Research Article - (2020) Volume 12, Issue 5
Kamrani SME1 and Hadizadeh F2*
1Department of Medicinal Chemistry, School of Pharmacy, Mashhad University of Medical Sciences, Mashhad, Iran
2Biotechnology Research Center, Pharmaceutical Technology Institute, Mashhad University of Medical Sciences, Mashhad, Iran
Received date: August 09, 2020, Accepted date: September 16, 2020, Published date: September 23, 2020
Citation: Kamrani SME, Hadizadeh F (2020) An ANN-Based QSAR Model to Predict Anti-Staphylococcus aureus Activity of Oxadiazoles. Arch Med Vol.12 No.5: 30. DOI: 10.36648/1989-5216.12.5.331.
Background: Antibiotic-resistant bacteria are likely to be one of the most critical problems in the future; hence, more effort is needed to design and develop new types of antimicrobial agents. Quantitative Structure-Activity Relationship (QSAR) is a procedure which helps other researchers to design better chemical agents with more potent biological activity. QSAR can be defined as a quantitative relationship between the chemical structure and biological activity.
Results: In this study, a previously-synthesized oxadiazole library was used to build a QSAR model based on the Group Method of Data Handling (GMDH) method and Partial Least Squares (PLS) regression. Owing to their high correlation coefficients (R2) for test and training data, both methods are sufficiently reliable. In this study, the active compounds of the library were used as a template to design new chemical compounds predicted to have a great anti-Staphylococcus aureus (S. aureus) activity according to PLS, GMDH, and docking methods. GMDH and PLS are highly flexible such that they can include other information like absorption, distribution, metabolism, and excretion (ADMT) and toxicity.
Conclusion: The selected methods can be used to handle huge amounts of data from a large library of chemical compounds and help research and development (R&D) process. Additionally, the designed model and the proposed compounds can help other researchers to find the best anti- Staphylococcus aureus chemical compounds.
Oxadiazole; Quantitative structure-activity relationship; Staphylococcus aureus
R&D: Research and Development; GMDH: Group Method of Data Handling; PLS: Partial Least Square; QSAR: Quantitative Structure Activity Relationship; ADME: Absorption, Distribution, Metabolism and Excretion; MRSA: Methicillin-Resistant Staphylococcus aureus; Pmic: -Log: Minimum Inhibitory Concentration
Antibiotic-resistant bacteria, whether gram-positive or gramnegative, are one of the most important health problems which are likely to turn into a critical threat in the future. This problem arises from the indiscriminate use of antibiotics. One of these resistant strains of bacteria is Methicillin-Resistant S. aureus (MRSA) [1]. Regarding the wide range of highly resistant strains observed in hospitals and communities, it is necessary to study and design new antibiotics to overcome the high rate of infectious diseases and their adverse consequences for the societies. One of these classes of chemical structures is 1,2,4-oxadiazole which has been extensively synthesized and investigated [2].
Developing Quantitative Structure-Activity Relationship (QSAR) models that delineate the relationships between the chemical descriptors and biological activities of the chemical compounds may have considerable advantages. These models can be developed based on artificial neural networks (e.g. MLP and RBF) or regressions (e.g. PLS, PCR, and MLR).
Chang and Mobashery synthesized a series of 1,2,4-oxadiazole and measured the plC50 of these compounds [3,4]. Leemans [5] used a library of 102 members of 1,2,4-oxadiazole to build a 3D-QSAR model with good predictive power (correlation coefficient for the training and test data sets were 0.88 and 0.61, respectively) [5] (Figure 1). In the present study, Partial Least Squares (PLS) and Group Methods of Data Handling (GMDH) were used to develop a robust QSAR model for predicting the Pmic of oxadiazoles with high correlation coefficients. To this end, data mining was conducted to find appropriate descriptors having a better correlation with Pmic.
Ivahnenko (1971) presented a predictive model for identifying a nonlinear relationship between inputs and outputs (Figure 2). Like other neural networks, the GMDH is also capable of predicting outcomes from the multivariate data and achieving the maximum accuracy and reliability by testing all possible structures of the polynomial regression models [6,7].
image
Partial Least Squares (PLS) regression is a popular method for soft modelling in QSAR. PLS is a method for constructing predictive models when there are multiple highly collinear factors. Today, PLS regression is one of the most widely used techniques in chemo metrics and related areas. It is also used in bio-informatics, sensometrics, neuroscience, and anthropology [8].
In this study, the Leemans' library was used to generate PLS- and GMDH-based QSAR model with a good correlation coefficient (R2). These methods, which are defined as Matlab codes (.m), can be used by other drug designers to produce potent and nontoxic oxadiazoles with good pharmacokinetic characteristics, low toxicity, and few side effects.