eISSN 2409-370X
International Journal of
Biology
and Chemistry
Al-Farabi Kazakh National University
al-Farabi Kazakh National University, al-Farabi ave., 71, 050040, Almaty, Republic of Kazakhstan website: http://ijbch.kaznu.kz/
Any inquiry for subscriptions should be sent to:
Prof. Mukhambetkali Burkitbayev, al-Farabi Kazakh National University al-Farabi ave., 71, 050040, Almaty, Republic of Kazakhstan
e-mail: Mukhambetkali [email protected]
The most significant achievements in the field of natural sciences are reached in joint collaboration, where important roles are taken by biology and chemistry. Therefore publi- cation of a Journal, displaying results of current studies in the field of biology and chem- istry, facilitates highlighting theoretical and practical issues and distribution of scientific discoveries.
One of the basic goals of the Journal is to promote the extensive exchange of informa- tion between the scientists from all over the world. We welcome publishing original papers and materials of biological and chemical conferences, held in different countries (by prior agreement, after the process of their subsequent selection).
Creation of International Journal of Biology and Chemistry is of great importance, since scientists worldwide, including other continents, might publish their articles, which will help to widen the geography of future collaboration.
The Journal aims to publish the results of the experimental and theoretical studies in the field of biology, biotechnology, chemistry and chemical technology. Among the em- phasized subjects are: modern issues of technologies for organic synthesis; scientific basis of the production of physiologically active preparations; modern issues of technologies for processing of raw materials; production of new materials and technologies; study on chemi- cal and physical properties and structure of oil and coal; theoretical and practical issues in processing of hydrocarbons; modern achievements in the field of nanotechnology; results of studies in various branches of biology, chemistry and related technologies.
We hope to receive papers from the leading scientific centers, which are involved in the application of the scientific principles of biological and chemical sciences on practice and fundamental research, related to production of new materials, technologies and ecological issues.
IRSTI 28.23.23 https://doi.org/10.26577/ijbch.2021.v14.i2.01
A.S. Makhmet1 ,M.G. Sharaev2 , A.E. Dyusembaev1 ,A.M. Kustubayeva1*
1Аl-Farabi Kazakh National University, Almaty, Kazakhstan
2Skolkovo Institute of Science and Technology, Moscow, Russia
*e-mail: [email protected]
Machine learning for brain signal analysis
Abstract. Machine learning (ML) is an effective tool for analysing signals from the human brain. Machine Learning techniques provide new insight into the understanding of brain function in healthy subjects and patients with neurological and mental disorders. Here we introduce the application of machine learning to brain signal analysis, specifically using two widely used brain signal collection methods: functional magnetic resonance imaging (fMRI) and Electroencephalography (EEG). The article provides a brief overview of the theoretical concept of machine learning and its types: supervised, unsupervised and reinforcement learning.
The potential of machine learning applications in pathology is discussed. Differences between EEG and fMRI methods regarding machine learning application and an overview of the techniques employed in different research studies are reviewed. The new machine learning methods invented for analysis of brain signals in the resting state and during the performance of the different cognitive tasks would be useful and worth considering in other domains, not limited to medicine.
Key words: EEG, fMRI, machine learning, MVPA, brain signal analysis.
Introduction
Machine learning applications are effective in solving many modern problems. Its widespread use has induced the emergence of a large amount of lit- erature. The present article aims to provide a brief introduction to machine learning techniques and its application in brain signals analysis, specifically to fMRI and EEG signals.
First, we provide a brief overview of the theo- retical concept of machine learning. Machine learn- ing is a scientific discipline in which techniques are designed so that machines can computationally ef- ficiently extract patterns, structures or relationships from data. It is a relatively new discipline, which lies at the intersection of two fields: mathematics and computer science. The majority of machine learning techniques are based on the following mathematical realms: linear algebra, analytic geometry, matrix de- composition, probability theory, vector calculus and optimization [1].
A learning process outcome would be called a model. Machine learning models describe relation- ships within observed data. A typical dataset that is considered in ML problems consists of dependent and independent variables. Independent variables can also be referred to as features, while a dependent
variable, which depends on independent variables, is referred to as a target.
Machine learning techniques are based on regres- sion, dimensionality reduction, density estimation and classification [1]. The key objective in regression and classification problems is to build a model that would map inputs x to corresponding values y, where x represents feature matrix and y represents a target vector [1]. The difference is that in regression the out- puts are continuous values, whereas in classification the outputs are discrete (or categorical) values. The key objective in dimensionality reduction problems is to reduce the number of features in the feature ma- trix, with minimum loss of potentially valuable infor- mation [1]. The key objective in density estimation problems is to describe a dataset from the perspective of a probability distribution [1].
Nowadays there are many different types of learning that exist in Machine learning. Traditionally, machine learning approaches are divided into three broad categories: supervised, unsupervised, and re- inforcement learning. Reinforcement learning is a type of learning, in which an ‘agent’ learns how to act in an environment by continuously getting ‘feed- back’ from that environment. In supervised learning, a model is trained on labeled data to learn a relation- ship between features and a target variable. Next, to
evaluate the performance of the model, the model is tested on previously ‘unseen’ data. Model training is the process during which a machine learns patterns and structure from available data. In contrast to the supervised learning approach, the unsupervised ap- proach works with unlabeled data, and a model is built to discover patterns and structures within the data.
Supervised learning is used in regression and clas- sification algorithms, whereas unsupervised learning is used in dimensionality reduction, density estima- tion, anomaly detection, autoencoding and cluster- ing techniques. In supervised learning, we are able to evaluate our model by comparing predicted and actual labels of test data. In contrast, in unsupervised learning there are no labels to predict, and, therefore, there is no direct evaluation for unsupervised learn- ing. However, an output of an unsupervised learning task is often used to construct an input (informative features) to the subsequent supervised learning task, and it can be evaluated via results of the subsequent supervised learning task and answer the question,
“Was the pattern discovered in unsupervised learning useful?” [2].
Machine learning application in clinics Medicine and pathology in particular are arg- Wuably the most promising domains to apply ma- chine learning: what could be more inspiring than contributing to saving millions of lives from complex diseases or improving the quality of life of those who are paralyzed? With each year there is an increased number of publications presenting applications of machine learning algorithms to medical data, yet this has not resulted in many meaningful contributions to clinical care [2]. The main reason is possibly low AI/ML algorithms trustworthiness, which comprises several parts: is the algorithm accurate and robust, how fair and transparent are its decisions, and finally, how interpretable are these decisions to the medical community [3]. While high accuracy and robustness (for example to low quality data) are well-known desired properties of AI/ML systems in all applied fields, model interpretability becomes crucial, espe- cially in a clinical setting. Another factor that makes it difficult to apply machine learning to medical data is that acquisition of data is a costly and lengthy pro- cess. For example, for brain signal analysis to acquire data using fMRI, one must have an fMRI scanner, which is considered to be expensive, and not many research groups can afford it. Comparatively, get- ting other types of data is much easier. For example, Netflix possesses all the data that is needed in order
to build its own recommendation system, because it analyses traffic that a user generates when interacting with a Netflix website or application.
Both supervised and unsupervised learning are used in pathology and human functions. One example is analysis of data obtained from patients with Heart Failure with preserved Ejection Fraction (HFpEF) [2]. HFpEF is a complex condition and reflects mul- tiple dominant pathophysiologic processes. The idea of the analysis was to group patients on the basis of qualitative echocardiographic and clinical variables.
Initially, there were 67 different features; after remov- ing highly correlated features, there were 46 predic- tors (features). Next, a regularized form of a cluster- ing algorithm was applied: clusters were determined by using multivariate Gaussian distributions and us- ing means and standard deviation assigned to each feature [2]. The clusters were formed by calculating a joint probability of membership for each patient.
Results of the comparison of the calculated clusters have demonstrated the differences across many phe- notypic variables. These phenotypic clusters resulted in becoming features in the supervised learning mod- el that predicted survival of HFpEF patients.
Study of signals of the brain, the most complex structure in the body, may help understand brain functioning in normal versus neurological or mental disorder brain conditions. Many studies use super- vised learning approaches to diagnose a brain func- tion pathology or classify symptom severity from concurrent neuroimaging data. Among brain function disorders are attention-deficit/hyperactivity disorder, autism, depression, and schizophrenia [4]. Moreover, machine learning technology is able to detect devia- tions from normative development trajectories as risk factors for psychopathology. Defining an age of an individual based on brain activity network patterns can be used to elucidate atypical development in chil- dren and adults with Tourette syndrome [4]. There is a growing number of studies using brain connectivity approaches [5]. This method is based on graph theo- ry and determines functional, structural, and causal dynamical networks. Therefore, brain connectivity measurements appear to serve as variables to deter- mine whether it is possible to predict subsequent di- agnosis or treatment outcomes.
Apart from supervised learning in brain signal analysis, an unsupervised learning approach can be used to cluster patients into subgroups with categori- cally different patterns of neuroimaging features.
The application of reinforcement learning is also considered in the field of neuroscience [6-7]. It is worth mentioning that the Reinforcement learning
approach itself gets its aspiration from the cognitive neuroscience field, as it tries to mimic brain function.
There are different techniques for measuring and mapping brain activity. These include Electroenceph- alography (EEG), functional magnetic resonance im- aging (fMRI), Positron emission tomography (PET) and Magnetoencephalography (MEG). Our focus in this article is on the two most popular noninvasive and safe methods to obtain brain signals during the cognitive task or resting states: fMRI and EEG.
Machine learning in functional MRI signal analysis
FMRI measures brain neural activity via mag- netic properties of blood, namely, the blood-oxygen level dependent (BOLD) signal. The method is based on the fact that with an increase in the activity of a particular area of the brain, blood flow to this area also increases, which means that the parameters of blood movement and level of oxygen in the vascu- lar bed change. A typical neuroimaging experiment holds for several sessions (runs) per subject. In fMRI, the whole brain is scanned for the duration of a ses- sion, resulting in many brain images per time unit, called volumes. The scanning rate affects the spatial and temporal resolutions of images. FMRI has good spatial resolution and a satisfactory level of temporal resolution, even though these spatial and temporal resolutions are attained at the expense of each other.
For one to perform an effective fMRI analysis, the data collected should first undergo the prepro- cessing stage. Preprocessing typically includes Re- aligning and Unwarping the Data, Slice-time correc- tion, Co-registration, Segmentation, Normalization, and Smoothing steps. The last Smoothing step is in many cases omitted to avoid distortion of neural ac- tivity intensity per voxel.
The first types of analyses that prevailed in the analysis of fMRI data were univariate and mass- univariate. In univariate analysis, an amplitude of a signal elicited from a voxel, which is a 3-dimen- sional pixel of a brain (usually 3mm*3mm*3mm), is analyzed in the context of each voxel separately. In mass-univariate analysis, a statistical inference about brain region responses to particular stimuli is made on the basis of the average activation value of a re- gion calculated using univariate analysis conducted for each voxel in that region of a brain. Nowadays univariate and mass-univariate analysis is enhanced by Multivoxel pattern analysis (MVPA), the first concept of which appeared in the early 2000s. It also
operates at a level of a voxel, however, the MVPA approach considers the fMRI analysis problem as a classification problem. Multivoxel pattern analysis has become a new paradigm for fMRI analysis in the world of neuroimaging.
The MVPA approach allows to ‘decode’ fMRI signals and maps them to sensory and motor events or participant’s mental state [8]. A brain activity that was triggered by a certain experimental condition is recorded and represented as a pattern of voxels for that condition. In MVPA, each voxel constitutes a dimension in space, correspondingly, every pattern of voxels i.e., brain activity can be represented as a dot in that voxel space. Thus, many points (voxel patterns) form clouds in the voxel space. Figure 1 shows a simple voxel space structure: the number of dimensions was simplified to 3 voxels, red dots rep- resent condition A and blue dots represent condition B. The plane separates two clouds. Such represen- tation of fMRI data makes it possible to apply dif- ferent supervised methods, such as support-vector machine (SVM) and linear discriminant analysis (LDA).
Functional MRI analysis is complicated by the vast number of voxels that are treated as features in classification methods. This creates a problem of “Curse of dimensionality”: a number of avail- able voxels reaches more than 30,000, whereas the number of trials (samples) are at the highest 100 [9].
Different approaches are employed to reduce data dimensionality and thus select only ‘useful’ voxels.
Two of the main standard approaches are region of interest (ROI) and Searchlight. Regions in ROI can be selected on the basis of anatomical structure, or on the results of application of the ANOVA method.
In the latter, voxels are selected on the ANOVA test of response of each voxel to the experimental con- ditions [10]. After ROI selection, a functional con- nectivity matrix can be calculated and vectorized in different ways. Obtained low-dimensional feature vectors are then used in ML models as predictors of various diseases, see for example [11].
In the Searchlight method, a classifier is conse- quently trained on a small spherical cluster of voxels, centered at each voxel of the indicated areas (usually the whole brain is selected). Classification accuracies for each spherical cluster are calculated and are as- signed respectively to the central voxels of the clus- ters. The subsets of clusters with good classification accuracy are identified, then either all voxels of these subsets are selected or the central voxels of the clus- ters of these subsets are selected [12].
7
Int. j. biol. chem. (Online) International Journal of Biology and Chemistry 14, № 2 (2021) A.S. Makhmet et al.
Common statistical approaches to dimensional- ity reduction, like ICA and PCA can also be success- fully applied to select features. An example of ICA application is Hunyadi’s [12] study of localization of epileptogenic zones, where ICA was applied to select epileptic components that later have been fed to classification algorithms [12]. Graph-theoretical approaches could be applied to fMRI-based connec- tivity matrices resulting in informative features for depression diagnostics [13].
Bleigh-Cohen [14] has proposed a new classifica- tion method called SBFE: patients with complex psy- chopathology were classified using data driven ROI search coupled with whole-brain machine-learning.
SBFE showed 91% accuracy in classifying schizo- phrenia patients with and without OCD [14]. They state that in contrast to standard ROI driven analy- ses, SBFE was able to classify schizophrenia patients with and without OCD with 91% accuracy. Thus, the application of this approach can have promising re- sults when needed to delineate patients of one com- plex psychiatric morbidity with the presence of dif- ferent symptoms.
De Martino [15] in his work has introduced re- cursive feature elimination (RFE) strategy. It uses SVM reclusively to remove irrelevant voxels and assess informative spatial patterns. The method has increased sensitivity for discriminative patterns.
According to Duff [16], the number of pre-pro- cessing and feature generation methods impacts pre-
diction accuracy more consistently than the choice of a classifier.
In many projects, the most popular choice of learning algorithms in fMRI analysis is SVM [17].
Among other different machine learning approaches in MVPA are Clustering algorithms [18], deep neural networks [19], and Representational Similarity Anal- ysis [20]. In some recent work, there are first attempts to make deep neural networks models on fMRI data interpretable for particular clinical diagnostics tasks, see [21-22] for details. An advantage of using deep learning over non-deep learning approaches is that features can be automatically learned by neural net- works, thus eliminating the need to conduct manual feature extraction and selection steps. Feature selec- tion and extraction steps are performed by special structures of deep learning architecture, called con- volutional and pooling layers.
In fMRI, due to the scarcity of samples compared to the number of features, the cross-validation tech- nique is often used. Cross-validation is one of the ML techniques to improve the efficiency of the algorithm by iteratively splitting data into different training/
testing sets, and thus artificially increasing the over- all training sample of the model. There are different strategies for selecting the validation fold. One of the popular strategies in MVPA is leave-subject-out or also known as leave-one-out (LOO-CV), where re- cords of one subject are assigned to be a test dataset [23].
Figure 1 – Representation of brain activities of different conditions in voxel space
Usually, experiments in neuroscience are con- structed in such a way that there is a balanced number of records with different labels in a dataset. Neverthe- less, studies with unbalanced datasets can also take place: an example is the study of intrusive memory formation, in which Clark [24] attempted to predict mental health symptoms by reconstructing idiosyn- cratic cognitive events. The labels of the conditions were not known at the time of the experiment. There- fore, balancing techniques were needed to be em- ployed prior to classifying the records.
As in any medical problem, sensitivity and speci- ficity are highly important metrics for evaluating model testing results in fMRI. Accuracy itself will not be enough to provide a full picture of the effective metric. In many studies, the ROC and Precision-Re- call (PR) curve is employed as an evaluation metric.
Machine learning in EEG analysis
In contrast to fMRI, EEG is an economical and easy-to-operate tool for recording brain activity [25]. It records the brain’s electrical activity over a period of time. Electrochemical processes occurring in the neuronal activity of the brain resulted in elec- trical oscillations on the brain surface with differ- ent amplitude and frequency – alpha, beta, gamma, theta, and delta rhythms. The relationship between these rhythms depends on external stimuli and the state of the human brain. These rhythms may differ in different brain conditions. For example, Gollan’s study of Frontal alpha EEG asymmetry confirms that depressed patients have a significantly higher difference in alpha rhythms between the left and right frontal part of the hemisphere than healthy participants [26].
Analysis of EEG has made it possible to develop brain-computer-interface systems. Brain-computer interface (BCI) system is an application that reads EEG signals in real-time and sends the decoded sig- nals to an external device.
The list of the neurological disorders that can be studied using EEG signals includes but is not limited to epilepsy, seizure prediction, Alzheimer’s disease, Mild Cognitive Impairment (MCI), Parkin- son’s disease, Creutzfeldt-Jakob Disease, sleep stud- ies, schizophrenia, analysis of emotional states [21].
Using machine learning classification methods, it is possible to predict and prevent the development of depression [27].
EEG signals have better temporal resolution than fMRI signals, but the spatial resolution is low [28].
EEG signals are collected by electrodes placed on the participant’s head. Each electrode represents an EEG channel and records a brain signal from a part of the brain that is closer to the electrode. Nevertheless, the location of an electrode may not correspond to the exact location of the brain source [29]. This consti- tutes an inverse problem that exists in EEG concern- ing localizing a brain source that elicited a particular EEG signal [30].
Although both EEG and fMRI methods employ signal processing techniques for feature extraction, it should be noted that EEG mostly employs time se- ries processing techniques, whereas fMRI combines image processing and time-series processing tech- niques. In contrast to fMRI, where features are the signal intensity value at each voxel, EEG features can be EEG bands spectral powers, coherence and inter- hemispheric asymmetry and other possible measured parameters from time-series.
EEG analysis is conducted in either time domain or frequency domain. Time-domain techniques en- compass wavelet transform and connectivity metrics, whereas frequency domain encompasses Fourier transform and further work with signal spectrum.
Physical stimuli induce changes in EEG signals called Event-related potentials. These potentials can be associated with mental activity and occur during stimulus perception or preparation and execution of actions. All types of EEG parameters may serve as features for machine learning to predict the brain ac- tivity in specific conditions. For example, the authors used coherence parameters of resting state to clas- sify depressed patients and healthy participants [31].
Machine learning is currently being applied to EEG data collected from healthy and depressed patients to predict performance differences between these two groups during a decision task [32], during an emo- tional regulation task [33] and vigilance objectives [34].
Deep learning approaches are becoming more popular in EEG analysis. Thus, Acharya [21] uses Convolutional Neural Network as a classifier to predict depressed or healthy people. The proposed solution by Achraya [21] used a Backpropagation algorithm to train the network, adaptive moment es- timation to optimize the parameters of the network structure and dropout technique to avoid overfitting.
A cross-validation strategy with 10 folds was used to test the dataset. Classification with Convolutional Neural Network has helped to achieve accuracy, sen- sitivity, and specificity over 90% for each left and right hemisphere.
Conclusion
The studies reviewed in the present article dem- onstrate the progress that has been made due to the use of machine learning techniques in fMRI analysis, EEG analysis and in the brain signal processing field overall. With technological advances and increasing computational efficiency (Moore’s law), the accuracy of the classification models in fMRI and EEG analy- sis may increase, even without changing the method- ology of the machine learning application.
Machine learning has made enormous progress in the last two decades largely due to the growth of computing power and the emergence of deep learning; and its techniques have proved to be a valuable tool in gaining more insights from data in any domain that can possess a vast amount of data.
Therefore, researchers in any field that deals with big data should be aware of the applicability and ca- pability of machine learning techniques to be able to leverage them to their benefits. The new ML meth- ods discovered or invented while solving cognitive science-related problems may be useful and worth considering in solving problems in other domains beyond medicine.
Funding
Research was supported by research grant from Ministry of Education and Science of Kazakhstan to A.M. Kustubayeva (AP08856595 “EEG/MRI study of brain development, emotional-cognitive functions, and genetic markers in different age groups”).
M.G. Sharaev was supported by RFBR, research project 18-29-01032.
References
1. Deisenroth M.P., Faisal A.A., Ong C.S. (2021) Mathematics for Machine Learning. Cambridge Uni- versity Press, 14-16 p.
2. Deo R.C. (2015) Machine Learning in Medicine. Circulation., vol. 132, no. 20, pp.
1920–1930. http://dx.doi.org/10.1161/CIRCULA- TIONAHA.115.001593.
3. Qayyum A., Qadir J., Bilal M., Al Fuqaha A.
(2020) Secure and Robust Machine Learning for Healthcare: A Survey. IEEE Reviews in Biomedi- cal Engineering., vol. 14, pp. 156-180. https://doi.
org/10.1109/rbme.2020.3013489.
4. Nielsen A.N., Gratton C., Church J.A., Dosen- bach N.U.F., Black K.J., Petersen S.E., et al.. (2019)
Atypical Functional Connectivity In Tourette Syn- drome differs between Children and Adults. Biologi- cal Psychiatry., vol. 87, no. 2, pp. 164-173. http://
dx.doi.org/10.1016/j.biopsych.2019.06.021
5. Rubinov M., Sporns O. (2010). Complex net- work measures of brain connectivity: Uses and inter- pretations. NeuroImage, vol. 52, no. 3, p. 1059–1069.
doi:10.1016/j.neuroimage.2009.10.003. https://doi.
org/10.1016/j.neuroimage.2009.10.003.
6. Botvinick M., Wang J.X., Dabney W., Mill- er K.J., Kurth-Nelson Z. (2020) Deep Reinforce- ment Learning and Its Neuroscientific Implications.
Neuron,. vol. 107, no. 4, pp. 603-616. https://doi.
org/10.1016/j.neuron.2020.06.014.
7. Niv Y. (2009). Reinforcement learning in the brain. Journal of Mathematical Psychology., vol.
53, no. 3, pp. 139–154.https://doi.org/10.1016/j.
jmp.2008.12.005.
8. Haynes J-D, Rees G. (2006) Decoding mental states from brain activity in humans. Nature Reviews Neuroscience., vol. 7, no. 7, pp. 523–534. http://
dx.doi.org/10.1038/nrn1931.
9. Hirose S, Nambu I, Naito E. (2015) An empiri- cal solution for over-pruning with a novel ensemble- learning method for fMRI decoding. Journal of Neu- roscience Methods., vol. 239, pp. 238–245. http://
dx.doi.org/10.1016/j.jneumeth.2014.10.023.
10. Jamshy S., Perez O., Yeshurun Y., Hendler T., Intrator N. (2012) Searchlight Based Feature Extrac- tion. In: Langs G., Rish I., Grosse-Wentrup M., Mur- phy B. (eds) Machine Learning and Interpretation in Neuroimaging. Lecture Notes in Computer Science, vol. 7263. Springer, Berlin, Heidelberg. https://doi.
org/10.1007/978-3-642-34713-9_3.
11. Sharaev M., Artemov A., Kondrateva E., Su- shchinskaya S., Burnaev E., Bernstein A., et al. (2018) MRI-Based Diagnostics of Depression Concomitant with Epilepsy: In Search of the Potential Biomarkers.
Proceedings of IEEE international conference on data science and advanced analytics. Turin, Italy, P. 555.
https://www.doi.org/10.1109/DSAA.2018.00071.
12. Hunyadi B, Tousseyn S, Dupont P, Van Huffel S, De Vos M, Van Paesschen W. (2015) A prospec- tive fMRI-based technique for localising the epilep- togenic zone in presurgical evaluation of epilepsy.
NeuroImage., vol. 113, pp. 329–339. http://dx.doi.
org/10.1016/j.neuroimage.2015.03.011.
13. Sharaev M., Artemov A., Kondrateva E., Ivanov S., et al. (2018) Learning Connectivity Pat- terns via Graph Kernels for fMRI-Based Depression Diagnostics. Proceedings of IEEE international con- ference on data mining workshops. Singapore, Sin-
gapore, P. 308. https://www.doi.org/10.1109/ICD- MW.2018.00051.
14. Bleich-Cohen M., Jamshy S., Sharon H., Weizman R., Intrator N., Poyurovsky, M., et al.
(2014) Machine learning fMRI classifier delineates subgroups of schizophrenia patients. Schizophre- nia Research., vol. 160 no. 1-3, pp. 196–200. http://
dx.doi.org/10.1016/j.schres.2014.10.033.
15. De Martino F, Valente G, Staeren N, Ash- burner J, Goebel R, Formisano E. (2008) Combining multivariate voxel selection and support vector ma- chines for mapping and classification of fMRI spa- tial patterns. NeuroImage., vol. 43, no. 1, pp. 44–58.
http://dx.doi.org/10.1016/j.neuroimage.2008.06.037.
16. Duff K. (2012) Evidence-Based Indicators of Neuropsychological Change in the Individual Pa- tient: Relevant Concepts and Methods. Archives of Clinical Neuropsychology., vol. 27, no. 3, pp. 248–
261. http://dx.doi.org/10.1093/arclin/acr120.
17. Yamamoto M, Bagarinao E, Kushima I, Taka- hashi T, Sasabayashi D, Inada T, Suzuki M, Iidaka T, Ozaki N. (2020) Support vector machine-based classification of schizophrenia patients and healthy controls using structural magnetic resonance imaging from two independent sites. PLoS One., vol.15, no.
11. https://doi.org/10.1371/journal.pone.0239615.
18. Aljobouri H.K., Jaber H.A., Koçak O.M., Al- gin O, Çankaya I. (2018) Clustering fMRI data with a robust unsupervised learning algorithm for neurosci- ence data mining. Journal of Neuroscience Methods., vol. 299, pp. 45–54. http://dx.doi.org/10.1016/j.jneu- meth.2018.02.007.
19. Svanera M, Savardi M, Benini S, Signoroni A, Raz G, Hendler T, et al. (2019) Transfer learn- ing of deep neural network representations for fMRI decoding. Journal of Neuroscience Methods., vol. 328, 108319. https://doi.org/10.1016/j.jneu- meth.2019.108319.
20. Kriegeskorte N. (2008) Representational similarity analysis – connecting the branches of sys- tems neuroscience. Frontiers in Systems Neurosci- ence., vol. 2, no 4., http://dx.doi.org/10.3389/neu- ro.06.004.2008.
21. Li X., Dvornek N.C., Zhou Y., Zhuang J., Ventola P., Duncan J.S. (2019) Graph Neural Net- work for Interpreting Task-fMRI Biomarkers. Pro- ceedings of international conference on medical im- age computing and computer assisted intervention.
Shenzhen, China, P. 485.
22. Yan W., Calhoun V., Song M., Cui Y., et al.
(2019) Discriminating schizophrenia using recurrent neural network applied on time courses of multi-site
FMRI data. EBioMedicine. https://doi.org/10.1016/j.
ebiom.2019.08.023.
23. Lueken U, Straube B, Yang Y, Hahn T, Bees- do-Baum K, Wittchen H-U, et al. (2015) Separat- ing depressive comorbidity from panic disorder: A combined functional magnetic resonance imaging and machine learning approach. Journal of Affec- tive Disorders., vol. 184, pp. 182–192. http://dx.doi.
org/10.1016/j.jad.2015.05.052.
24. Clark I.A., Niehaus K.E., Duff E.P., Di Sim- plicio M.C., Clifford G.D., Smith S.M., et al. (2014) First steps in using machine learning on fMRI data to predict intrusive memories of traumatic film foot- age. Behaviour Research and Therapy., vol. 62, pp.
37–46. http://dx.doi.org/10.1016/j.brat.2014.07.010.
25. Acharya U.R., Oh S.L., Hagiwara Y., Tan J.H., Adeli H., Subha D.P. (2018) Automated EEG- based screening of depression using deep convolu- tional neural network. Computer Methods and Pro- grams in Biomedicine., vol. 161, pp. 103–113. http://
dx.doi.org/10.1016/j.cmpb.2018.04.012.
26. Gollan J.K., Hoxha D., Chihade D., Pflieger M.E., Rosebrock, L., Cacioppo, J. (2014). Frontal al- pha EEG asymmetry before and after behavioral acti- vation treatment for depression. Biological Psychol- ogy., vol. 99, pp. 198–208. https://doi.org/10.1016/j.
chiabu.2020.104358.
27. Bachmann M, Päeske L, Kalev K, Aarma K, Lehtmets A, Ööpik P, et al. (2018) Methods for clas- sifying depression in single channel EEG using lin- ear and nonlinear signal analysis. Computer Methods and Programs in Biomedicine., vol. 155, pp. 11–17.
http://dx.doi.org/10.1016/j.cmpb.2017.11.023.
28. Liley D.T.J., Frascoli F. (2010) Approaches to Modelling the Dynamical Activity of Brain Function Based on the Electroencephalogram. In: Dewar R.L., Detering F. (eds) Complex Physical, Biophysical and Econophysical Systems. World Scientific Lecture Notes in Complex Systems, vol. 9. World Scientific Publishing Co., Singapore, Singapore. https://doi.
org/10.1142/9789814277327_000.
29. Teplan M. (2002) Fundamentals of EEG mea- surement. Measurement Science Review., vol. 2, no.
2, pp. 1-11
30. Michel C.M., Brunet D. (2019) EEG Source Imaging: A Practical Review of the Analysis Steps. Frontiers in Neurology., vol 10. https://doi.
org/10.3389/fneur.2019.00325
31. Ledovsky A., Mnatsakanian E., Sharaev M., Burnaev E. (2019) O-21 Separation of major depres- sion patients from healthy controls using machine learning approach to resting-state EEG. Clinical
Neurophysiology., vol. 130, no. 7, p. e28. https://doi.
org/10.1016/j.clinph.2019.04.337.
30. Kustubayeva A., Kamzanova A., Kudai- bergenova S., Pivkina V., Matthews G. (2020) Major depression and brain asymmetry in a decision-mak- ing task with negative and positive feedback. Symme- try., vol. 12, no. 12, pp. 1–25. https://doi.org/10.3390/
sym12122118.
33. Kustubayeva A., Tolegenova A., Mat- thews G. (2013) EEG-brain activity in different
strategies of emotions’ self-regulation: Suppres- sion and reappraisal. Psikhologicheskii Zhurnal.,, 34(4), pp. 58–68. https://doi.org/10.3390/brain- sci9090225.
34. Kamzanova A., Kustubayeva A., Matthews G. (2012) Diagnostic monitoring of vigilance dec- rement using EEG workload indices. Proceedings of the Human Factors and Ergonomics Society An- nual Meeting. Boston, Massachusetts, USA. P. 203.
https://doi.org/10.1177/1071181312561019.
IRSTI 34.31.17 https://doi.org/10.26577/ijbch.2021.v14.i2.02
M. Мutalkhanov , K. Sissemali* , A. Alnurova , A. Akilbekova , M. Tauasarova , Z. Bassygarayev , K. Boguspaev
Аl-Farabi Kazakh National University, Almaty, Kazakhstan
*e-mail: [email protected]
Determination of the correlation between the accumulation of pigments and natural rubber in Tau-saghyz
Abstract. Natural rubber is a valuable polymer used in a wide variety of industries, from tire manufactur- ing to medical products and applications in the food industry. The tau-sagyz rubber plant (Scozonera tau- saghyz Lipcsh et Bosse) is one of the few plants in the world that can be used in the production of natural rubber when cultivated in culture. To increase the productivity of plants, it is necessary to establish factors affecting the biosynthesis of natural rubber. In our studies, the level of pigments (chlorophyll–a, b, carot- enoids) was chosen as a factor that may have a positive effect on the accumulation of rubber in the roots of S. tau-saghyz. For this purpose, samples (leaves, roots) were taken from 3 groups of plants (north, south, east) in the Karatau mountains of the Kurman-Tas range. The extraction of pigments from leaf samples was carried out with acetone, and the extraction of rubber from the roots was carried out with hexane. The ob- tained data on the content of pigments and rubber in the samples were statistically processed using the t-test and Pearson correlation analysis. As a result, a statistical difference in the values of pigment accumulation in leaves and rubber in roots between the 3 studied groups of plants is shown. On the other hand, a very low level of correlation was found between the accumulation of pigments in the leaves and rubber in the roots in all three groups. The obtained results reliably showed that the level of pigment accumulation in the leaves does not affect the accumulation of rubber in the roots of Tau-sagyz.
Key words: Tau-saghyz, pigment accumulation, natural rubber, correlation.
Introduction
The importance of development of alternative crops for the natural rubber production. Natural rub- ber (NR) is a high molecular weight polymer found in more than 2,500 plant species [1;2]. Unfortunately, only few of these species can produce NR with desir- able characteristics. NR is used as the raw material for the multiple types of products, such as medical devices, surgical gloves, aircraft, car tires, engineer- ing components, pacifiers, clothes, toys and other [3]. NR possesses several properties that makes it so attractive for industrial use, including high tensile, tear, fatigue strength, ability to stick to itself and to other materials, moderate resistance to environmen- tal damage by heat, light and ozone water, bacterial and viral barrier properties [4].
Currently the majority of NR is collected from tropical tree, member of family Euphorbiaceae He- vea brasiliensis (Muell Arg). The tree is native to rainforests of Amazon region in South America. This includes forests of Brazil, Venezuela, Ecuador, Co- lombia, Peru, and Bolivia [5]. However, due to the
increase of demand for NR, Hevea were introduced to other countries. Currently the main producers of NR are Asian countries, in particular: Thailand with 35 % of worldwide NR supply and Indonesia with 30 % of worldwide NR supply [6]. However, the fact that NR is extracted exclusively from Hevea is a big disadvantage and possesses a big threat for entire rubber industry. An example of such threats is patho- gens that can either significantly decrease quantity and quality of extracted rubber, as well as wipe an entire plantation of Hevea [7]. This became possible because of lack of genetic variability that could slow the infection’s progression. Strategic pathogens in- clude Microcyclus ulei (South American leaf blight (SALB), Phytopthora spp. (abnormal leaf fall), Oid- ium Heveae (powdery mildew), Cornyespora cassii- cola (Corneyspora leaf fall), Corticium salmonicolor (pink disease), Rigidoporus spp. (white root disease) [8]. Another disadvantage of Hevea exclusive rub- ber production is its cultivation restrictions. Hevea has very strict environmental growing requirements.
This prevents its cultivation on landscapes located outside of certain tropical region, while simultane-
ously makes it extremely susceptible to climate changes [1]. This characteristic also makes its cul- tivation a significant part of deforestation problem, since to establish the plantation a corporation or in- dividual farmer must clear the land from forest to plant rubber trees [9]. The cost of Hevea originated NR is another disadvantage. Since latex harvesting is done manually by tapping the bark of rubber tree it is very labour-intensive process [10]. Combine it with fast increase of tapper’s age due to unwillingness of younger generation to become tappers since it con- sidered a low-class traditional job and not classified as a profession [11]. Combined with annually increasing demand for NR, increasing shortage of skilled tappers it leads to an increase of Hevea harvested rubber cost [1]. However, Hevea alone cannot satisfy the demand for NR. According to “Statista” website the worldwide demand for NR keeps growing annually as indus- tries keep growing and developing. For example, in 2010 in the world has consumed around 10.7 million tons of NR (cis-1,4-polyisoprene), while in 2019 this number reached up to 13.7 million tons of NR [12].
This means that the shortage of rubber in 2019 were equal to around 100 thousand tones. This situation has sparked an interest of in alternative rubber producing crops. Currently guayule (Parthenium argentatum) and Russian dandelion (Taraxacum kok-saghyz) are the most well-known and promising ones [13].
Scorzonera tau-saghyz a potential rubber crop.
Tau-saghyz inhabits stony-rubble slopes of Karatau mountain plateaus at 500-2000 meters above sea level [14;15]. Tau-saghyz is a member of Asteraceae family, perennial, dicotyledon plant with diploid set of chromosomes (n=14) [16]. Tau-saghyz was ini- tially discovered during Soviet Union times during exploration of mountains of South Kazakhstan. This discovery pushed the All-Union Research Institute of Rubber and Gutta-Percha to establish research centers in villages Burnyi and Atabayevo as well as a research station in the central part of the Karatau ridge [14]. Tau-saghyz could become a strategic crop for Kazakhstan because its ability to synthesize and store NR. Scorzonera tau-saghyz is one of the plant species capable of synthesizing NR in quantities suf- ficient to be considered a rubber crop. In Scorzonera tau-saghyz NR is synthesized and stored mainly in- side the underground stem (caudex) and root. Previ- ously obtained data on rubber content of tau-saghyz showed that it could accumulate up to 40 % of dry mass of the roots consists of rubber. Plant stores rub- ber in its roots and underground stem. If the root is damaged latex appears on the surface of the wood in form of milky-white or yellowish-green coloured substance which quickly coagulates on air (Figure 1).
Figure 1 – Latex appearing on the surface of cut root.
Photo by Мutalkhanov M., 2019
Unfortunately, the population of tau-saghyz has been reduced drastically during the years of World War II (1941-1945). During this period for the needs of military industry, more than 12 million roots were harvested, dry weight of which were equal to ap- proximately 908 metric tons is to [17]. These events had significant impact on population of tau-saghyz, which was significantly damaged by the end of these years. As result in 1978 Scorzonera tau-saghyz was included into The Red Book of the USSR and further into the Red Book of the KazSSR and Red Book of the Republic of Kazakhstan [14]. There are several variations of tau-saghyz. The most notable ones were identified at following locations: Kaynar-Bastau (eastern part of Karatau mountain), Jelagan-Ata (cen- tral part Karatau mountain) and Leontyevka (eastern part of Karatau mountain) [18].
Currently the population of Scorzonera tau- saghyz has started to recover. However, there are number of factors limiting its recovery in natural con- ditions. These factors include its weak competitive- ness, low seed germination rate and intensive con- version of the territory into pastures and plantations.
Combination of these factors significantly slows down recovery of population both inside and outside of natural reserve [14]. One of the ways to solve this problem is application of microclonal propagation from apical meristem. As well as cultivation at con- trolled conditions within a laboratory and on a field.
As result of such projects population of Scorzonera tau-saghyz started to recover. Currently tau-saghyz is included into the decree of the Government of the Republic of Kazakhstan of October 31, 2006 “On ap- proval of the Lists of rare and endangered species of animals and plants” and is protected by law [19]. The
plant can be found on the territory of Karatau state natural reserve.
Morphology of Tau-saghyz. Scorzonera tau- saghyz is a woody semi shrub plant. The height of its aerial part could reach up to 25-40 cm. The aerial part of the plant takes form of a rosette, shaped as squat hemispherical cushions. Compactization of these cushions could either be sparse or mildly dense. On Figure 2 a typical look of S. tau-saghyz cushion is presented.
Figure 2 – Aerial part of Scorzonera tau-saghyz. Photo by Мutalkhanov M., 2019
These cushions are made of multiple short, densely branched, thick, woody, perennial branches.
These branches typically grow from area close to the root collar. In general, these rosettes reach up to 30 cm. in diameter. However, the environmental condi- tions in which plant was growing could significantly alter the diameter, sometimes it could reach up to 100 cm [20].
The stem is covered by multiple dry leaves. The colour of these leaves varies from gray-brown to dark green with weak gloss, the texture of covering leaves is rough. Almost all of the leaves are sedated into deltoid base. Typically, the rosette contains between 6 and 24 leaves. Typically, the colour of leaves is greyish green, ends of the leaves have sharp edge of reddish-brown colour. The surface of the leaf may be either smooth or mildly pubescent. The lower part of the leaf has a distinct furrow on it. The central bigger one forms a distinct keel; the lateral ones are smaller and located along the central. The furrows are white coloured with distinct gloss. Leaves are expanded
into an axil shaped like a delta. On the exterior the axil is smooth, cream coloured and glossy. An inte- rior of the axil is densely covered by silky, slightly curly, light-brown, shiny hairs, up to 3-4 mm long [20]. The rosette’s base is approximately 1 cm wide.
There are several rows of scales covering this base.
The external scales are 9 mm long and wide, round shaped but with distinct point at the top, internal ones are up to 3 cm long, lanceolate and elongated. The scales are semi-transparent, with dark brown colora- tion. The central vein of these leaves is more promi- nent, and the multiple (up to 10) lateral veins are less noticeable. They’re bare outside, and slightly pubes- cent inside at the foundation, with a tiny amount of hair [20]. The underground part of the plant (caudex) compared to the short but highly branched aerial part is significantly longer, with scree habitats resulting in highest length of underground stem. The bark of the underground part is bare, with rough texture, and has a dark brown colour. Each branch of caudex ends with a rosette of leaves or sometimes it can end with a flowering shoot [14; 20]. Caudex has several purposes, for example: nutrients and water storage, nutrients transmission, as well as participation in vegetative reproduction of the plant, which remains the main mean of reproduction. Inside the caudex S. tau-saghyz have a multinucleated tube-like cell called laticifers. In these laticifers occurs the reac- tion of polymerization, which leads to the synthesis of cis-1,4-polyisoprene or NR.
Rubber biosynthesis. NR is a cis-1,4-polyiso- prene polymer that consists of Isopentenyl pyrophos- phate (IPP) monomers. IPP is an isoprenoid precur- sor which acts as an intermediate in the mevalonate (MVA) and in the non-mevalonate (MEP) pathway of isoprenoid precursor biosynthesis. IPP for rubber bi- osynthesis could either be derived through the MVA pathway occurring in cytosol or through MEP path- way occurring in plastids [8]. Studies performed on Hevea revealed that latex carbon used during rubber biosynthesis does not come directly from photosyn- thetic apparatus, but from stored carbohydrates like starch [21].
Indication of phonotypical traits could be asso- ciated (directly or indirectly) with increased accu- mulation of rubber is the main goal during the de- velopment of commercially viable rubber crop. It’s suggested that increased quantities of pigments could have a beneficial effect, due to the increased supply of carbon which further will be used during rubber biosynthesis.
Materials and methods
Plant materials. Samples of the leaves, caudices and roots of Scorzonera tau-saghyz were collected by digging from wild fields at Karatau mountains in the Turkestan district. In particular, on the Kurman- tas ridge (Figure 3-a) and Kanyon-Teris Akkan (Fig- ure 3-b) in the national park. Samples were collected from May to June 2019.
The samples were collected from the plants growing on two ridges of mountains: south, north and west. The number of samples taken constituted 34 from south and west ridges and 32 from north ridge. Obtained samples were transported to Labora- tory of ecological biotechnology at Al-Farabi Kazakh National University (KazNU). To preserve harvested
samples during the transportation collected leaves, caudices and roots were stored in liquid nitrogen.
After the samples got delivered, they were stored at -80℃ until use.
Pigment extraction and quantification. Determina- tion of pigment quantity were performed using Arnon (1945) method [22]. 1 g of defrosted leaves were taken and grounded in 5 ml of 90 % acetone using pestle and mortar. Obtained extract were transferred to centri- fuged at 5000 rpm for 5 min. After centrifugation com- plete the supernatant were transferred into clean tubes, while avoiding disrupting the pellet. The examination of the supernatant was performed on spectrophotom- eter. Light absorption read at wavelength of 663, 645, and 441 nm. Quantity of pigment was calculated using Holm-Wettstein formulas (1.1-1.4):
a b
Figure 3 – Karatau mountains area where samples taken. Photo by Мutalkhanov M., 2019.
Note: a – Kurmantas ridge; b – Caniyon-Teris Akkan
where:
A662- light absorption at wavelength of 662nm A644-light absorption at wavelength of 644nm A441 – light absorption at wavelength of 441nm Rubber extraction and purification. Root samples washed with running water to clean them of the soil.
С𝐶𝐶𝐶А ���/ml� � ����4 �𝐴𝐴662� � ����� �𝐴𝐴644� (1.1) С𝐶𝐶𝐶𝑏𝑏 ���/ml� � 2��426 �𝐴𝐴662� � 4�6�� �𝐴𝐴644� (1.2)
С𝐶𝐶𝐶𝐶𝐶�𝑏𝑏 ���/ml� � ����4 �𝐴𝐴662� � 2��4�6 �𝐴𝐴644� (1.3)
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶�4,695 (𝐴𝐴441� � 0.268 (С𝐶𝐶𝐶𝐶𝐶�𝑏𝑏� (1.4) After washing, the roots were air dried in the shade.
These roots transferred to KazNU where they have been stored at -80℃ for 3 months. The root samples were grinded into powder by mill. Approximately 4g of powder is then transferred to Erlenmeyer flask with 50ml hexane. The remaining hexane were removed by
IKA RV05 basic rotary evaporator (IKAWerkeGmbH
& Co. KG, Germany) with IKA HP4 basic water bath (IKAWerkeGmbH & Co. KG, Germany) have been used. After that the remainings were dried in a VD115 vacuum dryer (BINDERGmbH, Germany) at 24 °C for 22‒24 h. All procedures performed at 23-25 °C.
Statistical analysis. The difference in pigment and rubber content between the groups examined using two-tailed t-test. Correlation analysis between pigment and rubber content performed using Pearson correla- tion. The calculations were performed using Microsoft Excel Add-on XLStat (ver. 2020.4) by Addinsoft.
Results and discussion
Differences in pigments and rubber content.
Determination of pigment content in leaves of Tau- saghyz revealed that the content of pigments in leaves strongly depends on location of slope. So, the mean pigment content in leaves from East sam- ples were highest, followed by South and North samples respectively (Table 1). The same pattern observed during determination of carotenoid con- tent. T-test have shown that groups from different slopes .
Table 1 – The results of the t-test analysis of pigments and rubber content between groups
Substance Groups tObserved tCritical DF alpha
Chlorophyll-A
East-North 3.809 1.998 65 0.05
South-North 2.420 1.998 65 0.05
South-East 1.376 1.998 66 0.05
Chlorophyll-B
East-North 3.902 1.998 65 0.05
South-North 1.431 1.998 65 0.05
South-East 1.696 1.997 66 0.05
Chlorophyll total
East-North 3.951 1.998 65 0.05
South-North 2.200 1.998 65 0.05
South-East 1.412 1.997 66 0.05
Carotenoid
East-North 3.755 1.998 65 0.05
South-North 2.296 1.999 65 0.05
South-East 1.260 1.997 66 0.05
Rubber
East-North 0.280 1.998 65 0.05
South-North 0.125 1.998 65 0.05
South-East 0.195 1.997 66 0.05
The t-test showed that tObserved> tCritical (3.809>1.998 with DF = 65, alpha = 0.05), which shows a statisti- cally significant difference between East and North with 95 % probability. A similar analysis of the fol- lowing groups established a significant difference between South and North (2.420>1.998; DF = 65, alpha = 0.05). At the same time, South and East (1.376<1.998; DF = 66, alpha = 0.05) showed no dif- ference between each other. Content of chlorophyll- B differed between East and North (3.902>1.998; DF
= 65, alpha = 0.05). However, no difference indicated between South-North (1.431<1.998; DF = 65, alpha
= 0.05) and South-East (1.696<1.998; DF = 66, al- pha = 0.05). Total chlorophyll content differed be- tween the East-North (3.951>1.998; DF = 65, alpha
= 0.05) and South-North (2.200>1.998; DF = 65, alpha = 0.05) slopes. At the same time, South-East (1.412<1.998; DF = 66, alpha = 0.05) showed no dif- ference between each other. Carotenoid content were
different between the East-North (3.755>1.998; DF = 65, alpha = 0.05) and South-North (2.296>1.998; DF
= 65, alpha = 0.05) slopes. At the same time, South- East (1.260<1.998; DF = 66, alpha = 0.05) showed no difference between each other. Rubber accumu- lation shown no statistically difference between the groups East-North (0.280>1.998; DF = 65, alpha = 0.05), South-North (0.125>1.998; DF = 65, alpha = 0.05) and South-East (0.195<1.998; DF = 66, alpha = 0.05). Results show that plants from South and East have significantly higher quantities of pigments com- pared to North. On the other hand content of rubber showed no difference between the groups.
Correlation analysis of the rubber content and the pigments content. To determine the degree of association between the accumulation of photosyn- thetic pigments in the leaves and the accumulation of rubber in the roots, the linear correlation coefficient r-Pearson used (Table 2).
Table 2 – The results of the correlation analysis between the pigments content in leaves and the content of rubber in roots in samples from Karatau mountains slopes
Rubber
Slopes North East South
Ch-a 0.169 -0.090 -0.263
Ch-b 0.094 -0.088 -0.286
Ch-a+b 0.150 -0.088 -0.293
Car 0.225 -0.098 -0.099
Data from the correlation analysis between the rubber content in roots and the pigment content in leaves of Scorzonera tau-saghyz shows that rubber content in roots has very weak correlation with all four pigment variants. Among the groups only north, which showed lowest mean quantity of pigments, dis- played positive correlation. On the other hand south and east, which showed the higher mean quantity of pigments, displayed negative correlations. However, due to the weak association between pigments and rubber quantities, it is suggested that pigment quan- tity not associated with rubber accumulation in roots.
Obtained data aligns with data obtained by Thal- er P. et al. (2016) [23], who tracked the movement of carbon isotope 13C throughout the body of Hevea, which allowed them to track the distribution of car- bon in organism of plant. According to their findings carbon for rubber biosynthesis do not come direct- ly from photosynthetic apparatus. Instead it “accu- mulated in mixed pool of carbohydrates within the plant”. Possible reason for the lack of correlation may be the fact that there is a direct competition for organic compounds between the processes of bio- mass accumulation and rubber accumulation. Taking into account the fact that rubber is a secondary me- tabolite, its possible that the accumulation of biomass will be of paramount importance for the organism.
Moreover, since the samples of leaves and roots col- lected during the flowering phase of tau-saghyz, the costly process of rubber formation were suppressed.
Conclusion
NR is biopolymer used in wide variety of indus- tries. S. tau-saghyz is a plant that can be used as an alternative rubber crop, which in term could have a beneficial effect on economy and industry of the country. During experiment samples of leaves and roots collected from 3 groups of wild plants(South, East, North) undergo the pigment and rubber ex- traction processes. Results of t-test indicated clear difference in pigment content between North and
South, East, while non have been indicated between South and East. In case of rubber accumulation. t-test showed lack of statistically significant difference be- tween all three groups.
Pearson correlation analysis showed that all of the pigment variants have very weak correla- tion with rubber accumulation. Because of that it were considered negligible. Possible reason for it being direct competition for organic compounds between the processes of biomass accumulation and rubber accumulation. Its possible that col- lection of rubber while plant is in dormant state could increase rubber yield due to the absence of the competition.
Acquired data shows that pigment content cannot be used as a mean to identify plants with high rubber productivity. Researches on other mechanisms are re- quired in order to identify phenotypical traits associ- ated with increased rubber content.
Funding
The work was financially supported by MES RK within the Project No. AP08053131 Efficient editing of the Scorzonera Tau-Saghyz genome using CRIS- PR/Cas9 technology to obtain genetically improved plants with an increased content of natural rubber (for 2020-2022 yy).
References
1 van Beilen J., Poirier Y. (2007) Establish- ment of new crops for the production of natural rub- ber. Trends Biotechnol., vol. 25, no. 11, pp. 522-529.
https://doi: 10.1016/j.tibtech.2007.08.009.
2 Rasutis D, Soratana K, McMahan C.M., Lan- dis A.E. (2015) A sustainability review of domestic rubber from the guayule plant. Ind Crops Prod., vol.
70. pp. 383–394. doi:10.1016/j.indcrop.2015.03.042 3 Beliczky L. S., Fajen J. (2012) Encyclopae- dia of Occupational Health and Safety 4th Edition.
Chapter 80 – Rubber Industry. https://www.iloen-
cyclopaedia.org/contents/part-xii-57503/rubber-in- dustry.
4 Industrial Rubber Goods: The Buyer’s Guide. (from July 1, 2020, corrected at July 3, 2019, accessed at July 6, 2021). http://www.industrialrub- bergoods.com/natural-rubber.html.
5 Rainforest-Alliance. (from June 24, 2021, corrected at June 29, 2021, accessed at July 3, 2021).
https://www.rainforest-alliance.org.
6 Bottier C. (2020) Biochemical composition of Hevea brasiliensis latex: A focus on the protein, lipid, carbohydrate and mineral contents. ADV BOT RES., vol. 93, pp. 201–237. https://doi.org/10.1016/
bs.abr.2019.11.003.
7 Buranov A. U., Elmuradov B. J. (2010) Ex- traction and Characterization of Latex and Natural Rubber from Rubber-Bearing Plants. J. Agric. Food Chem., vol. 58, no. 2, pp. 734–743
8 Cao X., Yan J., Lei, J., Li, J., Zhu, J., Zhang, H. (2017) De novo Transcriptome Sequencing of Me- JA-Induced Taraxacum kok-saghyz Rodin to Identify Genes Related to Rubber Formation. Sci. Rep., vol.
7. https://doi.org/10.1038/s41598-017-14890-z.
9 Warren-Thomas E.M., Edwards D.P., Beb- ber D.P. et al. (2018) Protecting tropical forests from the rapid expansion of rubber using carbon pay- ments. Nat. Commun., vol.9, pp. 911. https://doi.
org/10.1038/s41467-018-03287-9.
10 Mooibroek H., Cornish K. (2000) Alter- native sources of Natural Rubber. Appl. Micro- biol. Biotechnol., vol. 53, pp. 355-365. https://doi.
org/10.1007/s002530051627.
11 Hazir M., Kadir R., Gloor E., Galbraith D.
(2017) Effect of agroclimatic variability on land suitability for cultivating rubber (Hevea brasil- iensis) and growth performance assessment in the tropical rainforest climate of Peninsular Malaysia.
Clim. Risk Manag., vol. 27, pp. 100-203. https://doi.
org/10.1016/j.crm.2019.100203.
12 Statista. (from June 29, 2021, corrected at July 2, 2021, accessed at July 4, 2021). https://www.
statista.com/topics/3268/rubber/#dossierKeyfigures.
13 Cornish K. (2011) Similarities and differ- ences in rubber biochemistry among plant species.
Phytochemistry., vol. 57. – pp. 1123-1134. https://
doi.org/10.1016/s0031-9422(01)00097-8.
14 Boguspaev K.K., Adil’baev Zh.A., Faleev D.G., Zhanataev Sh.A., Turasheva S.K., Sambetov K.K. (2013) Perspektivy razrabotki tehnologij voss- tanovlenija populjacii rastenij tau-sagyza (Scorzon- era tau-saghyz (Lipsch. et Bosse) v Karatauskom
gosudarstvennom prirodnom zapovednike [Prospects for the development of technologies for restoring the plant population of tau-sagyz (Scorzonera tau-saghyz (Lipsch. et Bosse) in the Karatau State Nature Re- serve]. Vestnik KazNU. Serija jekologicheskaja. vol.
2, pp. 38. https://pandia.ru/text/80/317/71142.php.
15 Plants For A Future Database (from June 16, 2021, with corrections at June 20, 2021, accessed at June 26, 2021). https://pfaf.org
16 Bordbar F., Makki P., Alifarsangi F. (2019) Karyotype Investigations in Iranian Scorzonera tor- tuosissima and S. intricata, Asteraceae. Cytologia, vol. 84, no. 1, pp. 69–72. https://doi.org/10.1508/cy- tologia.84.69.
17 Arias M., Hernandez M., Remondegui N., Huvenaars K., van Dijk P., Ritter E. (2016) First ge- netic linkage map of Taraxacum koksaghyz Rodin based on AFLP, SSR, COS and EST-SSR markers.
Sci. Rep. vol.6. https://doi.org/10.1038/srep31031.
18 Bondarenko P. V., Tarakanov S.G. (1941) Tau-sagyz. Priemy vyrashhivanija v Srednej Azii [Tau-sagyz. Cultivation techniques in Central Asia].
Tashkent: AN SSSR, Uzbekskij filial. Izdatel’stvo UzFAN.
19 Collection of acts of the President of the Republic of Kazakhstan and the Government of Re- public of Kazakhstan. (2006) “On approval of the list of environmental protection objects of special eco- logical, scientific and cultural significance” . vol. 40, pp.444.
20 Кul’tijasov M.V. (1938) Tau-Sagyz i vve- denie ego v kul’turu [Tau-Sagyz and its introduction into culture]. Moskva: Izdatel’stvo Akademii nauk SSSR. pp. 45-51
21 Men, X., Wang, F., Chen, G. Q., Zhang, H. B., & Xian, M. (2018). Biosynthesis of Natural Rubber: Current State and Perspectives. Int. J. Mol.
Sci., vol. 20, no. 1, pp. 50. https://doi.org/10.3390/
ijms20010050.
22 Yavorska N., Vorobets N. (2019) Photosyn- thetic Pigments in Shoots of Vaccinium corymbosum L. (Cv. Elliott). Agrobiodivers. improv. nutr., health life qual., vol. 3. pp. 93-100. https://doi.org/10.15414/
agrobiodiversity.2019.2585-8246.093-100.
23 Thaler P., Duangngam O., Kasemsap P., Sa- thornkich J., Chayawat C., Satakhum D., Priault P., Desalme D., Chantuma P.,Ghashghaie J., Epron D.
(2016) The source of latex. Tracing carbon from leaf photosynthesis to latex metabolism in rubber trees us- ing carbon sTable isotopes. CRRI and IRRDB Inter- national Rubber Conference. Cambodia, P. 260-268.
IRSTI 34.29.01 https://doi.org/10.26577/ijbch.2021.v14.i2.03
A.S. Abdygalieva* , Zh.O. Ospanbaev
Kazakh research Institute of agriculture and crop production, Almaty region, Kazakhstan
*e-mail: [email protected]
Assessment of the current state of populations and study of phytochemical features of Conium maculatum L.
in the Ili-Alatau mountains
Abstract. A comprehensive study of the anatomical structure of the plant Conuim maculatum L. from the genus Ariaceae Lindl., which has valuable medicinal properties is presented. Currently, the medicinal plant Conuim maculatum L. is widely used in folk medicine. This plant is used by specialists for the prevention of various diseases, especially for the treatment of various types of cancer, as well as for the preparation of active painkillers and anti-inflammatory drugs. The article analyzes the current state of the plant Conuim maculatum L. in the Іli-Alatau, diagnostic signs were identified and the main features of its anatomical structure were analyzed. The article analyzes in detail the biological features of the medicinal plant Conuim maculatum L., the main promising distribution zones of the plant in the country, and also describes the systematic classification, phytochemical features and chemical structure of this species. Eco-phytocenotic features of Conuim maculatum L. are shown on a population growing at the foothills of the Іli-Alatau. The work on determining the chemical composition of the species Conuim maculatum L. was carried out at the research center of medicinal plants of the al-Farabi Kazakh National University.
Key words: Conuim maculatum L., botany, population, phytochemistry, аnatomy, morphology.
Introduction
Currently, one of the pressing problems facing domestic scientists is the production of medicinal preparations necessary for medicine from plants with medicinal properties, increasing their effectiveness.
Undoubtedly, the medicine made at the expense of medicinal plants has a number of advantages over synthetic ones [1,2]. This is due to the fact that phytopreparations obtained on the basis of medicinal plants are highly effective in the treatment of neglected diseases in the human body and cause minimal damage to the environment. Therefore, it is necessary to turn plants with medicinal properties into the main source of raw materials for the pharmaceutical industry, by integrating scientific work with practical work aimed at the systematic study of medicinal plants and the use of medicinal plants as raw materials for the production of medicines [3]. Inflammation as a pathological process is the most common form of the disease among people, the exacerbation of such diseases, the long-term persistence of symptoms of this disease in the human body leads to a decrease in the ability to work of a person, so the development of drugs to stabilize this problem is one of the most pressing issues today [4-6].
The flora of modern Kazakhstan includes about 6,000
plant species. Among them, more than 1,500 species of plants with healing properties have been registered.
However, only more than 60 medicinal plants are officially included in the State Pharmacopoeia of the Republic of Kazakhstan, but despite this, such medicinal plants as Conuim (C.) maculatum L. require a full-scale systematic study [7]. This plant, which has a small source base, is used only in folk medicine for the treatment of inflammatory diseases, asthma, seizures and other diseases [8].
C. maculatum L. contains biologically active constituents of various chemical structures. Thera- peutic effect of this plant on the human body is very large, as it is a part of the phenolic combined type, and a valuable medicinal plant in the fight against diseases and bacteria.
Recently, the range of application of C. macu- latum L. significantly expanded, and now they are used not only in the treatment of rheumatological diseases, but also for the prevention of thrombosis in immunocompetent diseases and the prevention of the initial stage of atherosclerosis. They can be used in small operations, in the treatment of cardiovascular diseases, in oligomenorrhea, in Alzheimer’s disease, in dementia and oncology. In particular, it is widely used for the prevention of colon cancer [9-11]. For
