Polymer Coating Materials and Their Fouling Release Activity: A

Subscriber access provided by GAZI UNIV

Article

Polymer coating materials and their fouling release activity: A cheminformatics approach to predict properties Bakhtiyor Rasulev, Farukh Jabeen, Shane J. Stafslien, Bret J. Chisholm, James Bahr, Martin Ossowski, and Philip Boudjouk ACS Appl. Mater. Interfaces, Just Accepted Manuscript • DOI: 10.1021/acsami.6b12766 • Publication Date (Web): 16 Dec 2016 Downloaded from http://pubs.acs.org on December 17, 2016

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

ACS Applied Materials & Interfaces is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 38

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Applied Materials & Interfaces

Polymer coating materials and their fouling release activity: A cheminformatics approach to predict properties Bakhtiyor Rasuleva,c*, Farukh Jabeena, Shane Stafslienb, Bret J. Chisholmc, James Bahrb, Martin Ossowskia, Philip Boudjouka,d*

a

Center for Computationally Assisted Science and Technology, North Dakota State University,

Fargo, ND, United States b

c

Research and Creative Activities, North Dakota State University, Fargo, ND, United States

Department of Coatings and Polymeric Materials, North Dakota State University, Fargo, ND,

United States d

Department of Chemistry and Biochemistry, North Dakota State University, Fargo, ND, United

States

* - corresponding author

ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Abstract A novel cheminformatics-based approach has been employed to investigate a set of polymer coating materials designed to mitigate the accumulation of marine biofouling on surfaces immersed in the sea. Specifically, a set of 27 non-toxic, amphiphilic polysiloxane-based polymer coatings was synthesized using a combinatorial, high-throughput approach and characterized for fouling-release (FR) activity towards a number of relevant marine fouling organisms, including bacteria, microalgae and adult barnacles. In order to model these complex systems adequately, a new computational technique was used in which all investigated polymer-based coating materials were considered as mixture systems comprising several compositional variables at a range of concentrations. By applying a combination of methodologies for mixture systems and a quantitative structure-activity relationship approach (QSAR), seven unique QSAR models were developed that were able to successfully predict the desired FR properties. Furthermore, the developed models identified several significant descriptors responsible for FR activity of investigated polymer-based coating materials, with correlation coefficients ranging from r2test=0.63 to 0.94. The computational models derived from this study may serve as a powerful set of tools to predict optimal combinations of source components to produce amphiphilic polysiloxane-based coating systems with effective, broad-spectrum FR properties.

Keywords: polymer coating materials, QSAR, mixture, poly-siloxane, antifouling, foulingrelease


Page 2 of 38

Page 3 of 38

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Introduction Coating materials are widely used in industry and medicine. In this regard, polymer-based coating materials have innumerable applications because they are versatile, cost-effective, and can be tailored to a variety of specifications

1-4

. The science of polymer synthesis allows for

excellent control over the properties of a bulk polymer system. However, control over surface properties is a major challenge in modern polymer science. It is particularly formidable to predict the surface property of complex polymer materials when the initial mixture of polymer components changes 5-6. Marine biofouling affects all submerged surfaces causing detrimental effects on shipping and leisure vessels, heat exchangers, oceanographic sensors and aquaculture systems. Biofouling of ship hulls is an ongoing issue that has major economic and environmental impact. Biofouling increases hydrodynamic drag that translates to dramatic reductions in fuel efficiency

7-8

. In this

regard, early antifouling (AF) strategies involved embedding metal and/or organometallic-based biocides into the polymer matrix of an AF paint. For example, in the 1960s, organotin-based AF coatings were developed and found to be highly effective 9. However, the broad-spectrum activity of the organotin group, primarily tributyltin (TBT), resulted in adverse effects on nontargeted organisms 10, and was found to be unacceptably persistent in the marine environment 11. As a result, the International Maritime Organization Convention on the Control of Harmful AF Systems banned the use of TBT in marine coatings after 2008. Due to this ban, new AF coatings have been primarily based on copper-containing biocides 12. Although the toxic effects of copper to non-target organisms are not as severe as those of tin, the use of copper still represents a serious environmental concern.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 38

As a non-toxic alternative to AF coatings, fouling-release (FR) coatings were developed and commercialized. Unlike AF coatings, FR coatings do not contain biocides, and combat ship hull fouling by minimizing the adhesion strength between the attached organism and the coating surface so that the fouling can be removed through hydrodynamic shearing. Some relationships between the type of surface (hydrophobic or hydrophilic) and fouling adhesion was discussed by Krishnan et al

13-16

., where the authors indicated that different marine organisms have different

adhesion properties to hydrophobic and hydrophilic surfaces. For example, the authors showed that while Navicula cells released more easily from hydrophilic surfaces, Ulva sporelings showed higher removal from hydrophobic surfaces. Ideally, the adhesion strength between the attached fouling organisms and coating surface is low enough to enable release of the majority of the attached fouling when the vessel is underway. Thus, commercially available FR coatings are generally based on polysiloxanes

17

It is generally accepted that the relatively good FR

performance of polysiloxane-based coatings is a result of both the low surface energy of polysiloxanes and their low moduli

18-19

. As a result, it has been of interest to investigate the

modification of polysiloxanes with hydrophilic moieties to reduce biofouling. For example, polydimethylsiloxane (PDMS) elastomers were surface modified with PEOs using platinumcatalyzed hydrosilation and modified surfaces provided a 90% reduction of fibrinogen adsorption compared to the unmodified PDMS control 20. Polysiloxane-based amphiphilic polymer coatings materials have been recently investigated for the fouling release (FR) activity by other researchers

21-22

. In previous work

22

these materials

were synthesized using combinatorial analysis and investigated for the best components’ composition to obtain a polymer coating with desired properties. However, combinatorial analysis enables synthesis of a limited number of polymer coating materials and does not directly


Page 5 of 38

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


specify the main structural factors that are responsible for imparting optimal release of biofouling from coating surfaces. In this case, computational methods, such as quantitative structure-activity/property relationships (QSAR/QSPR) can be helpful to in elucidating these key attributes and generate ideas for further improvement of these coating systems. QSAR analysis is based on the premise that the structure of a molecule is the principal determinant of its physicochemical, toxicological, and biomedicinal properties

23-26

. Presently,

QSAR is widely applied to the development of rationales to enhance desirable properties by tuning the structure within the congeneric series of compounds, including polymers. This popular approach utilizes statistical methods to determine a correlation amongst structural features of compounds and the studied property(s). Most QSAR/QSPR models tend to follow a similar strategy 27. This includes the following: (1) data set selection; (2) molecular structure generation; (3) geometry optimization of molecular structures using appropriate procedures; (4) various descriptors generation; (5) variable selection and/or data reduction methods; (6) model generation; (7) and validation and predictability evaluation of the developed models 28-29. To date, there are only a few publications related to QSARs of polymers, since polymers are inherently

complicated

macromolecular

systems,

possessing

large

conformationally-labile structures and non-systematic cross-linking, etc

molecular 30-32

sizes,

. For example,

attempts have been made to develop special descriptors that encode polymer structures 31. As the next step, several efforts were made to build QSAR models for various properties of polymers 30, 32

. For this purpose, researchers often use the information based on monomer structure 30, 32. Yu

et al 32 developed two artificial neural network (ANN) models to predict reactivity parameters ln Q and e of acrylate monomers, applying data from density functional theory (DFT) calculations at the B3LYP/6-31G(d,p) level. The authors found that the resonance and polar effects of



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 38

acrylate monomers can be reflected by quantum-chemical descriptors such as Mulliken and atomic polar tensor (APT) charges, the total dipole moment (µT), the lowest unoccupied molecular orbital energy (ELUMO) and the total energy (ET). In other work, authors developed a QSPR model for refractive indices of 234 structurally diverse polymers

30

. The reported model

involved a single molecular descriptor and a conformation-independent approach, in which the most appropriate polymer structures were investigated by considering 1–5 monomeric repeating units. Another study, which is very close to the current study attempting to investigate and predict surface adhesion property of a library of 496 polymers, was published by Winkler and co-authors33. The authors used a dataset provided by Yang et al 34 and applied a QSAR approach to build neural network predictive models. The authors generated satisfactory models with r2test=0.63, however, no detailed information on the dataset and which polymer structures were used for descriptors generation (linear or cross-linked) were provided in this paper, which makes it difficult to reproduce the results and investigate further applications. In this regard, systematic and reproducible investigations of a number of datasets is still needed for these kinds of systems, to more adequately address new challenges in this field. For polymer coating systems, the situation is even more complicated, since these materials often comprise highly cross-linked polymers, combining several different monomers in the polymeric system. As a result of this complexity, the classical QSAR approaches are not readily applicable for polymer coating systems and no publications on this topic are presently known. Molecular modeling approaches to deal with polymer coatings, computational investigations, and rational design have been unsuccessful because of the highly complicated structures of the systems which prevent direct modeling.


Page 7 of 38

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


As mentioned above, the prediction of surface properties for complex polymer coating systems when the initial mixture of polymer substrates changes is an exceedingly difficult and challenging task. However, we recently reported preliminary studies with application of a novel mixture-QSAR approach for polymer coatings

35-36

. In this paper, an attempt to develop a first

QSAR-based model(s) is reported which provides a possible avenue in which to computationally predict FR properties of the coatings investigated. Described herein is a computational methodology applied to amphiphilic polysiloxane coatings and structure-activity modeling results obtained for several marine organisms, including bacteria, microalgae and adult barnacles.

Materials and Methods Polymer coating materials A set of 27 polymer coatings was produced using a combinatorial approach and characterized for activity towards a number of relevant marine fouling organisms, including bacteria, microalgae and adult barnacles

22

(Table 1). The details on coating preparation and fouling-release

experiments were described previously22. An automated water-jet method was used to evaluate the adhesion of two marine bacteria in a rapid manner, Cellulophaga lytica and Halomonas pacifica, and a microalgae diatom, Navicula incerta, to coatings prepared in multi-well plates using methods that have been previously described in detail12b. In addition, the coatings were evaluated for their ability to prevent or minimize the adhesion strength of barnacles (Amphibalanus Amphitrite) using a rapid laboratory reattachment assay that has been previously described in detail22. All components of the polymer coatings were identified structurally and characterized computationally 21,22.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

In order to model these complex systems, all coatings investigated were considered as mixture systems, having components in various initial concentrations (Fig. 1). The main components involved in the study are listed in Table 2. A set of parameters/descriptors was generated that encoded each component of the system. Descriptors generation: To relate components’ structure changes to FR properties for each initial component of the polymer coating, a set of properties/descriptors was computationally generated that encoded the chemical structure. The structures were built and prepared for further use by Chemaxon software37. Dragon 6 software 38 was used to generate a set of descriptors. This software provides more than 4500 various descriptors corresponding to 0D-, 1D-, 2D-, and 3Dindexes. The descriptors are comprised of 20 different classes - constitutional, topological, walk and path counts, connectivity indices, information indices, 2D autocorrelations, edge adjacency indices, Burden eigenvalues, topological charge indices, eigenvalue based indices, randic molecular profiles, geometrical descriptors, RDF descriptors, 3D-MoRSE descriptors, WHIM descriptors, GETAWAY descriptors, functional groups, atom-centered fragments, charge descriptors, and molecular property descriptors

38-39

. Constant and near to constant descriptors

were eliminated. After filtering out constant and close to constant descriptors, about 1200 descriptors in total were generated per each main component of the polymer coating. Mixture Descriptors Due to the complexity of polymer coatings a new approach to describe the polymer coating system was applied. For this work, the following methodology was utilized. Each polymer coating was considered as a mixture system, where components were in various concentrations, and descriptors for each polymer coating were calculated based on structures and concentrations


Page 8 of 38

Page 9 of 38

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


of each component (so-called mixture descriptors) in the coating. Figure 1 represents the overall idea and complexity of the polymer coating.

Figure 1. A representation of a highly cross-linked polymer coating (represented as a mustard colored material within blue box) derived from multiple reactive components.

To implement this approach the following equations were applied: Dmix = f (C1×Dn , C2×Dn , C3×Dn , C4×Dn … Cn×Dn),

(1)

where Dmix is a mixture descriptor, Cn - concentration of an individual component in the mixture and Dn is a descriptor of individual component Amix = f (P1, P2, P3, P4, ….Pn),

(2)

where Amix – activity of the mixture system, P – component property, P = (Cn×Dmix_n)

Amix = (C1×Dmix1) + (C2×Dmix2)...+(Cn×Dmix3), where C – coefficient, Dmix_n – mixture descriptor, Amix – activity of the mixture system


(3)


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 38

The first equation, Eq. 1, shows how the mixture descriptors were obtained, by including the same type of descriptor (Dn) from each component to describe the mixture system. Eq. 2 represents the activity of the mixture system that depends on properties of the components in this system. Next equation, Eq. 3, represents the summary of Eq. 1 and Eq. 2, which can be considered as an overall MLR-based QSAR equation, covering the presence of mixture descriptors and coefficients in the model. The overall scheme of the applied approach is represented in Fig. 2.

Figure 2. The scheme of applied mixture-QSAR approach to polymer coatings, where Comp 1…Comp 5 are components of the polymer coating, C1…C5 are concentrations of components, and Dn are descriptors’ values for the components.


Page 11 of 38

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


QSAR modeling: After the calculation of all mixture descriptors the next step was QSAR modeling. For that, the set of 27 coatings was divided into training (21 coatings) and test sets (6 coatings), 75% training set and 25% test set, respectively. Overall, QSAR is a mathematical relationship between a biological activity of a molecular system and its geometric, chemical or physical characteristics. QSAR attempts to find a consistent relationship between biological activity and molecular properties, so that these “rules” can be used to evaluate the activity of new chemical systems. Once a valid QSAR has been determined, it should be possible to predict the physical property or biological activity of related compounds or drug candidates before they are put through expensive and time-consuming biological testing. In some cases, only computed values need to be known to make an assessment. In the current study, the correlation between activity and structural properties was developed by using the variable selection Genetic Algorithm (GA) and Multiple Linear Regression Analysis (MLRA) methods. Thus, preliminary model selection was performed by means of the GAMLRA 40-43 technique as implemented in the QSARINS 2.2 44-45 program. It is worth noting that genetic algorithms have been applied in recent studies as a powerful tool to address many problems in QSAR studies

40-43

. The method is based on the mechanism of natural evolution,

where the higher descriptor weights are more preserved in the mathematic evolution process and finally in the model, while the lower weight descriptor is eliminated. In this study, the GA variable selection technique was used to reduce number of descriptors that is applied for final model and GA variable selection started with a population of 500 random models and 2000 iterations to evolution with the mutation probability specified at 40%. The MLRA technique was



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 12 of 38

used to develop final QSAR models, since it is transparent, easy to interpret, and ideal to obtain reproducible results. Several QSAR models were developed (one model per each endpoint), followed by statistical analysis with evaluation by squared correlation coefficient r2, root mean square error RMSE, Fisher coefficient F, and non-collinearity of descriptors in the model. A final set of QSARs was generated by using GA-MLR approach and tested by applying the “leave-one-out” technique (the process of removing a molecule from the set, then creating and validating the model against the individual molecules, which was performed for the entire training set), q2. Thus, we utilized the following equations to calculate correlation coefficient, r2 (eq. 4) and the root mean square error of calibration (training) RMSEC, as the measures of goodness-of-fit for each developed model (eq.5): = 1 −

)

(4)

)

(5)

∑

(

∑ ) (

∑ ( =

To verify stability of the models (sensitivity to the composition of the training set), in each case we calculated the cross-validated coefficient q2LOO (leave-one-out method) and root mean square error of cross-validation RMSECV. Both statistics were calculated according to Eq. 6 and Eq.7: !! = 1 −

"# )

∑

(

% =

) ∑ $($

(6)

"# )

∑ (

(7)

Following the recommendations by Gramatica and Lin

46-47

, we calculated Concordance

Correlation Coefficient (CCC) as a more restrictive parameter for expressing external


Page 13 of 38

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


predictivity of each model in comparing to the commonly used external (test set) validation coefficient q2Ext (eq.8) and root mean square error of prediction (RMSEP) (eq.9). In this work, we applied all these three statistics. &'( = 1 − ∑.*/0()*+,- −)*1234 ) / ∑.*/0()*+,- −)6 +,- )

(8) 7 = 8

)

∑9 ($ $

.

(9)

where: yjobs – experimental (observed) value of the property for the ith/ jth compound; yjpred – predicted value for ith/ jth compound; ), ŷ – the mean experimental value of the property in the training and validation set, respectively; n, k – the number of compounds in the training and validation set, respectively.

Additionally, the chemical applicability domain (AD) for the models obtained was calculated by the leverage approach to verify predictive reliability 29, 48. To visualize the applicability domain of the QSPR models, the Williams plot was used. Thus, Williams plot of standardized crossvalidated residuals (RES) vs. leverage (Hat diagonal) values (HAT) clearly depicts both the response outliers (Y outliers) and structurally influential compounds (X outliers) in a model. All initial descriptors were normalized before calculation of mixture descriptors. The normalization of descriptors was carried out using Matlab 49. Decision Trees method. The classification analysis was established by Decision Trees method 50 usingMatlab 49. The quality of all developed models was determined using equations 10, 13, 14: (EF(G

:;

Polymer Coating Materials and Their Fouling Release Activity: A

Recommend Documents