Abstract
Reliance on deep learning techniques has become an important trend in several science domains including biological science, due to its proven efficiency in manipulating big data that are often characterized by their non-linear processes and complicated relationships. In this study, Convolutional Neural Networks (CNN) has been recruited, as one of the deep learning techniques, to be used in classifying and predicting the biological activities of the essential oil-producing plant/s through their chemical compositions. The model is established based on the available chemical composition’s information of a set of endemic Egyptian plants and their biological activities. Another type of machine learning algorithms, Multiclass Neural Network (MNN), has been applied on the same Essential Oils (EO) dataset. This aims to fairly evaluate the performance of the proposed CNN model. The recorded accuracy in the testing process for both CNN and MNN is 98.13% and 81.88%, respectively. Finally, the CNN technique has been adopted as a reliable model for classifying and predicting the bioactivities of the Egyptian EO-containing plants. The overall accuracy for the final prediction process is reported as approximately 97%. Hereby, the proposed deep learning model could be utilized as an efficient model in predicting the bioactivities of, at least Egyptian, EOs-producing plants.
Subject terms: Biochemistry, Computational biology and bioinformatics, Plant sciences
Introduction
Recently, Artificial Intelligence (AI) has become one of the vigorous science that infiltrated a huge number of modern life issues such as chemical engineering, water treatment, and biological domain like genomic and proteomic studies which are especially characterized by complicated and non-linear processes1. Deep learning is one of the most promising branches of artificial intelligence with proven power in taking the raw features extracted from the extremely large data sets, such as the data produced from genomics, chemistry, and pharmaceutical laboratories. Processing these data result in inferred patterns and training process-based predictive models2.
Essential Oils (EOs) are biologically effective organic compounds extracted from different parts of the aromatic plants such as flowers, leaves, and barks to name a few3. Due to their wide range of biological activities, these natural products are widely used in complementary and alternative medicine (CAM). Replacing the inorganic chemistry by natural alternatives is still hot topic in recent biological area of research. This is because the inorganic chemical products may bereave harmful influences when used in health-related industries, such as medicine, pharmaceutics, cosmetics, food, and beverages. Therefore, the modern researches head to find the alternative natural products, including EOs, due to their greater ability to adapt to alive organs of the human body with, sometimes, limited side effects4.
Generally, the bioactivity of the EO-producing plant depends on the chemical structure and EOs content, which determine the overall bioactivity of such plant. EOs are, in fact, composed of different combinations of low nuclear weight natural blends with complete organic locomotion. According to their structure, these dynamic blends can be categorized into some significant pools (e.g. hydrocarbons, oxygenated mixtures, and sulfur or conceivably nitrogen). These mixes’ pools are the mystery key for the biological activity of each EO3.
The biological activities of the EOs may include antiseptics, antimicrobials, antifungals, antioxidant, antitumor, antivirals, and/or anti-inflammatories. Moreover, these activities vary according to the chemical constitution, which may differ from plant to another according to their geographic location, agriculture conditions, climatic or seasonal changes4. Noteworthy, the evaluation of the EOs' bioactivities cannot be constantly credited to one single compound in the EOs mixture. The genuine connections between the EO’s cocktail and its biological activities are highly non-linear, especially when considered across variable pools of chemical structures. Hence, reliance on traditional techniques in predicting the biological impact for such activity data with this variety of structures is a troublesome issue5. Therefore, developing a deep learning-based computational model to categorize and predict the biological activities of EOs-producing plants based on their chemical construction’s variations, without recourse to in-vitro experiments, could save time and cost.
Machine learning (ML) algorithms, especially Artificial Neural Networks (ANN), have been proposed to contribute in solving several biological issues in the recent decades6. ANN, in general, can be depicted as a numerical model of a particular structure, comprising of some of the single processing components (i.e. nodes and neurons), constructed between inter-connected layers. Each entire layer is mainly composed of hidden neurons which are responsible for transforming the input values and sending the outputs to the other associated neurons1. Recently, due to the expansion of the biological information, the fully connected neural network would have a huge number of parameters, which needs full processing inside the network layers to deliver the desired output. Deep learning approaches have proven their efficiency in the applications whose data are characterized by their large quantities, high dimensionality, and highly structured. Thus, deep learning approaches are widely used in image processing due to the nature of the image which contains many thousands of variables (pixels) that can be clearly grouped into well- defined objects6. However, deep learning approaches are no longer limited to image processing domain, where it is recently considered an attractive solution for some types of text classification such as DNA sequences classification problems7. From this standpoint, deep learning can be an efficient learning approach for dealing with the complex composition of the chemical compounds and their interrelationships with biological activities.
One of the efficient models for the deep learning is the Convolutional Neural Network (CNN). The CNN is characterized by two novel types of layers: convolution and pooling layers. These layers are based on using filters to convolve the range of the input data to a smaller range, detecting important or specific parts within this range8. The CNN usually consists of Input Layer , Convolution Layer (i.e. produces a matrix of dimension smaller than the input matrix), ReLU or Rectified Linear Unit . ReLU is mathematically expressed as Max(0, x) (i.e. it means that any number below 0 is converted to 0, while any positive number is allowed to pass as it is), Max pooling (i.e. passes the maximum value from amongst a small collection of elements of the incoming matrix to the output) and the final output layer (i.e. a fully-connected neural network layer, which makes the output based on the activation function), as shown in Fig. 19.
This study seeks to classify and predict the biological activities of the Egyptian essential oil-producing plants based on their EOs content as an experimental case study. The classification is implemented based on two algorithms: Multiclass Neural Network (MNN) and Convolutional Neural Network (CNN), to evaluate the efficiency of both machine and deep learning techniques. The effective algorithm is adopted in developing a biological activity prediction model EOs-producing cases cultivated in Egypt. The research skills in this study are organized as follow; first, presenting the results obtained from implementing the two algorithms; MNN as a machine learning algorithm and CNN as a deep learning algorithm, when applied on the Egyptian essential oils dataset. Next, these results were discussed and evaluated for the existing dataset and the new untested datasets. Finally the methodologies followed in this work are discussed.
Results
Using MNN- and CNN-based algorithms to classify the plants’ EOs bioactivity
The results recorded from the classification process for both MNN and CNN algorithms are summarized in Tables 1 and 2. In the CNN and MNN algorithms, the training processes show an overall accuracy of 100% and 99.2%, of correct classification of essential oils activity, respectively. Whilst, for the testing stage, the overall accuracy achieved by CNN and MNN is 98.13% and 81.88%, respectively.
Table 1.
Bioactivity class Classifiers |
True positive | True negative | False positive | False negative | ||||
---|---|---|---|---|---|---|---|---|
MNN | CNN | MNN | CNN | MNN | CNN | MNN | CNN | |
Antiviral | 50% | 97.7% | 94.4% | 98.2% | 5.6% | 1.8% | 50% | 2.3% |
Antiwormal | 100% | 100% | 98.3% | 98.5% | 1.7% | 1.5% | 0 | 0% |
Anti-inflammatory | 90% | 100% | 80% | 100% | 20% | 0% | 10% | 0% |
Anticancer | 85.7% | 94% | 96.2% | 93.3% | 3.8% | 6.7% | 14.3% | 6% |
Antioxidant | 75.9% | 98.4% | 80.6% | 98.6% | 19.4% | 1.4% | 24.1% | 1.6% |
Antimicrobial | 78.8% | 97.4% | 66.7% | 97.3% | 33.3% | 2.7% | 21.2% | 2.6% |
Antifungal | 81.3% | 98% | 86.4% | 99.2% | 13.6% | 0.8% | 18.8% | 2% |
Cytotoxic activity | 58.3% | 96.6% | 93.8% | 98.7% | 6.3% | 1.3% | 41.7% | 3.4% |
Table 2.
Bioactivity class Metrics |
Average accuracy | Precision | Recall | F1 score | ||||
---|---|---|---|---|---|---|---|---|
MNN | CNN | MNN | CNN | MNN | CNN | MNN | CNN | |
Antiviral | 0.72 | 0.99 | 0.9 | 0.98 | 0.5 | 0.99 | 0.64 | 0.99 |
Antiwormal | 0.99 | 0.99 | 0.98 | 0.99 | 1 | 1 | 0.99 | 0.99 |
Anti-inflammatory | 0.85 | 1 | 0.82 | 1 | 0.9 | 1 | 0.86 | 1 |
Anticancer | 0.9 | 0.94 | 0.96 | 0.93 | 0.86 | 0.94 | 0.9 | 0.94 |
Antioxidant | 0.78 | 0.99 | 0.8 | 0.99 | 0.76 | 0.98 | 0.78 | 0.98 |
Antimicrobial | 0.73 | 0.97 | 0.7 | 0.97 | 0.79 | 0.97 | 0.74 | 0.97 |
Antifungal | 0.84 | 0.99 | 0.86 | 0.99 | 0.812 | 0.98 | 0.83 | 0.99 |
Cytotoxic activity | 0.76 | 0.98 | 0.9 | 0.99 | 0.58 | 0.97 | 0.71 | 0.98 |
The confusion matrix is the most suitable way to validate the classification performance. Here, the confusion matrix of the MNN and CNN classification is presented in Table 1. It consists of four outcomes of binary classifiers: True Positive, False Positive, True Negative, and False Negative. Also, accuracy, precision, recall, and F1 score are different metrics that are used for evaluating the classification efficiency based on the values of the confusion matrix10. The four metrics for the bioactivities classes have been calculated and documented in Table 2.
Using a CNN-based algorithms to build a prediction model for the EOs bioactivity
In accordance with the completion of the classification process, the CNN algorithm show high accuracy for the training and testing processes. This is due to its capabilities in dealing with the huge number of data and focusing on the high impact features in the dataset. Thus, the biological activity prediction model has been built based on the CNN proposed algorithm. The overall accuracy for predicting the biological activities for previously unknown Egyptian essential oils components has been recorded as approximately 97%. The concluded metrics values and the discrimination outcomes of the proposed CNN prediction model are reported in Tables 3 and 4, and in Fig. 2.
Table 3.
Bioactivity class Classifiers |
True positive | True negative | False positive | False negative |
---|---|---|---|---|
Antiviral | 98% | 96.3% | 3.7% | 2% |
Antiwormal | 100% | 98.3% | 1.7% | 0% |
Anti-inflammatory | 100% | 100% | 0% | 0% |
Anticancer | 97% | 94.7% | 5.3% | 3% |
Antioxidant | 100% | 95% | 5% | 0% |
Antimicrobial | 94% | 93.5% | 6.5% | 6% |
Antifungal | 100% | 96.6% | 3.4% | 0% |
Cytotoxic activity | 94% | 93.1% | 6.9% | 6% |
Table 4.
Bioactivity class Metrics |
Average accuracy | Precision | Recall | F1 score |
---|---|---|---|---|
Antiviral | 0.97 | 0.96 | 0.98 | 0.97 |
Antiwormal | 0.99 | 0.98 | 1 | 0.99 |
Anti-inflammatory | 1 | 1 | 1 | 1 |
Anticancer | 0.96 | 0.95 | 0.97 | 0.96 |
Antioxidant | 0.98 | 0.95 | 1 | 0.98 |
Antimicrobial | 0.94 | 0.94 | 0.94 | 0.94 |
Antifungal | 0.98 | 0.97 | 1 | 0.98 |
Cytotoxic activity | 0.94 | 0.93 | 0.94 | 0.94 |
Discussion
Adopting the idea of using AI, especially machine learning and deep learning algorithms, has become a vital topic in medical and biological problem solving. EOs are considered as one of the well-known natural products which have medical-relevant use and biological-defense activities against several types of viruses, bacteria, and cancer11. However, the lack of the complete information about their intrinsic chemical variability and their function makes it difficult to confirm its consistent activities4.
Many researchers have adopted the AI in EO's researches aiming to reduce the in-vitro procedures and to make reasonable predications of the experiment results. For instance, Ragno et al. (2020) have developed an unsupervised machine learning algorithm to cluster the EOs and to identify the EOs that have strong ability in inhibiting bacterial growth of all bacterial strains12. Moreover, Artini et al. (2018) have exhibited a binary classification model based on ML to classify the essential oils activities from different Mediterranean plants against pseudomonas aeruginosa13. Similarly, Daynac et al. (2015) have used the fast artificial neural networks (FNN) to predict the antimicrobial activity of 49 EOs against four types of pathogens. The FNN algorithm predicted more than 70% of the antimicrobial activities within a 10 mm maximum error range4.
Egypt is one of the biggest countries in the world in exporting the high-quality raw material of more than 150 medicinal and aromatic plants. The variety and massiveness of aromatic plant species in Egypt, stemming from the climatic conditions in its environment, may stimulate the accumulation of high concentrated secondary metabolites. This makes the Egyptian aromatic plants are considered between the most promising sources for many biologically active compounds14. Thus, there is an imperious need to extend the scientific knowledge base of the aromatic species in Egypt by the modern AI methodologies using machine and deep learning.
The experimental study, here, is applied on the Egyptian case for the essential oils. The data were collected from several peer researches article including in-vitro experiments. For each case, the in-vitro experiment may only focus in analyzing the chemical composition of the EO, or apply one or more of EOs on a specific infection type (e.g. bacteria, viruses, or cancer). This encouraged the authors, here, to experiment the machine and deep learning algorithms in predicting the biological activities for some of the Egyptian essential oils-producing plants. The training and testing processes are conducted on a novel dataset of EOs-producing plants from Egypt, that are collected and manipulated by authors from peer reviewed scientific researches11–55. The resulting dataset consists of a sample of one hundred and twenty (120) plants. The total chemical compounds extracted from this dataset are 573 compounds. In the proposed experimental case, the biological activities of the essential oils are classified according to eight categories of bio-activities (anticancer, antioxidant, antimicrobial, antifungal, antiviral, anti-wormal, anti-inflammatory, and cytotoxic activities).
In order to classify the EO's according to their biological activities, two classification models rely on the supervised learning are constructed; the first one is based on the Multiclass Neural Network (MNN) (i.e. the Multiclass N.N module in Azure). The second model depends on the Convolutional Neural Network (CNN) and it has been implemented by Python and executed as a module on the open source Azure Machine learning studio56. Figure 3 presents an inclusive flowchart for the proposed methodology stages. The two proposed supervised learning algorithms are applied on nearly 68,760 values that represent the percentage of chemical compounds concentrations in the EO’s dataset on study. The training process for both MNN and CNN algorithms is done on 60% of the dataset, whilst the remaining 40% is divided between the testing and the prediction processes. That means the actual size for the training, testing, and validation processes are 45,840, 11,460, and 11,460 values, respectively. Figure 4 shows a sample for the EOs dataset stored in azure format.
In the proposed MNN algorithm, the training model runs a sequence of binary classifiers and trains each to decide a separate classification outcome according to the softmax activation function results. The architecture of the proposed MNN is a fully connected layers network with one hidden layer contains 50 nodes, and an output layer with 8 nodes each one represents an output class. The outputs of the hidden layer and the output layer are calculated by Eq. 1 and 3 respectively57.
1 |
where
2 |
3 |
where
4 |
In general, MNN proves its efficiency when the total amount of data is limited. Hereby, more significantly, another efficient classification model based on the CNN is created. The proposed CNN is developed in such a way to handle the ambiguity and inconsistency that appeared in the chemical compositions values which could not be fully treated with the MNN. One of the mysterious characteristics in the essential oil(s) is that, they may contain some chemical compounds that have no biological influence activity, and this may be due to the inconsistency among these compounds. Thus, a number of compounds/oils that have no effect in a certain EO’s pool may be found, but they may show a clear influence if they are found with other compounds in another EO pool58.
The problem in this study is categorized as multi-label classification problem, where the essential oils can have multiple activities (i.e. outputs) at the same time. Thus, in regards with the proposed dataset, the CNN could produce eight output labels for each essential oil (i.e. the eight biological activities), where an essential oil may have all of these activities or some of them. The output labels have been encoded in the form of a one-hot encoded vector with multiple ones in it, as a special form of the one-hot encoding method. For instance, the Essential oil “Pluchea dioscoridis” is known with its activities as antimicrobial, antioxidant, and anticancer, so its label will be [0,0,0,1,1,1,0,0] for the target vector [Antiviral, Antiwormal, Anti-inflammatory, Anticancer, Antioxidant, Antimicrobial, Antifungal, and Cytotoxic Activity]. Table 5 shows a sample of the CNN training results on the dataset where it documents the score probabilities that outcome from the sigmoid activation function. These score values refer to predicted biological activities labels for a number of Essential oils.
Table 5.
Eos name Score probabilities |
Antiviral | Antiwormal | Anti-inflammatory | Anticancer | Antioxidant | Antimicrobial | Antifungal | Cytotoxic activity |
---|---|---|---|---|---|---|---|---|
Citrus deliciosa var. tangarina—fina clementine—peels | 0.9888 | 0.0014 | 0.0032 | 0.0001 | 0.9854 | 0.9981 | 0.9983 | 0.9966 |
Citrus deliciosa var. tangarina—fina clementine—leaves | 0.9999 | 0.9986 | 0.0005 | 0.0003 | 0.9992 | 0.9984 | 0.9983 | 0.0013 |
Citrus deliciosa var. tangarina—Nour Clementine—peels | 0.9465 | 0.0022 | 0.3243 | 0.0002 | 0.9894 | 0.9854 | 0.9684 | 0.7711 |
Citrus deliciosa var. tangarina—Nour Clementine—leaves | 0.9999 | 0.0013 | 0.0641 | 0.0002 | 0.9995 | 0.9997 | 0.9996 | 0.0003 |
Citrus deliciosa var. tangarina—Spinosa Clementine—peels | 0.8294 | 0.0795 | 0.0001 | 0.0003 | 0.8291 | 0.9294 | 0.9993 | 0.6654 |
Citrus deliciosa var. tangarina—Spinosa Clementine—leaves | 0.9977 | 0.0015 | 0.0320 | 0.0084 | 0.9968 | 0.9982 | 0.9961 | 0.0049 |
Citrus deliciosa var. tangarina—Thornless Clementine—peels | 0.9991 | 0.0007 | 0.0931 | 0.0034 | 0.9987 | 0.9999 | 0.998 | 0.2611 |
Citrus deliciosa var. tangarina—Thornless Clementine—leaves | 0.9999 | 0.0020 | 0.0134 | 0.0082 | 0.9999 | 0.9998 | 0.9996 | 0.0007 |
Callistemon comboynensis | 0.9975 | 0.0037 | 0.1375 | 0.0011 | 0.9947 | 0.9955 | 0.9945 | 0.0031 |
Cupressus sempervirens L Semperuirens L |
0.8982 | 0.0011 | 0.0004 | 0.0012 | 0.9947 | 0.9967 | 0.0002 | 0.0035 |
Cuminum cyminum | 0.8923 | 0.9747 | 0.9384 | 0.0005 | 0.9297 | 0.9997 | 0.7909 | 0.0001 |
Ocimum basilicum L. (sinai ) | 0.0004 | 0.0005 | 0.0883 | 0.9977 | 0.9991 | 0.9990 | 0.9089 | 0.0003 |
Tagetes minuta L | 0.4107 | 0.0011 | 0.0034 | 0.0253 | 0.5146 | 0.7167 | 0.6179 | 0.1418 |
Achillea fragrantissima | 0.0004 | 0.0003 | 0.0003 | 0.0007 | 0.0008 | 0.0013 | 0.0001 | 0.9985 |
Pluchea dioscoridis | 0.0052 | 0.7622 | 0.0054 | 0.9863 | 0.9931 | 0.9991 | 0.8932 | 0.9915 |
Myrtus communis—leaves | 0.0022 | 0.9887 | 0.0554 | 0.0018 | 0.0003 | 0.0021 | 0.0023 | 0.0028 |
Myrtus communis—fruits | 0.0007 | 0.6754 | 0.0001 | 0.0005 | 0.0013 | 0.0023 | 0.0011 | 0.0004 |
Eugenia supraxillaris—leaves | 0.001 | 0.9998 | 0.3451 | 0.0075 | 0.0050 | 0.0075 | 0.0065 | 0.0032 |
Eugenia supraxillaris—fruits | 0.9955 | 0.7865 | 0.0003 | 0.0003 | 0.9995 | 0.9934 | 0.0273 | 0.8539 |
Conclusion
Developing a computational model based on deep learning for classifying and predicting the EOs' biological activities without resorting to in-vitro experiments is the challenge of this study.
Due to the efficacy of the EOs as antiseptics, antimicrobials, antifungals, antioxidant, antitumor, antivirals, and anti-inflammatories, they have a great attention from the health concerns industries, such as medicine, pharmaceutics, cosmetics, and others. However, there is a significant challenge in deciding the relevance between the chemical compounds that form the EO and its biological activities through the traditional in-vitro experiments. In this study, two classification models are implemented to classify and predict the biological activities of 120 types of Egyptian essential oils as an experimental study case. This experiment has been implemented based on two types of supervised learning algorithms, Multiclass Neural Network and Convolutional Neural Network in order to evaluate the efficiency of both machine and deep learning techniques. The comparison between the accuracy and relevance metrics for both MNN and CNN algorithms in the testing stage showed that the CNN outperformed the MNN as it scored an accuracy rate 98.13%, while the MNN recorded 81.88%.
Method
The model of the CNN, which is utilized in this experiment, comprises of a three-layers fully connected network with two convolution layers and two pooling layers followed by one hidden layer as shown in Fig. 5. The input information for the proposed CNN is a 2D matrix of [120 × 573] that represents “the number of essential oils and the numbers of their chemical compounds, respectively”. The following sequential layers start with a convolution layer which applies a convolutional process to the input matrix. The output of each node in the convolution layer is the result of a convolution operation by each filter. The next layer is the pooling layer which is used to compress information and generalize features to reduce the overfitting of the training data. In this experiment, the local max-pooling, which produces the maximum value from small divided regions in the input matrix, is used. The convolution and pooling outputs are calculated by the following equation;
5 |
where is the output vector of the lth layer with kth kernel, is the input vector, is the weight of the convolution/pooling filter, and is the bias coefficient.
During the CNN learning process, seven hyper-parameters are tuned (convolution filter size (k), number of filters (f), stride size (s), pooling size (p), number of nodes in the hidden layer, and the activation function). The numbers of iterations for the learning process are specified around 100. Table 6 displays the values of the assumed hyper-parameters and the output layer dimensions. The individual layer dimension can be calculated by8;
6 |
Table 6.
Layers | Filter size | Stride | Output size | Output Dense |
---|---|---|---|---|
Input | – | – | 120 × 573 | 68,760 |
1st convolution layer | 120 × 5 × 100 | 1 | 1 × 569 × 100 | 56,900 |
1st max pooling layer | 1 × 2 × 100 | 2 | 1 × 285 × 100 | 28,500 |
2nd convolution layer | 1 × 5 × 100 | 2 | 1 × 141 × 100 | 14,100 |
2nd max pooling layer | 1 × 2 × 100 | 2 | 1 × 71 × 100 | 7100 |
Finally, the sequence of the convolutional and pooling layers ends with a fully connected feed-forward neural layer that uses a “sigmoid” activation function. The sigmoid function is selected in the proposed CNN implementation because its function depends mainly on converting each score of the final node to a probability value between 0 to 1, independent of what the other scores are. So, the input could be classified into multiple independent classes (Suppl Information).
Supplementary Information
Acknowledgements
This project was supported financially by the Academy of Scientific Research and Technology (ASRT), Egypt, Grant No. 6474.
Author contributions
N.E.E. and W.A.A. designed and carried out the experiments and developed the machine and deep learning algorithms. M.K.H. and O.A. revised and adopted the dataset. All authors contributed in writing, editing, and revising the manuscript.
Data availability
The complete data set is available in the supplementary file (sup-file). The dataset is formatted to be suitable to be processed by Azure ML modules.
Footnotes
Publisher's note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
is available for this paper at 10.1038/s41598-020-78449-1.
References
- 1.Muniz I, et al. Classification of food vegetable oils by fluorimetry and artificial neural networks. Food Control. 2015;47:86–91. doi: 10.1016/j.foodcont.2014.06.030. [DOI] [Google Scholar]
- 2.Littmann M, et al. Validity of machine learning in biology and medicine increased through collaborations across fields of expertise. Mach. Intell. Nat. 2016 doi: 10.1038/s42256-019-0139-8. [DOI] [Google Scholar]
- 3.Dhifi, W., Bellili, S., Sabrine Jazi, N. B. W. M. Essential oils ’ chemical characterization and investigation of some biological activities . Medicines3, 1–16 (2020). [DOI] [PMC free article] [PubMed]
- 4.Arjmand, Z. & Dastan, D. Chemical characterization and biological activity of essential oils from the aerial part and root of Ferula haussknechtii. Flavour Fragr. 114–123. 10.1002/ffj.3544 (2020).
- 5.Harper G, Bradshaw J, Gittins JC, Green DVS, Leach AR. Prediction of biological activity for high-throughput screening using binary kernel discrimination. J. Chem. Inf. Comput. Sci. 2001;41:1295–1300. doi: 10.1021/ci000397q. [DOI] [PubMed] [Google Scholar]
- 6.Jones DT. Setting the standards for machine learning in biology. Nat. Rev. Mol. Cell Biol. 2019;20:659–660. doi: 10.1038/s41580-019-0176-5. [DOI] [PubMed] [Google Scholar]
- 7.Aoki G, Sakakibara Y. Convolutional neural networks for classification of alignments of non-coding RNA sequences. Bioinformatics. 2018 doi: 10.1093/bioinformatics/bty228. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Dhillon A, Verma GK. Convolutional neural network: A review of models, methodologies and applications to object detection. Prog. Artif. Intell. 2019 doi: 10.1007/s13748-019-00203-0. [DOI] [Google Scholar]
- 9.Gao RX, et al. Deep learning for smart manufacturing : Methods and applications Deep learning for smart manufacturing : Methods and applications. J. Manuf. Syst. 2018;48:144–156. doi: 10.1016/j.jmsy.2018.01.003. [DOI] [Google Scholar]
- 10.Tharwat, A. Applied computing and informatics classification assessment methods. Appl. Comput. Inform. 1–13. 10.1016/j.aci.2018.08.003 (2018).
- 11.Elansary, H. O., Abdelgaleil, S. A. M., Mahmoud, E. A., Yessoufou, K. & Elhindi, K. Effective antioxidant , antimicrobial and anticancer activities of essential oils of horticultural aromatic crops in northern Egypt. 1–10 (2018). [DOI] [PMC free article] [PubMed]
- 12.Ragno, R. et al. Essential oils against bacterial isolates from cystic fibrosis patients by means of antimicrobial and unsupervised machine learning approaches. Sci. Reports Nat. Res.10, 1–11 (2020). [DOI] [PMC free article] [PubMed]
- 13.Artini, M., Vrenna, G. & Pepi, F. Antimicrobial and antibiofilm activity and machine learning classification analysis of essential oils. Molecules23 (2018). [DOI] [PMC free article] [PubMed]
- 14.Fayez S, Abouziena HF, El-razik TMA, Amer HM, Hussein MS. Winter weeds and its control in the medicinal plants in Egypt : A survey study. Egypt. Pharm. J. 2019;18:16–26. doi: 10.4103/epj.epj_13_18. [DOI] [Google Scholar]
- 15.El-askary HI, et al. Chemical composition of the essential oil and botanical study of the flowers of Mentha suaveolens chemical composition of the essential oil and botanical study. Inf. Healthc. 2014;52:688–697. doi: 10.3109/13880209.2013.865239. [DOI] [PubMed] [Google Scholar]
- 16.Kassem FF, Harraz FM, Material P. Composition of the essential oil of Egyptian Cupressus L. cones. Flavour Fragr. J. 1991;6:205–207. doi: 10.1002/ffj.2730060308. [DOI] [Google Scholar]
- 17.El-azim MHMA, Abdelgawad AAM, El-mesallamy AMD. Chemical composition and antimicrobial activity of essential oil of Egyptian Ocimum basilicum L indo. Am. J. Pharm. Sci. 2015;2:837–842. [Google Scholar]
- 18.Aziz EE, Badawy EM, Zheljazkov VD, Nicola SM, Fouad H. Yield and chemical composition of essential oil of Egypt. J. Chem. 2019;62:533–540. [Google Scholar]
- 19.Aziz, E. E. et al. Chemical composition of Mentha pulegium L. (Pennyroyal) plant as influenced by foliar application of different sources of zinc. Egypt. Pharm. J.18, 53–59 (2019).
- 20.Khalil N, Ashour M, Fikry S, Naser A, Salama O. Chemical composition and antimicrobial activity of essential oils of selected Apiaceous plants growing in Egypt Future. Futur. J. Pharm. Sci. 2018;4:88–92. doi: 10.1016/j.fjps.2017.10.004. [DOI] [Google Scholar]
- 21.El-hawary, S. S. et al. Chemical composition and biological activities of peels and leaves essential oils of four cultivars of Citrus deliciosa var. tangarina. Am. J. Essent. Oils Nat. Prod.1, 1–6 (2013).
- 22.El, A., El, N. & Abd-elgawad, A. M. Essential oils constituents of aerial parts of Cyperus capitatus L. and Cyperus difformis L. grown wild in Egypt. J. Essent. Oil Bear. Plants 20, 1659–1665 (2017).
- 23.Abd-elgawad, A. M. et al. Chemical composition variations, allelopathic , and antioxidant activities of Symphyotrichum squamatum (Spreng.) Nesom essential oils growing in heterogeneous habitats. Arab. J. Chem. 10.1016/j.arabjc.2019.07.005 (2019).
- 24.Amin SM, et al. Comparative chemical study and antimicrobial activity of essential oils of three Artemisia species from Egypt and Saudi Arabia. Wiley Online Libr. Flavour Fragr. J. 2019;34:450–459. doi: 10.1002/ffj.3525. [DOI] [Google Scholar]
- 25.Saqqa, G. Al, Alian, A., Ismail, F. & Ramzy, S. Chemical composition of rocket , thyme and parsley essential oils and their effect on some fungi and aflatoxin production. MedCrave Online J. Toxicol.4, 277–282 (2018).
- 26.Ahl, H. A. H. S., Gendy, A. S. H., Mahmoud, A. A. & Mohamed, H. F. Y. Essential oil composition of Marrubium vulgare L. cultivated in Egypt. Int. J. Plant Sci. Ecol.1, 138–141 (2015).
- 27.Tarek N, et al. ScienceDirect comparative chemical and antimicrobial study of nine essential oils obtained from medicinal plants growing in Egypt. Beni-Suef Univ. J. Basic Appl. Sci. 2014;3:149–156. doi: 10.1016/j.bjbas.2014.05.009. [DOI] [Google Scholar]
- 28.Ghareeb, M. A., Refahy, L. A., Saad, A. M. & Ahmed, W. S. CODEN (USA): PCHHAX chemical composition , antioxidant and anticancer activities of the essential oil from Eucalyptus citriodora (Hook.) leaves. Der Pharma Chem.8, 192–200 (2016).
- 29.El-ghorab AH, Ramadan MM, El-moez SIA, April M, April M. Research journal of pharmaceutical, biological and chemical sciences essential oil, antioxidant, antimicrobial and anticancer activities of Egyptian Pluchea dioscoridis extract. Res. J. Pharm. Biol. Chem. Sci. 2015;6:1255–1265. [Google Scholar]
- 30.Halim, Ahmed F; Eldahshan, O. Comparison of the composition and antimicrobial activities of the essential oils of green branches and leaves of Egyptian navel orange (Citrus sinensis L.) O Sbeck. Chem. Biodivers.13, 681–685 (2016). [DOI] [PubMed]
- 31.Ohamed, M. I. E. M. & Bdelgaleil, S. A. M. A. Chemical composition and insecticidal potential of essential oils from Egyptian plants against Sitophilus oryzae (L .) (Coleoptera : Curculionidae) and Tribolium castaneum (Herbst) (Coleoptera : Tenebrionidae). Appl. Entomol. Zool.43, 599–607 (2008).
- 32.El Gohary, A. E., Hassan, E. A. & Ali, A. F. Enhancement of growth , yield and chemical constituents of rosemary (Rosmarinus officinalis , L.) plants by application of compost and biofertilization treatments. Middle East J. Agric. Res.4, 99–111 (2015).
- 33.Saad MMG. Chemical composition and biological activities of four citrus essential oils. J. Plant Protein Path. 2013;4:767–780. [Google Scholar]
- 34.Aboutabl, E. A., El-tantawy, M. E. & Shams, M. M. Chemical composition and antimicrobial activity of volatile constituents from the roots , leaves , and seeds of Arctium lappa L . (Asteraceae) grown in Egypt. Div. Pharm. Drug Ind. Res. Natl. Res. Center12, 173–176 (2013).
- 35.Soliman, F. M., Yousif, M. F., Zaghloul, S. S. & Okba, M. M. Seasonal variation in the essential oil composition of Origanum majorana L. cultivated in Egypt. Z Naturforsch C J. Biosci64, 611–615 (2009). [DOI] [PubMed]
- 36.Gad HA, Id IMA, Wink M. Phytochemical profiling and seasonal variation of essential oils of three Callistemon species cultivated in Egypt. PLoS ONE. 2019;17:1–15. doi: 10.1371/journal.pone.0219571. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Kouachea B, Brada M, et al. Chemical composition and acaricidal activity of Thymus algeriensis essential oil against Varroa destructor. NPC Nat. Prod. Commun. 2017;12:3135–3138. [PubMed] [Google Scholar]
- 38.Ibrahim, M. E., Ahmed, S. S., E.-S. & S. A. Essential oil composition of Iphiona scabra (DC) grow wild in Egypt. Int. Food Res. J.23, 1823–1826 (2016).
- 39.Aboutabl EA, Meselhy KM, Elkhreisy EM, Nassar MI, Fawzi R. Journal of essential oil bearing plants composition and bioactivity of essential oils from leaves and fruits of Myrtus communis and Eugenia supraxillaris (Myrtaceae) grown in Egypt. J. Essent. Oil Bear. Plants. 2013;14:37–41. [Google Scholar]
- 40.Hamdan DI, Abdulla RH, Mohamed ME, El-shazly AM. Chemical composition and biological activity of essential oils of Cleopatra mandarin (Citrus reshni) cultivated in Egypt. J. Pharmacogn. Phyther. 2013;5:83–90. doi: 10.1016/j.phcgj.2013.02.005. [DOI] [Google Scholar]
- 41.El-Kholany E. Utilization of essential oils from Citronella and Geranium as natural preservative in mayonnaise Ebtehal. Int. J. Microbiol. Biotechnol. 2016;1:49–59. [Google Scholar]
- 42.El-kashoury EA, El-askary HI, Kandil ZA. Chemical composition and biological activities of the essential oil of Mentha suaveolens Ehrh. Z. Naturforsch. C. J. Biosci. 2012;67:571–579. doi: 10.1515/znc-2012-11-1207. [DOI] [PubMed] [Google Scholar]
- 43.Rowayshed GH, Abd-Elhameed AA, et al. Effective chemical compounds and antibacterial activities of marjoram leaves, teucrium leaves and fennel fruits essential oils. Middle East J. Appl. Sci. 2014;4:637–647. [Google Scholar]
- 44.Vanessa R, et al. Chemical composition and antifungal activity of essential oils from four Asteraceae plants grown in Egypt. Z. Naturforsch C J. Biosci. 2018;73:313–318. doi: 10.1515/znc-2017-0219. [DOI] [PubMed] [Google Scholar]
- 45.El-Shazly, A., Hafez, S. & Wink, M. Comparative study of the essential oils and extracts of Achillea fragrantissima (Forssk.) Sch. Bip. and Achillea santolina L. (Asteraceae) from Egypt. Pharmazie59, 226–230 (2004). [PubMed]
- 46.Kf, T. & Zta, S. Chemical composition and antibacterial activity of volatile oil of Sequoia sempervirens (Lamb.) grown in Egypt medicinal & aromatic plants. Med. Aromatic Plants5 (2016).
- 47.Soliman FM, Fathy MM, Salama MM, Saber FR. Chemical composition and bioactivity of the volatile oil from leaves and stems of Eucalyptus cinerea. Inf. Healthc. 2014;52:1272–1277. doi: 10.3109/13880209.2014.889177. [DOI] [PubMed] [Google Scholar]
- 48.Ramadan, M. M., Ali, M. M., Ghanem, K. Z. & El-Ghorabe, A. H. Essential oils from Egyptian aromatic plants as antioxidant and novel anticancer agents in human cancer cell lines. Grasas y Aceites66 (2015).
- 49.Mohamed AA, et al. Assessment of antioxidant and antimicrobial activities of essential oil and extracts of Boswellia carteri resin. Int. J. Pharmacogn. Phytochem. Res. 2015;7:502–509. [Google Scholar]
- 50.Ibrahim TA, El-hela AA, El-hefnawy HM, Al-taweel AM. Chemical composition and antimicrobial activities of essential oils of some coniferous plants cultivated in Egypt. Iran. J. Pharm. Res. 2017;16:328–337. [PMC free article] [PubMed] [Google Scholar]
- 51.Fayed SA. Antioxidant and anticancer activities of Citrus reticulate (Petitgrain Mandarin) and Pelargonium graveolens (Geranium) essential oils. Res. J. Agric. Biol. Sci. 2009;5:740–747. [Google Scholar]
- 52.Hanafi RS, Sobeh M, Ashour ML, El-readi MZ. Chemical composition and biological activity of essential oils of cumin and coriander fruits from Egypt chemical composition and biological activity of essential oils of cumin and coriander fruits from Egypt. Nat. Prod. J. 2014;4:63–69. [Google Scholar]
- 53.Abdelhady MI, Aly HAH. Original article antioxidant and antimicrobial activities of Callistemon comboynensis essential oil. Free Radic. Antioxid. 2012;2:37–41. doi: 10.5530/ax.2012.2.8. [DOI] [Google Scholar]
- 54.Mahmoud GI. Biological effects, antioxidant and anticancer activities of marigold and basil essential oils. J. Med. Plants Res. 2013;7:561–572. [Google Scholar]
- 55.Badawy MEI, Abdelgaleil SAM. Composition and antimicrobial activity of essential oils isolated from Egyptian plants against plant pathogenic bacteria and fungi. Ind. Crop. Prod. 2014;52:776–782. doi: 10.1016/j.indcrop.2013.12.003. [DOI] [Google Scholar]
- 56.https://studio.azureml.net/. 10.1109/ICCES.2016.7821969
- 57.Wang, Y., Li, Y., Song, Y. & Rong, X. Applied sciences the influence of the activation function in a convolution neural network model of facial expression recognition. Appl. Sci.10 (2020).
- 58.Daynac, M., Cortes-cabrera, A. & Prieto, J. M. Application of artificial intelligence to the prediction of the antimicrobial activity of essential oils. Evidence-Based Complement. Altern. Med.2015 (2015). [DOI] [PMC free article] [PubMed]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The complete data set is available in the supplementary file (sup-file). The dataset is formatted to be suitable to be processed by Azure ML modules.