studies on the mimics, the target value for the DC was chosen to be 80% (Abedin et al., 2014; Abedin et al., 2015c). The hydrophilic-rich phase mimics exhibited initial and secondary rates and the formation of microgels was associated with the secondary rate (Abedin et al., 2014; Abedin et al., 2015c). For the target value of the rate of polymerization, only the initial rate of the mimics were considered which ranged from 14.3 × 10-4 s-1 to 47 × 10-4 s-1 (Abedin et al., 2014; Abedin et al., 2015c). The secondary rates were significantly slower than the initial rates, and hence the shrinkage would most likely to be dominated by the initial rate. The target value for the rate of polymerization was chosen to be close to the average of the lowest and highest initial rates observed for the mimics in presence of sufficient photo-initiators (14.3 × 10-4 s-1 and 47 × 10-4 s-1). DC carried the highest scaling factor followed by log P and the rest of the properties carried equal but lowest scaling factor. The reasons for giving highest scaling factor to DC and then to log P are that a high DC is essential to ensure durable dentin/adhesive bond integrity within this phase, and it is critical that a hydrophilic photosensitizer is designed to ensure that it partitions at a higher concentration within the hydrophilic-rich phase. Four candidate photosensitizer molecules with low objective value out of 100 trials are shown in Figure 7 and Table 4 summarizes their properties along with the target values.

Figure 7. Molecular structure of the candidate photosensitizers.

Figure 7

Open in a new tab

Table 4. Summary of the properties of candidate photosensitizer molecules and the target values.

Photosensitizer ln(ξ) log P Relative Normalized PAE DC Rate (s-1) Objective
Target 9.00 -0.55 7.00 0.80 0.0030 0.00
Candidate 1 15.28 -0.47 13.10 0.76 0.0061 0.32
Candidate 2 14.24 -0.70 12.93 0.48 0.0019 0.33
Candidate 3 16.25 -0.37 13.83 0.79 0.0006 0.36
Candidate 4 11.66 -0.29 9.63 0.67 0.0076 0.42
Open in a new tab

As mentioned earlier with the conventional photosensitizer camphorquinone (CQ), almost no polymerization of the hydrophilic-rich phase was observed, and in the presence of an iodonium salt it could reach 20% to 32% depending on the co-initiator (Abedin et al., 2016). The proposed candidate molecules have a significantly higher DC which is an improvement over the camphorquinone for the hydrophilic-rich phase. The log P value of the candidates suggest that their hydrophilicity is also high. All the candidate molecules have iminium ions (C=NH2+ bond) in their structures, indicating that structures consisting this feature may be beneficial for achieving high DC. The C=N bond, C=C bond and benzyl groups within the candidate molecules are structural features which impart them the photosensitizing capability.

Conclusion

In this study, a framework to design water compatible visible light photosensitizer for application in dental adhesive were demonstrated. The target properties were molar extinction coefficient, octanol/water partition coefficient, relative normalized photon absorption efficiency (PAE), degree of conversion (DC) of the hydrophilic-rich phase and rate of polymerization (RT). The target properties are all linked to the photo-polymerization reaction of the adhesive monomers except for the property log P which gives a measure of the hydrophilicity of the photosensitizer. QSPRs developed using connectivity indices by linear regression exhibited high predictive capability. Four candidate photosensitizers have been proposed here and it was observed that they all possess iminium ion in their structures. The predicted DC of the hydrophilic-rich phase were higher for the candidate molecules than in the case with conventional photosensitizer CQ.

Highlights.

  • Framework to design water compatible photosensitizers for dental adhesives.

  • Insight into photosensitizer structure linked to improved dental adhesive lifetime.

  • Proposed four example candidate molecules showing high degree of conversion.

  • Iminium ions were identified to have positive impact on the degree of conversion.

Acknowledgments

This investigation was supported by the research grant R01 DE022054, National Institute of Dental and Craniofacial Research, National Institutes of Health, Bethesda, MD 20892

Footnotes

The opinions and information in this article are of the authors, and do not represent the views or policies of the FDA.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Abedin F. Characterization of hydrophilic-rich phase mimic in dentin adhesive and computer-aided molecular design of water compatible visible light initiators. The University of Kansas; 2016. [Google Scholar]
  2. Abedin F, Roughton B, Spencer P, Ye Q, Camarda K. Computational Molecular Design of Water Compatible Dentin Adhesive System. PSE2015 ESCAPE25. 2015a:233. [Google Scholar]
  3. Abedin F, Ye Q, Camarda K, Spencer P. Impact of light intensity on the polymerization kinetics and network structure of model hydrophobic and hydrophilic methacrylate based dental adhesive resin. Journal of Biomedical Materials Research Part B: Applied Biomaterials. 2015b:n/a–n/a. doi: 10.1002/jbm.b.33517. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Abedin F, Ye Q, Good HJ, Parthasarathy R, Spencer P. Polymerization- and solvent-induced phase separation in hydrophilic-rich dentin adhesive mimic. Acta Biomaterialia. 2014;10:3038–3047. doi: 10.1016/j.actbio.2014.03.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Abedin F, Ye Q, Parthasarathy R, Misra A, Spencer P. Polymerization Behavior of Hydrophilic-Rich Phase of Dentin Adhesive. Journal of Dental Research. 2015c;94:500–507. doi: 10.1177/0022034514565646. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Abedin F, Ye Q, Song L, Ge X, Camarda K, Spencer P. Effect of Partition of Photo-initiator Components and Addition of Iodonium Salt on the Photopolymerization of Phase-Separated Dental Adhesive. The Journal of the Minerals, Metals and Materials Society. 2016:68. doi: 10.1007/s11837-016-1816-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Bicerano J. Prediction of polymer properties. CRC Press; 2002. [Google Scholar]
  8. Braga RR, Ferracane JL. Contraction Stress Related to Degree of Conversion and Reaction Kinetics. Journal of Dental Research. 2002;81:114–118. [PubMed] [Google Scholar]
  9. Cadenaro M, Antoniolli F, Sauro S, Tay FR, Di Lenarda R, Prati C, Biasotto M, Contardo L, Breschi L. Degree of conversion and permeability of dental adhesives. European Journal of Oral Sciences. 2005;113:525–530. doi: 10.1111/j.1600-0722.2005.00251.x. [DOI] [PubMed] [Google Scholar]
  10. Camarda KV, Maranas CD. Optimization in Polymer Design Using Connectivity Indices. Industrial & Engineering Chemistry Research. 1999;38:1884–1892. [Google Scholar]
  11. Chavali S, Lin B, Miller DC, Camarda KV. Environmentally-benign transition metal catalyst design using optimization techniques. Computers & Chemical Engineering. 2004;28:605–611. [Google Scholar]
  12. Cook WD, Chen F. Enhanced photopolymerization of dimethacrylates with ketones, amines, and iodonium salts: the CQ system. Journal of Polymer Science Part A: Polymer Chemistry. 2011;49:5030–5041. [Google Scholar]
  13. Demarco FF, Corrêa MB, Cenci MS, Moraes RR, Opdam NJ. Longevity of posterior composite restorations: not only a matter of materials. Dental Materials. 2012;28:87–101. doi: 10.1016/j.dental.2011.09.003. [DOI] [PubMed] [Google Scholar]
  14. Pérez Mdel M, Pérez-Ocón F, Lucena-Martín C, Pulgar R. Stability and Reproducibility of Radiometric Properties of Light Curing Units (LCUs). Part I: QTH LCUs. Dental Materials Journal. 2008;27:284–291. [PubMed] [Google Scholar]
  15. Eslick JC, Ye Q, Park J, Topp EM, Spencer P, Camarda KV. A computational molecular design framework for crosslinked polymer networks. Computers & Chemical Engineering. 2009;33:954–963. doi: 10.1016/j.compchemeng.2008.09.019. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Folić M, Adjiman CS, Pistikopoulos EN. Computer-Aided Solvent Design for Reactions: Maximizing Product Formation. Industrial & Engineering Chemistry Research. 2008;47:5190–5202. [Google Scholar]
  17. Fujisawa S, Masuhara E. Determination of partition coefficients of acrylates, methacrylates, and vinyl monomers using high performance liquid chromatography (HPLC) Journal of Biomedical Materials Research. 1981;15:787–793. doi: 10.1002/jbm.820150603. [DOI] [PubMed] [Google Scholar]
  18. Gani R, Harper PM, Hostrup M. Automatic Creation of Missing Groups through Connectivity Index for Pure-Component Property Prediction. Industrial & Engineering Chemistry Research. 2005;44:7262–7269. [Google Scholar]
  19. Gani R, Nielsen B, Fredenslund A. A group contribution approach to computer-aided molecular design. AIChE Journal. 1991;37:1318–1332. [Google Scholar]
  20. Hashimoto M, Nagano F, Endo K, Ohno H. A review: Biodegradation of resin–dentin bonds. Japanese Dental Science Review. 2011;47:5–12. [Google Scholar]
  21. Lim BS, Ferracane JL, Sakaguchi RL, Condon JR. Reduction of polymerization contraction stress for dental composites by two-step light-activation. Dental Materials. 2002;18:436–444. doi: 10.1016/s0109-5641(01)00066-5. [DOI] [PubMed] [Google Scholar]
  22. Lumely T. Package ‘Leaps’ 2009 [Google Scholar]
  23. Marrero J, Gani R. Group-contribution based estimation of pure component properties. Fluid Phase Equilibria. 2001:183–184. 183–208. [Google Scholar]
  24. Neumann MG, Miranda WG, Jr, Schmitt CC, Rueggeberg FA, Correa IC. Molar extinction coefficients and the photon absorption efficiency of dental photoinitiators and light curing units. Journal of Dentistry. 2005;33:525–532. doi: 10.1016/j.jdent.2004.11.013. [DOI] [PubMed] [Google Scholar]
  25. Pashley DH, Tay FR, Yiu C, Hashimoto M, Breschi L, Carvalho RM, Ito S. Collagen Degradation by Host-derived Enzymes during Aging. Journal of Dental Research. 2004;83:216–221. doi: 10.1177/154405910408300306. [DOI] [PubMed] [Google Scholar]
  26. Pellosi DS, Estevão BM, Semensato J, Severino D, Baptista MS, Politi MJ, Hioka N, Caetano W. Photophysical properties and interactions of xanthene dyes in aqueous micelles. Journal of Photochemistry and Photobiology A: Chemistry. 2012;247:8–15. [Google Scholar]
  27. R-Development-Core-Team. R: A language and environment for for statistical computing 2010 [Google Scholar]
  28. Randic M. Characterization of molecular branching. Journal of the American Chemical Society. 1975;97:6609–6615. [Google Scholar]
  29. Roughton BC, Christian B, White J, Camarda KV, Gani R. Simultaneous design of ionic liquid entrainers and energy efficient azeotropic separation processes. Computers & Chemical Engineering. 2012a;42:248–262. [Google Scholar]
  30. Roughton BC, Topp EM, Camarda KV. Use of glass transitions in carbohydrate excipient design for lyophilized protein formulations. Computers & Chemical Engineering. 2012b;36:208–216. doi: 10.1016/j.compchemeng.2011.07.018. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Santini A, Miletic V. Quantitative micro-Raman assessment of dentine demineralization, adhesive penetration, and degree of conversion of three dentine bonding systems. European Journal of Oral Sciences. 2008;116:177–183. doi: 10.1111/j.1600-0722.2008.00525.x. [DOI] [PubMed] [Google Scholar]
  32. Sheikh MI. Renal handling of phenol red. II. The mechanism of substituted phenolsulphophthalein (PSP) dye transport in rabbit kidney tubules in vitro. The Journal of physiology. 1976;256:175–195. doi: 10.1113/jphysiol.1976.sp011319. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Shokati B, Tam LE, Santerre JP, Finer Y. Effect of salivary esterase on the integrity and fracture toughness of the dentin-resin interface. Journal of Biomedical Materials Research Part B: Applied Biomaterials. 2010;94B:230–237. doi: 10.1002/jbm.b.31645. [DOI] [PubMed] [Google Scholar]
  34. Siddhaye S, Camarda KV, Topp E, Southard M. Design of novel pharmaceutical products via combinatorial optimization. Computers & Chemical Engineering. 2000;24:701–704. [Google Scholar]
  35. Spencer P, Wang Y. Adhesive phase separation at the dentin interface under wet bonding conditions. Journal of Biomedical Materials Research. 2002;62:447–456. doi: 10.1002/jbm.10364. [DOI] [PubMed] [Google Scholar]
  36. Spencer P, Ye Q, Park J, Topp E, Misra A, Marangos O, Wang Y, Bohaty B, Singh V, Sene F, Eslick J, Camarda K, Katz JL. Adhesive/Dentin Interface: The Weak Link in the Composite Restoration. Annals of Biomedical Engineering. 2010;38:1989–2003. doi: 10.1007/s10439-010-9969-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Stahl F, Ashworth SH, Jandt KD, Mills RW. Light-emitting diode (LED) polymerisation of dental composites: flexural properties and polymerisation potential. Biomaterials. 2000;21:1379–1385. doi: 10.1016/s0142-9612(00)00029-6. [DOI] [PubMed] [Google Scholar]
  38. Struebing H, Ganase Z, Karamertzanis PG, Siougkrou E, Haycock P, Piccione PM, Armstrong A, Galindo A, Adjiman CS. Computer-aided molecular design of solvents for accelerated reaction kinetics. Nat Chem. 2013;5:952–957. doi: 10.1038/nchem.1755. [DOI] [PubMed] [Google Scholar]
  39. Venkatasubramanian V, Chan K, Caruthers JM. Computer-aided molecular design using genetic algorithms. Computers & Chemical Engineering. 1994;18:833–844. [Google Scholar]
  40. Venkatraman V, Alsberg BK. A quantitative structure-property relationship study of the photovoltaic performance of phenothiazine dyes. Dyes and Pigments. 2015;114:69–77. [Google Scholar]
  41. Venkatraman V, Åstrand PO, Kåre Alsberg B. Quantitative structure-property relationship modeling of Grätzel solar cell dyes. Journal of computational chemistry. 2014;35:214–226. doi: 10.1002/jcc.23485. [DOI] [PubMed] [Google Scholar]
  42. Wagner SJ, Skripchenko A, Robinette D, Foley JW, Cincotta L. Factors affecting virus photoinactivation by a series of phenothiazine dyes. Photochemistry and photobiology. 1998;67:343–349. [PubMed] [Google Scholar]
  43. Wainwright M, Burrow S, Guinot SR, Phoenix D, Waring J. Uptake and cell-killing activities of a series of Victoria blue derivatives in a mouse mammary tumour cell line. Cytotechnology. 1999;29:35–43. doi: 10.1023/A:1008098810928. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Ye Q, Park J, Parthasarathy R, Pamatmat F, Misra A, Laurence JS, Marangos O, Spencer P. Quantitative analysis of aqueous phase composition of model dentin adhesives experiencing phase separation. Journal of Biomedical Materials Research Part B: Applied Biomaterials. 2012;100B:1086–1092. doi: 10.1002/jbm.b.32675. [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES