QSAR study and the hydrolysis activity prediction of three alkaline lipases from different lipase-producing microorganisms
© Wang et al.; licensee BioMed Central Ltd. 2012
Received: 14 August 2012
Accepted: 12 September 2012
Published: 28 September 2012
The hydrolysis activities of three alkaline lipases, L-A1, L-A2 and L-A3 secreted by different lipase-producing microorganisms isolated from the Bay of Bohai, P. R. China were characterized with 16 kinds of esters. It was found that all the lipases have the ability to catalyze the hydrolysis of the glycerides, methyl esters, ethyl esters, especially for triglycerides, which shows that they have broad substrate spectra, and this property is very important for them to be used in detergent industry. Three QSAR models were built for L-A1, L-A2 and L-A3 respectively with GFA using Discovery studio 2.1. The models equations 1, 2 and 3 can explain 95.80%, 97.45% and 97.09% of the variances (R 2 adj ) respectively while they could predict 95.44%, 89.61% and 93.41% of the variances (R 2 cv ) respectively. With these models the hydrolysis activities of these lipases to mixed esters were predicted and the result showed that the predicted values are in good agreement with the measured values, which indicates that this method can be used as a simple tool to predict the lipase activities for single or mixed esters.
KeywordsQSAR GFA Ester Lipase activity Vegetable oils
Lipases are defined as triacylglycerol acylhydrolases (E.C. 188.8.131.52) that catalyze the hydrolysis of oils and fats at the oil–water interface to free fatty acids and glycerol. Microbial lipases have been proven to be useful biocatalysts for obtaining chiral, non-racemic compounds. Lipase from Burkholderia cepacia can efficiently the reaction of catalyze hydrolysis, alcoholysis, transesterification, aminolysis, acidolysis, and esterification[1–3]. In order to improve the usefulness of lipases as biocatalysts, an understanding of the lipase application in daily life is needed. They directly or indirectly form an integral part of the industries ranging from food, pharmaceuticals, and detergents[5, 6] to organic synthesis, cosmetics, leather, and tea industries. However, the single biggest market for their use is in detergents where their functional importance lies in the removal of fatty residues in laundry, dishwashers, and for cleaning clogged drains. Though the lipase function is usually connected with enzyme activity, the higher enzyme activity, the better washing performance is, sometimes the washing performance is not fully consistent with the lipase activity. One reason is that there are different methods for the determination of lipase activity. At present, the lipase activity is usually determined by titrimetric methods, spectrophotometry, nephelometry and turbidimetry, electric conductivity, and so on. And each of them based on a specific property of the lipase reaction system, which leads to the different activity measuring values for the same lipase. The other reason is that the substrates in detergency ability evaluation are different from that in the determination of lipase activities. In the washing performance evaluation, the substrates used are usually mixture of different fats or oil, for example, lipase decontamination capability was measured using emulsified olive oil as the substrate[15, 16]. Decontamination capability is related to lipase activity, while animal fat and plant oils are main oil pollution daily in our lives. The main components of these oil pollutions are triglyceride, diacylglycerol, free fatty acid, etc. However, the substrate used in the determination of lipase activity is usually a pure matter, the difference in substrates result in the difference between the washing performance and the activity of lipase. There are different lipase activities for different substrates, which results from the differences in substrate composition and structure. A lipase with better detergency ability should have higher hydrolysis ability to a broad spectrum of esters. In order to obtain comprehensive understanding of the lipase activity and substrate spectrum, the substrates with various composition and structure are required to evaluate them, further, a quantitative structure and activity relationship should be built. There are some studies on this aspect, for example, there are two distinct modeling strategies for predicting lipase activity highlights: structure-based approach and data-driven approach. The structure-based models start with a known active site structure of the lipase[17–19] and then identify the preferred substrates based on conformation, charge, and other force field calculations[20, 21]. On the other hand, data driven models such as quantitative structure–activity relationship (QSAR) approach develops a mathematical relationship between the enzyme activity and structural descriptors of substrates using available experimental data. In context of lipases, such QSAR approach has been reported in predicting the substrate specificity and enantioselectivity of a lipase in esterification/trans-esterification reactions. However, there are few reports on the systematic evaluation of the lipase detergency ability using different substrates existed in oil spill.
Previously, three kinds of lipase from the soil collected from the Bay of Bohai, P. R. China was found by our laboratory including Burkholderia cepacia L-A1, Acinetobacter johnsonii L-A2[25, 26] and Acinetobacter calcoaceticus L-A3. They have highly stability in the presence of various oxidizing agents, some commercial detergents and alkaline protease. The three enzymes hydrolyzed a wide range of oils and showed a high level of lipase activity in hydrolyzing glyceride. In order to systematic evaluate its ability to hydrolyze different esters including some usually existed in edible oils and fats, this study derived some quantitative structure and activity relationships (QSARs) between the experimental results and structural parameters important for the substrate specificity of Burkholderia cepacia L-A1, Acinetobacter johnsonii L-A2 and Acinetobacter calcoaceticus L-A3 towards triglyceride, ethyl oleate, methyl laurate and allyl phenylacetate, etc. Meanwhile, this study will be useful for developing a standard for lipase evaluation with their detergency ability.
Materials and methods
Alkaline lipase-producing microorganisms were isolated from the Bay of Bohai, P.R. China and they were numbered as Burkholderia cepacia L-A1, Acinetobacter johnsonii L-A2 and Acinetobacter calcoaceticus L-A3, respectively. Refined, edible vegetable oils were purchased locally. Glycerol tripalmitate from Alfa Aesar Chemical Co., LTD (Tianjin, China). Methyl hexadecanoate, methyl myristate, methyl laurate, methyl linoleate, ethyl tetradecanoate, ethyl palmitate, ethyl tetradecanoate and ethyl linoleate were from TCI (Shanghai, China). Triarachidin, ethyl palmitate, ellyl phenylacetate, eripalmitin were from Tokyo Chemical Industry Company (Japan). Methyl oleate, methyl gallate, mlyceryl monostearate, glycerol trioleate, ethyl oleate, ethyl stearate and allyl phenylacetate were bought from Sinopharm Medicine Holding Co., Ltd (Tianjin, China).
Lipase activity determination
X, enzyme activity,U/g (U/ml).
B, sample consumption volume of standard sodium hydroxide solution for titration, ml.
A, blank sample consumption volume of standard sodium hydroxide solution for titration, ml.
c, standard sodium hydroxide concentration, mol/L.
0.05, conversion factor of sodium hydroxide concentration of standard solution.
50, 1ml sodium hydroxide solution (0.05 mol/L) equivalent to 50μmol fatty acid.
1/60, the reaction time of 60 min with 1 min count.
In this study, the 17 esters commercially availed listed in Additional file1: Table S1 were used as substrates to examine 3 lipase activities, the ester hydrolytic activity data of three lipases determined using spectrophotometry were also listed.
Generation of the 3D structure of the esters
The ester series were further subjected to molecular modeling studies using ChemBioOffice Software version 11. The 2D structure of the ester compounds was drawn in ChemBioDraw Ultra version 11 and then copied to Chem 3D Ultra version 11 to create the three-dimensional (3D) model. These structures were then subjected to energy minimization using molecular mechanics (MM2). The minimized molecules were further subjected to optimization via the Austin model 1 (AM1) method using the closed-shell (restricted) wave function of the Gamess.
Descriptors for QSAR
More than 120 physiochemical properties of the esters used as descriptors for QSAR construction were obtained using the “Calculate Molecular Properties” module of the Discovery Studio 2.1 package. These descriptors include 2D (AlogP, Molecular_SurfaceArea, Num_RotatableBonds, Num_H_Donors, Molecular_Weight, Kappa_1 topological descriptors such as CIC, CHI_3_C, IAC_Mean, BIC, IC, IAC_Total and SIC, etc.) and 3D (Jurs descriptors, Dipole, Molecular Volume and shadowindices, etc.) parameters. All the definition of the descriptors can be seen in the help of DS2.1. The lipase activity in A U/ml was converted to the logarithmic scale before used for subsequent QSAR analyses as the response variable.
QSAR model development
The obtained QSAR models which are developed from the training set should be validated using new esters for checking the predictive ability of the developed models. Thus the original data set is divided into training and test sets for QSAR model development and validation respectively. The ability of a model to predict accurately the target property of compounds that were not used for the model development is based on the fact that a molecule which is structurally similar to the training set molecules will be predicted well because the model has captured features that are common to the training set molecules and is able to find them in the new molecule. In our study, the whole data set (n =16) was divided into training (n =12) and test (n =4) sets by function groups. This approach (clustering) ensures that the similarity principle can be employed for the lipase activity prediction of the test set. The splitting has been performed such that points representing both training and test sets are distributed within the whole descriptor space of the entire dataset, and each point of the test set has a closer point of the training set. Compared with the number of molecular physiochemical properties, the training set is comparatively very small. In order to obtain the model with statistical meaning, these properties should be cut down and the most suitable descriptors will be left for the final model. The difficult thing is how to select which properties as the most suitable descriptor set to build QSAR models. In this study, the genetic function approximation (GFA) technique was employed to deal with this problem. The principles of GFA can be seen elsewhere[34, 35]. It uses the multivariate adaptive regression algorithm accompanied with the genetic algorithm (GA) to evolve population of models (each model containing a subset of variables) that best fit the training set data. With this methodology, a series of potential QSAR models (the population of organisms) are generated and tested repeatedly until an approximate optimal solution is reached finally. In this study, the QSAR models having different numbers of descriptor terms were selected by GFA and all the descriptors in the QSAR trial descriptor pool were used as linear terms. Subsequently, genetic partial least squares (G/PLS) module was employed to optimize the obtained model further.
Statistical quality assessment and model validation method
The successful QSAR model should be robust enough to make accurate and reliable predictions of the lipase activities, thus, the obtained QSAR models from the training set should be subsequently validated. There are several methods to evaluate the quality of QSAR models. In this study, Friedman lack-of-fit (LOF) was selected as the rule for the selection of the GFA derived equations, while correlation coefficient R2 and adjusted R2 (R 2 adj ), were taken as objective functions for G/PLS equations’ selection. The predictivity of generated QSAR models were finally validated using leave-one-out cross-validation R2 (R 2 cv ). Because the descriptor number available normally exceeds that of the samples (training set compounds), how to prevent over-fitting of GFA is critical to the successful construction of a statistically significant QSAR model. In this study, the QSAR models having different numbers of descriptor terms were selected by GFA and all the descriptors in the QSAR trial descriptor pool were used as linear terms. LOF is designed to control the model size and to avoid the over-fitting. The smoothing factor was set to 0.5, the optimal QSAR model was considered to be obtained when descriptors used became constant and independent of an increasing number of crossover operations. All the descriptors were used as linear terms during the GFA to generate QSAR models in the QSAR trial descriptor pool.
QSAR model predictivity for the lipase hydrolysis ability to some natural mixed esters
Compositions of the vegetable oils
Composition vegetable oil
11.0 ± 0.8(%)
4.5 ± 0.4(%)
20.7 ± 1.0(%)
8.9 ± 1.8(%)
54.2 ± 2.4(%)
14.5 ± 1.3(%)
2.5 ± 1.2(%)
70.0 ± 1.0(%)
1.5 ± 0.5(%)
12.0 ± 1.0(%)
X mix , the lipase activity for hydrolysis of the natural oil (U/ml).
X i , the lipase activity for hydrolysis of i oil ester component.
y i , the proportion of fatty acid glycerides.
n, the ester numbers contained in natural oil.
mi, molar fraction of each triglycerides contained in natural oil with mass fractions >1%.
QSAR model predictivity for the lipase hydrolysis ability to natural mixed esters was assessed by the comparison of the X mix obtained from the experiment with that obtained from QSAR models.
Results and discussion
Activity comparison of three lipases
QSAR Modeling with 2D and 3D combined set of descriptors
The sample number N = 12; LOF = 0.0024; R2 =0.9841; R 2 adj =0.9709; R 2 cv =0.898; F = 74.44
Observed and predicted L-A1 activities, physiochemical properties of different substances from DS 2.1 used for the construction of QSAR models
Observed and predicted L-A2 activities, physiochemical properties of different Substances from DS 2.1 used for the construction of QSAR models
Observed and predicted L-A3 activities, physiochemical properties of different Substances from DS 2.1 used for the construction of QSAR models
In this study, R2, R 2 adj , R 2 pre and R 2 cv were employed to evaluate the obtained models. Eq.3, 4 and 5 can explain 96.94%, 97.45% and 97.09% of the variances (R 2 adj ) respectively while they could predict 88.9%, 95.4% and 89.8% of the variances (R 2 cv ) respectively. F > F(a=0.05) showes that the models are those for a (non-multiplicity-corrected) confidence level of 0.95. It can be seen from Equation 3 that Molecular_Volume and Jurs_PPSA_3 have positive contribution to the bioactivity of the lipase. However, Molecular_PolarSASA, ALogP_MR and Shadow_XYfrac have the negative effect on the bioactivities of the lipase L-A1.
The standardized regression coefficient for each variable is 54.54, 39.42, 6.048, 0.6085 and 19.85 respectively. Therefore, the relative importance of the descriptors according to their standardized regression coefficients is in the following order:
ALogP_MR>Molecular_Volume>Jurs_PPSA_3>> Molecular_PolarSASA >Shadow_XYfrac.
It was found that ALogP_MR, Molecular_Volume and Jurs_PPSA_3 play the key role for the bioactivity of lipase L-A1. L-A1 tends to catalyze the hydrolysis of the esters with high ALogP_MR and Jurs_PPSA_3 values. For example, glycerol trioleate has the highest Molecular_Volume and comparatively higher Jurs_PPSA_3 values. And they counteract the negative contribution of ALogP_MR to L-A1 bioactivity, which make L-A1 possess the highest activity of 33.4U/ml.
For Eq.4, it can be found that <56.961 −Jurs_PNSA_1> and Molecular_Weight have positive contribution to the bioactivity of the lipase. However, CHI_0, Dipole_Y and <Shadow_XYfrac −0.5246> have the negative effect on the bioactivities of the lipase. The relative importance of the descriptors according to their standardized regression coefficients is in the following order:
CHI_0 > Molecular_Weight> > Dipole_Y > <Shadow_XYfrac − 0.524612> > <56.961 − Jurs_PNSA_1 > (The standardized regression coefficient for each variable is 105.19, 105.75, 0.8094, 0.7227 and 0.0081 respectively). From this equation, it was found that L-A2 tends to hydrolyze glycerides with higher values of Molecular_Weight.
For Eq.5, the standardized regression coefficient for CHI_1, Dipole_X, Jurs_FNSA_3, Jurs_FPSA_3, and Shadow_XY is 13.68, 1.174, 3.802, 4.022 and 13.91 respectively. It can be seen that the relative importance of the descriptors is as follows:
Shadow_XY> CHI_1> Jurs_FPSA_3> Jurs_FNSA_3> Dipole_X.
Thus, Shadow_XY and CHI_1 play the key roles in determining the lipase activity. Jurs_FPSA_3, Jurs_FNSA_3 and CHI_1 have the opposite contribution to the lipase activity. The dimension of the actual lipase activity value is determined by the one with higher values. For example, substrate 5 has a far higher value of Shadow_XY than that of Dipole_X, which makes L-A3 possesses comparatively higher bioactivity for it.
In order to evaluate the predictivities of these models, the four esters listed in Tables2,3 and4 were used as test set and their activities were predicted with the three models were listed in Tables2,3 and4.
Prediction for the hydrolysis activity to vegetable oils
Measured and predicted lipase activities for olive oil and soybean oil
Predicted value activity(U/ml)
Measured value activity(U/ml)
It can be seen that they have good prediction for the hydrolysis ability of three lipases. For example, the predicted values of L-A1, L-A2 and L-A3 are 25.83 U/ml, 27.86 U/ml and 26.43 U/ml which is concord well with the measured values of 27.53 U/ml, 26.52 U/ml and 27.47 U/ml respectively. This result shows that these QSAR models not only can predict the lipase activity for one fat acid ester, but they can be used to predict the lipase activity for hydrolysis the natural oils composed of mixture of different esters.
In this study, three QSAR models for lipases L-A1, L-A2 and L-A3 respectively were obtained using GFA algorithm in DS 2.1. The prediction of these QSAR model were evaluated by internal validation and external validation. The results showed that they have good prediction for the hydrolysis ability of three lipases it can also be used to predict and evaluate the hydrolytic activity to mixed oils.
This work was supported by Tianjin Natural Science Foundation (No. 09JCZDJC17800, No. 07JCYBJC07900) and the Program of Weihai Science and Technology Development (IMJQ01110034).
- Jaeger KE, Reetz MT: Microbial lipases form versatile tools for biotechnology. Trends Biotechnol. 1998, 16: 396-403. 10.1016/S0167-7799(98)01195-0View ArticlePubMedGoogle Scholar
- Fernandes MLM, Saad EB, Meira JA, Ramos LP, Mitchell DA, Krieger N: Esterification and transesterification reactions catalysed by addition of fermented solids to organic reaction media. J Mol Catal B: Enzym. 2007, 44: 8-13. 10.1016/j.molcatb.2006.08.004.View ArticleGoogle Scholar
- Laumen K, Schneider MP: A highly selective ester hydrolase from pseudomonas sp for the enzymatic preparation of enantiomerically pure secondary alcohols - chiral auxiliaries in organic-synthesis. J Chem Society-Chem Commun. 1988, 22: 598-600.View ArticleGoogle Scholar
- Tsai SW, Tsai CS, Chang CS: Lipase-catalyzed synthesis of (S)-naproxen ester prodrug by transesterification in organic solvents. Appl Biochem Biotechnol. 1999, 80: 205-219. 10.1385/ABAB:80:3:205View ArticlePubMedGoogle Scholar
- Ito S, Kobayashi T, Ara K, Ozaki K, Kawai S, Hatada Y: Alkaline detergent enzymes from alkaliphiles: enzymatic properties, genetics, and structures. Biomed Life Sci. 1998, 2: 185-190.Google Scholar
- Bora LK, Mohan C: Production of thermostable alkaline lipase on vegetable oils from a thermophilic Bacillus sp. DH4, characterization and its potential applications as detergent additive. Chem Tech Biotechnol. 2008, 83: 19-24.View ArticleGoogle Scholar
- Maugard T, Legoy MD: Enzymatic synthesis of derivatives of vitamin A in organic media. J Mol Catal B: Enzym. 2000, 8: 275-280. 10.1016/S1381-1177(99)00078-8.View ArticleGoogle Scholar
- Saxena RK, Ghosh PK, Gupta R, Davidson WS, Bradoo S, Gulati R: Microbial lipases: Potential biocatalysts for the future industry. Curr Sci. 1999, 77: 101-115.Google Scholar
- Hemachander C, Puvanakrishnan R: Lipase from Ralstonia pickettii as an additive in laundry detergent formulations. Process Biochem. 2000, 35: 809-814. 10.1016/S0032-9592(99)00140-5.View ArticleGoogle Scholar
- Dharmsthiti S, Kuhasuntisuk B: Lipase from Pseudomonas aeruginosa LP602: biochemical properties and application for wastewater treatment. J Ind Microbiol Biotechnol. 1998, 21: 75-80. 10.1038/sj.jim.2900563.View ArticleGoogle Scholar
- Vorderwulbecke T, Kieslich K: Comparison of lipases by different assays. Enzyme Microb Technol. 1992, 14: 631-639. 10.1016/0141-0229(92)90038-P.View ArticleGoogle Scholar
- Tietz NW, Shuey DF, Astles JR: Turbidimetric measurement of lipase activity problems and some solutions. Clin Chem. 1987, 33: 1624-1629.PubMedGoogle Scholar
- Ballot C, Favre-Bonvin G, Wallach J: Conductimetric assay of a bacterial lipase, using triacetin as a substrate. Biochem Eng J. 1982, 15: 119-129.Google Scholar
- Saisubramanian N, Edwinoliver NG, Nandakumar N, Kamini NR, Puvanakrishnan R: Efficacy of lipase from Aspergillus niger as an additive in detergent formulations: a statistical approach. J Ind Microbiol Biotechnol. 2006, 33: 669-676. 10.1007/s10295-006-0100-9View ArticlePubMedGoogle Scholar
- Han D, Rhee JS: Characteristics of lipase-catalyzed hydrolysis of olive oil in AOT-isooctane reversed micelles. Biotechnol Bioeng. 2004, 28: 1250-1255.View ArticleGoogle Scholar
- Grbavcic S, Bezbradica D, Izrael-Zivkovic L, Avramovic N, Milosavic N, Karadzic I, Knezevic-Jugovic Z: Production of lipase and protease from an indigenous Pseudomonas aeruginosa strain and their evaluation as detergent additives: Compatibility study with detergent ingredients and washing performance. Bioresour Technol. 2011, 102: 11226-11233. 10.1016/j.biortech.2011.09.076View ArticlePubMedGoogle Scholar
- Hæffner F, Norin T, Hult K: Molecular Modeling of the Enantioselectivity in Lipase Catalyzed Transesterification Reactions. Biophys J. 1998, 74: 1251-1262. 10.1016/S0006-3495(98)77839-7PubMed CentralView ArticlePubMedGoogle Scholar
- Tendulkar AV, Wangikar PP, Sohoni MA, Samant VV, Mone CY: Parameterization and classification of the protein universe via geometric techniques. J Mol Biol. 2003, 334: 157-172. 10.1016/j.jmb.2003.09.021View ArticlePubMedGoogle Scholar
- Wangikar PP, Tendulkar AV, Ramya S, Mail DN, Sarawagi S: Functional sites in protein families uncovered via an objective and automated graph theoretic approach. J Mol Biol. 2003, 326: 955-978. 10.1016/S0022-2836(02)01384-0View ArticlePubMedGoogle Scholar
- Berti F, Forzato C, Nitti P, Pitacco G, Valentin E: A study of the enantio preference of lipase PS (Pseudomonas cepacia) towards diastereomeric dihydro-5-alkyl-4-hydroxymethyl-2(3 H)-furanones. Tetrahedron: Asymmetry. 2005, 16. 10.1-1102.View ArticleGoogle Scholar
- Schrag JD, Cygler M: A refined structure of the lipase from Geotrichum candidum. J Mol Biol. 1993, 230: 575-591. 10.1006/jmbi.1993.1171View ArticlePubMedGoogle Scholar
- Botta M, Cernia E, Corelli F, Manetti F, Soro S: Probing the substrate specificity for lipases. A CoMFA approach for predicting the hydrolysis rates of 2-arylpropionic esters catalyzed by Candida rugosa lipase. Biochim Biophys Acta. 1996, 1296: 121-126. 10.1016/0167-4838(96)00064-7View ArticlePubMedGoogle Scholar
- Tomic S, Kojic-Prodic B: A quantitative model for predicting enzyme enantioselectivity: application to Burkholderia cepacia lipase and 3-(aryloxy)-1, 2-propanediol derivatives. J Mol Graphics Modell. 2002, 21: 241-252. 10.1016/S1093-3263(02)00148-1.View ArticleGoogle Scholar
- Wang H, Liu R, Lu F, Qi W, Shao J, Ma H: A novel alkaline and low-temperature lipase of Burkholderia cepacia isolated from Bohai in China for detergent formulation. Ann Microbiol. 2009, 59: 105-110. 10.1007/BF03175606.View ArticleGoogle Scholar
- Wang H, Zhang J, Wang X, Qi W, Dai Y: Genome shuffling improves production of the low-temperature alkalophilic lipase by Acinetobacter johnsonii. Biotechnol Lett. 2012, 34: 145-151. 10.1007/s10529-011-0749-7View ArticlePubMedGoogle Scholar
- Wang H, Shao J, Wei YJ, Zhang J, Qi W: A Novel Low-Temperature Alkaline Lipase from Acinetobacter johnsonii LP28 Suitable for Detergent Formulation. Food Technol Biotechnol. 2011, 49: 96-102.Google Scholar
- Wang H, Zhong S, Ma H, Zhang J, Qi W: Screening and characterization of a novel alkaline lipase from acinetobacter calcoaceticus 1–7 isolated from bohai bay in china for detergent formulation. Braz J Microbiol. 2012, 43: 148-156. 10.1590/S1517-83822012000100016.PubMed CentralView ArticlePubMedGoogle Scholar
- Nahas E: Control of lipase production by Rhizopus oligosporus under various growth conditions. J Gen Microbiol. 1988, 4: 227-233.Google Scholar
- CambridgeSoft Inc: ChemBioOffice Ultra Version 11.0. 2008, Cambridge, USA.Google Scholar
- Schmidt MW, Baldridge KK, Boatz JA, Elbert ST, Gordon MS, Jensen JH, Koseki S, Matsunaga N, Nguyen KA, Su S: General atomic and molecular electronic structure system. J Comput Chem. 1993, 14: 1347-1363. 10.1002/jcc.540141112.View ArticleGoogle Scholar
- Accelrys Inc: Discovery Studio 2.1. 2010, San Diego, CA, USA.Google Scholar
- Leonard JT, Roy K: On selection of training and test sets for the development of predictive QSAR models. Qsar & Comb Sci. 2006, 25: 235-251. 10.1002/qsar.200510161View ArticleGoogle Scholar
- Roy K, Mandal AS: Development of linear and nonlinear predictive QSAR models and their external validation using molecular similarity principle for anti-HIV indolyl aryl sulfones. J Enzym Inhib Med Chem. 2008, 23: 980-995. 10.1080/14756360701811379.View ArticleGoogle Scholar
- Fan Y, Shi LM, Kohn KW, Pommier Y, Weinstein JN: Quantitative structure-antitumor activity relationships of camptothecin analogues: Cluster analysis and genetic algorithm-based studies. J Med Chem. 2001, 44: 3254-3263. 10.1021/jm0005151View ArticlePubMedGoogle Scholar
- Rogers D, Hopfinger AJ: Application of genetic function approximation to quantitative structure -activity relationship and quantitative structure -property relationship. J Chem Inf Comput Sci. 1994, 34: 854-866.View ArticleGoogle Scholar
- Friedman JH: Multivariate adaptive regression splines (with discussion). Ann Stat. 1991, 19: 1-141. 10.1214/aos/1176347963.View ArticleGoogle Scholar
- Dunn WJ, Hopfinger AJ, Catana C, Duraiswami C: Solution of the conformation and alignment tensors for the binding of trimethoprim and its analogs to dihydrofolate reductase: 3D-quantitative structure-activity relationship study using molecular shape analysis, 3-way partial least-squares regression, and 3-way factor analysis. J Med Chem. 1996, 39: 4825-4832. 10.1021/jm960491rView ArticlePubMedGoogle Scholar
- Zelles L, Bai QY: Fractionation of fatty acids derived from soil lipids by solid phase extraction and their quantitative analysis by GC-MS. Soil Biol Biochem. 1993, 25: 495-507. 10.1016/0038-0717(93)90075-M.View ArticleGoogle Scholar
- Lee DS, Noh BS, Bae SY, Kim K: Characterization of fatty acids composition in vegetable oils by gas chromatography and chemometrics. Anal Chim Acta. 1998, 358: 163-175. 10.1016/S0003-2670(97)00574-6.View ArticleGoogle Scholar
- Greenland S, Maclure M, Schlesselman JJ, Poole C, Morgenstern H: Standardized regression coefficients: a further critique and review of some alternatives. Epidemiol (Cambridge, Mass). 1991, 2: 387-392. 10.1097/00001648-199109000-00015.View ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.