Добавил:
kiopkiopkiop18@yandex.ru t.me/Prokururor I Вовсе не секретарь, но почту проверяю Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз: Предмет: Файл:

Ординатура / Хирургия / Библиотека им академика М.И. Перельмана / Книга_5609_Библиотеки_им_академика_М_И_Перельмана

.pdf
Скачиваний:
0
Добавлен:
02.09.2026
Размер:
21 Мб
Скачать
206
https://t.me/medicina_free
with other conditions, has been surging (Cragg and Newman, 2013). More recently, several research institutions have been displaying interest in plant extracts due to advancements in natural product screening techniques, including liquid chroma­tography (LC)/mass spectrometry (MS), LC/ultraviolet spectroscopy, LC/nuclear magnetic resonance spectroscopy, high performance liquid chromatograpy–MS/ MS, high-resolution Fourier transform mass spectrometry, and photo-diode arrays, providing ample structural congurations, elucidation for lead compound identica­tion (Hostettmann et al, 2001)
Drug discovery strategies include phenotypic- and target-based computer-aided drug design approaches, with the target-based strategy involving target identication related to disease state. While these investigations drive potent chemical optimiza­tion of lead compounds, clinical trials may have limitations (Swinney, 2013; Zheng et al, 2013). According to a recent study, invalidated illness targets result in many unsuccessful medication candidates in Phase II and III clinical trials. Computational technologies such as machine learning (ML) tools have been widely used to improve the hit rate in drug development molecules (Jaén-Oltra et al, 2000; Marrero-Ponce et al, 2005).
Singh and colleagues established a Bayesian classication model based on structural ngerprints and physicochemical property descriptors. They applied it to virtually screen an independent data set of 200k molecules, exhibiting that the model can screen top hits of PubChem Bioassay actives with up to 76% accu­racy. Computational approaches such as deep learning and ML have been highly employed in improving drug discovery hit rates for many synthetic and naturally derived drugs (Singh et al, 2012). Ekins and colleagues developed Bayesian mod­els to predict compound activity toward Mycobacterium tuberculosis, then com­putationally screened 82,403 molecules and chose 550 for in vitro testing, yielding 124 actives against Mycobacterium tuberculosis. However, there has been little study on categorization predictions for phenotypic screening of neuroprotective drugs so far (Ekins et al, 2014).
The objective of this book chapter is to provide an overview of multiple natural products and their diverse phytochemical constituent prole against neurotoxicity, oxidative stress, and inammation. In addition, recent tactics of ML and deep learn­ing in natural product screening against neuroprotection have also been emphasized.
NeuroPhytomedicine
11.2 PHYTOCHEMICAL CONSTITUENTS OF NATURAL
PLANT PRODUCTS
The majority of the bioactive components found in different parts (leaves, seeds, owers, roots, and stems) of natural plants are reported to be avonoids, terpenes, polyphenols, and CT, at varying concentrations. Phytochemical compounds display multiple therapeutic mechanisms, including kinase-mediated aberrant signal trans­duction and antioxidant and anti-inammatory stimulation. Flavonoids are classi­ed as secondary metabolites of polyphenolic nature, found mainly in plants, with dietary consumption values. They have a generalized 15-carbon skeletal structure consisting of a heterocyclic pyrane ring connecting (with an embedded O group) two benzene rings, C6-C3-C6. Presently, nearly 6000 types of avonoids have been
Deciphering the Deep Learning and Machine Learning Tactics
https://t.me/medicina_free
identied from several plant species, including proanthocyanins, avones, bioa­vonoids, avonol glycosides, and acylated avonol glycosides (Ullah et al, 2020). Alteration in the structural skeleton affects and denes the structural basis for its related derivatives. They exert several benecial effects in the prevention of cancer, cardiovascular disorders, NDD, and diabetes (Safe et al, 2021). CT are a sub-class of terpenoids, tetraterpenoids, producing organic pigments for plants, algae, and certain fungi. They are further divided into two categories, carotenes (containing hydrocarbons with no O-atoms) and xanthophylls (containing O-atom). They are composed of four terpene units with ten carbon atoms each, providing a total of 40 carbon atoms, with their absorbance wavelengths ranging from 400 to 550 nM. Data from several epidemiological studies suggest that carotenoid consumption in humans and animal models appears to be protective against breast, prostate, head, and neck cancer and Parkinson’s disease (PD) (Tan and Norhaizan, 2019). Terpenes are unsaturated hydrocarbon-containing natural products produced by plants and contain nearly 30,000 compounds. They are further classied by the number of car­bon present, such as monoterpenes, diterpenes, and sesquiterpenes, and are regarded as benecial phytocompounds for their anti-apoptotic, anti-platelet, anti-microbial, and anti-inammatory activities (Ninkuu et al, 2021).
207
11.3 ALZHEIMER’S DISEASE – A CRITICAL PROSPECT OF NEUROTOXICITY
Alzheimer’s disease (AD) is a neurological condition that gradually deteriorates cognition and perception prior to progressing to advanced dementia. The underly­ing pathogenic cause of AD is considered to be neurobrillary tangles (NFTs) and amyloid-β (Aβ) brils. The clinical effectiveness of natural plants for treating neu­ral conditions can be achieved by the evaluation of these two pathological protein aggregates in the brain tissue. Numerous studies suggest that bioactive plants and plant extracts such as curcumin, Ginkgo biloba, and resveratrol are known to pos- sess potent anti-AD activities for their varied constitution of bioactive phytochemicals (Shareena and Kumar, 2022). EGB 761-treated P301S TG micedepicted declining levels of p-tau, enhanced learning retention, spatial memory, mitigated p38, glycogen synthase kinase-3-beta (GSK-3β) signaling, rescued cyclic AMP-response element binding protein phosphorylation, and synaptophysin loss in cultured cortical neurons (Qin et al, 2018). Further, dietary EGb-treated-TF/CRND8 AD mice were shown to mitigate AD pathology via downregulating Aβ aggregate formation and modulation of β-secretase enzyme activity while reducing Aβ -induced proinammatory cyto­kine release, such as tumor necrosis factor alpha (TNF-α), interleukin (IL)-1β, pre- venting Aβ-mediated aberrant microglial activation (Colciaghi et al, 2004). In human SH-SY5Y neuroblastoma cell lines, curcumin suppresses Aβ-mediated tau and GSK­3β phosphorylation at Ser396, Thr231, and Ser9 residues and ameliorates HDAC6 overexpression and Aβ toxicity. However, the neuroprotective impact of curcumin on Aβ-mediated GSK-3β dephosphorylation is not directly associated with oxidative stress. It abolishes Aβ-induced downregulation of Ak strain transforming (AKT) and PDPK1 phosphorylation at Thr308, Ser473, and Ser241, respectively, suggesting that the second response, phosphatidylinositol (3,4,5)-trisphosphate (PIP3), consists of
208 NeuroPhytomedicine
https://t.me/medicina_free
curcumin-protective signaling. Likewise, insulin receptor (IR)/phosphatidylinositol­3-kinase (PI3K) pathway, as a regulatory response signal of PIP3, does not involve Aβ-mediated AKT deactivation (Thr308, Ser473 dephosphorylation). In addition, Aβ expression leads to increased levels of phosphatase and tensin homolog (PTEN), a negative PIP3 regulator, which is suppressed by curcumin treatment, implying that curcumin is effectively Aβ-mediated tau phosphorylation engaging PTEN/AKT/ GSK-3β pathway (Huang et al, 2014). In another study, curcumin, in combination with resveratrol, downregulates ROS generation, ameliorates oxidative stress, pre­vents tau hyperphosphorylation at T181 and T205 residue sites, and exhibits neu­roprotective actions on SH-SY5Y cells from Aβ oligomer damage (Yu et al, 2022). In P301L/rTg4510 mice, cornel iridoid glycoside (CIG) treatment improves memory capabilities, spatial learning, and cognitive impairment, represses synapse, and neuro­nal loss, mitigates brain atrophy, enhances synaptic proteins, preserves cytoskeleton, prevents tau hyperphosphorylation, aggregation in rTg4510 mice’s cerebral tissues. Mechanistically, CIG enhances the PP2A activity, upregulates PP2Ac methylation at Leu309, reduces PP2Ac phosphorylation at Tyr307, and promotes protein expressions of PTPA, PTP1B, and LCMT-1 in the brain (Various plant products and their extract’s potent anti-AD activity is represented in Figure 11.1) (Ma et al, 2019). Novel osmotin
FIGURE 11.1 The schematic diagram displays the therapeutic effect of natural plant prod­ucts, such as curcumin, resveratrol, and osmotin, on the pathogenic AD species: Tau and aβ protein species, preventing NFT hyperphosphorylation, aggregation, aβ-oligomer toxicity, synaptic dysfunctions, neuronal death, interneuronal aβ and Tau species migration, oxidative stress, cytokine release, in addition to suppressing activated microglia and astrocytes.
Deciphering the Deep Learning and Machine Learning Tactics
https://t.me/medicina_free
extract signicantly inhibits Aβ1-42-mediated synaptic decits, memory impairment, BACE-1 expression, synaptotoxicity, and Aβ protein aggregates and accumulation. Moreover, osmotin-treated Aβ1-42-mice attenuates Aβ1-42-mediated tau protein hyperphosphorylation at Ser413 via regulation of aberrant p-GSK-3β (Ser9), p-PI3K, p-AKT (Ser473) phosphorylation, prevents neuronal apoptosis via p53-induced caspase-associated apoptotic pathway and neurodegeneration along with alleviating cellular neurotoxicity (Ali et al, 2015).
209
11.4 COGNITION
A common characteristic of many pathological conditions is cognitive decline or impairment. Memory, learning, thinking, and information processing are pri­marily affected by the physiological changes that result in the body. For their abilities to improve cognition, many plant extracts have been used extensively in conditions like dementia, AD, and PD related to memory loss as the body ages (Eckert, 2010). Angelica gigas (AG), in combination with Bombyx mori silk­worm (SW), notably protects hippocampal neuronal cells from H202-mediated cell death in HT22 mouse cell lines. It recovers Scopolamine-induced cognitive impairment and spatial learning. In specic, it upregulates mRNA and protein levels of phosphorylated p38 and extracellular signal-regulated kinase (ERK)1/2 and reduces Bax/Bcl2 apoptotic index expression (Guo et al, 2021). Gypenoside LXXV, derived from Gynostemma pentaphyllum, substantially alleviated cog­nitive decits in db/db mice and enhanced lipid metabolism and glucose toler­ance. It drastically improved glucose uptake by the brain and markedly elevated p-AKT/total AKT, GLUT4, and PPARγ expressions with a negative impact on p-IRS-1/total IRS-1 levels (Meng et al, 2022). Dihydromyricetin (DHM) decreased protein carbonyl, lipid peroxidation levels, augmented catalase, and superoxide dismutase in the brain tissue in Pb-mediated mice. It mitigated Pb-mediated cell apoptosis, with an evident reduction in cleaved caspase-3, Bax. DHM facilitated AMP-activated protein kinase (AMPK) phosphorylation, pre­vented activation of TLR4, p38, MyD88, and GSK-3, inhibited inammatory cytokines (iNOS, IL-6, TNF-α, and COX-2), Aβ levels, and NF-κB nuclear trans­location, indicating that DHM could enhance cognitive functional prole via prevention of oxidative stress, inammation, and apoptosis (Meng et al, 2022). Salidroside, obtained from Rhodiola rosea, prevents cognitive impairment due to hypoperfusion, and cerebral hypoxia, which was attributed to alteration of PI3K, NF-κB, HIFα, mitogen-activated protein kinase (MAPK), and matrix metalloproteinases (MMPs) signaling pathways. It enhances lipopolysaccharides (LPS)-mediated learning impairment, memory decit, and neuroinammation via modulating the SIRT-1dependent nuclear factor erythroid 2–related factor 2 (NRF-2)/HO-1/NF-κB pathway. Salidroside improves cognitive prole in ortho­pedic surgery-mediated cognitive impairment mice, decreasing the expressions of M1 marker genes (CCL5, CD16, CXCL1, CXCL10, IFN-γ, and TNF-α), ele­vating M2 marker genes (Arg1, TGF-β, YM1, IL-4, and IL-10) and activating AMPK kinase/PPARγ pathway (Pan et al, 2022).
210
https://t.me/medicina_free
NeuroPhytomedicine
11.5 BIOACTIVITIES OF NATURALLY DERIVED PHYTOCONSTITUENTS
11.5.1 PolyPhenols
Numerous polyphenolic compounds are a class of natural products that can exhibit potent anti-tumor, antioxidant, and anti-inammatory properties through modula­tion of ROS levels, ER stress, Ca+2 levels, and autophagy. B16-F10 cells, on treatment with P2Et, (Caesalpinia spinosa extract), mediated ER stress and apoptosis in mela- noma cells. P2Et extract mediated ER stress signaling via (PKR)-like endoplasmic reticulum kinase (PERK) kinase phosphorylation, which can modulate Ca+2 levels and promote anti-tumor responses. It induced expression of ICD-associated DAMPs (ATP, Ecto-CRT, and HMGB1) release in a PERK-dependent manner (Prieto et al,
2019). Dietary supplementation of green tea-derived EGCG rescues synaptic devel-
opment and dendritic defect in CDKL5-KO mice neurons. It restored the spine and brain maturation density, along with attenuating defective spine maturation and PSD95+ puncta cell population (Trovò et al, 2020). EGCG-treated BAEC cells resulted in increased NO production and phosphorylation of AKT, eNOS, and Fyn, depicting that ECGC has potent endothelial-dependent vasodilatory effects mediated by ROS and Fyn-dependent intracellular pathways, which lead to PI3K/AKT, eNOS activation. Mulberry polyphenol extracts (MPE) substantially suppressed phosphor­ylated ERK, RAS, and Β-galactosidase levels and enhanced NO and iNOS synthase levels. NO increased the activation of AMPK and reduced HMG-CoA reductase activity via its phosphorylation. It promoted the association of cyclins to their CDK kinases and hyperphosphorylated RB. MPE prevented K-Ras-mediated A7r5 vascu­lar smooth muscle cell senescence via enhancing AMPK and iNOS-dependent path­ways, suggesting its powerful role in mitigating age-induced atherosclerosis (Chen et al, 2022). In p53/wild-type HCT116 colorectal cancer cells, Artemisia annua L. polyphenols (pKAL) induced late apoptosis, morphological changes in cancer cells, associated with elevated acidic vesicles, loss of Golgi integrity, reduced DNA content via upregulation and downregulation of γ-H2AX/p53/p21/Bak cleavage/ phospho-c-Jun N-terminal kinases (JNK)/p62/MAPK1/LC3B-I axis and AKT/β- catenin/cyclophilin A/GM130, respectively, implying that p-KAL plays an essential survival role in HCT116 cell lines, via p-JNK/p62 signaling and by inhibiting ROS­independent p53-dependent cell death signaling (Jung et al, 2021). Annurca apple polyphenols selectively prevent malondialdehyde (MDA)-MB-231 cell proliferation and viability. APE promoted G2/M phase cell cycle arrest, correlated with p-cdc25C and p27 upregulation, along with p21 decrement. It also reduced the expression of several oncoproteins, including β-catenin, NF-κB, and c-myc, and suppressed AKT activation. Moreover, it lowered Dusp-1 levels, enhanced ROS production, and JNK­c-Jun phosphorylation, portraying the underlying pathways of APE-mediated cell death in MDA-MB-231 TNBC cell lines (Martino et al, 2019).
11.5.2 sAPonin
Hederoside C (HedC), a pentacyclic triterpene saponin isolated from P. Koreana, demonstrated potent anti-cancer properties toward MG63, and U20S osteosarcoma
Deciphering the Deep Learning and Machine Learning Tactics
https://t.me/medicina_free
cell lines wherein HedC prevented proliferative activity of both U20S and MG63 cell lines, and induced apoptosis (downregulation of cleaved poly(ADP-ribose) polymerase (PARP), terminal deoxynucleotidyl transferase dUTP nick end labeling–positive cells, cleaved caspase-9 and -3), in time- and dose-dependent man­ner. HedC-treated cells exhibited elevated p21, p53, Bax, and lowered Bcl-2, whereas HedC-induced apoptosis was followed by reduced STAT3, MAPKs (JNK, ERK1/2, and p38) phosphorylation. In in vivo xenograft mouse model, HedC showed an anti­metastatic effect via suppression of neoplastic migration and invasion, ameliorated and enhanced expressions of p-STAT3, PCNA, and p53, cleaved caspase-3, respec­tively, asserting that HedC has anti-metastatic, anti-tumor potentials (Park et al,
2021). Ginsenoside Ro (GRo) (Panax ginseng’s oleanolic saponin) notably attenuates LPS-mediated lung injury and proinammatory mediators, including IL-6, IL-1β, TNF-α. In addition, it inhibits the binding of LPS488 (uorescence-labeled LPS) to membranes of RAW264.7 macrophages, suppresses MAPK, NF-κB phosphoryla­tion, p65 nuclear translocation, dose-dependently. Moreover, docking and molecular dynamics (MD) simulation analysis suggest that GRo is effectively docked into LPS binding regions of MD2/TLR4 complex with stable binding conformation, implying that GRo serves as a therapeutic anti-inammatory agent (Xu et al, 2022). Findings display that cycloastragenol-treated mice brains were evaluated for their effects against Aβ-mediated neurogenic dysfunctions, mitochondrial apoptosis, and oxida­tive stress. Treatment with cycloastragenol effectively inhibited expressions of MAP kinases such as p-p38, p-JNK, and ERK1/2, reduced inammatory markers (Iba-1, IL-1β, TNF-α, and GFAP), mitigated activated microglia and astrocytes. It improved Aβ-mediated cognitive impairment, regulated anti-apoptotic effects, upregulating and downregulating Bcl-2, Bax, Bim, and Casp-3, respectively, and promoted neu­ronal nuclei (NeuN), p-CREB expressions in the frontal cortex and hippocampal regions of mice’s brains, conrming its neuronal survival effect (Ikram et al, 2021).
211
11.5.3 cArotenoiDs (ct)
CTs are a class of tetraterpenoids, constituting conjugated polyene chains with eight units of C5 isoprenoids. CTs are found in several sources, such as photosynthetic and non-photosynthetic plants, animals, and aquatic sources. With the presence of varied functional groups, number of conjugated double bonds, and cumulative polarities, CT exerts potent ROS scavenging, antioxidant, anti-inammatory, neuroprotective, anti­drug resistance, and anti-cancer properties (Kabir et al, 2022). Carotenoid, identied from Spondias mombin, exhibits potent anti-angiogenic, anti-proliferative activi­ties in 7,12-dimethylbenz[a]-anthracene breast cancer models in female Wistar rats. Carotenoid isolate treatment remarkably downregulates the expressions of MMP-2, HIF-1, VEGF, VEGFR, and EGFR in mammary tumors while upregulating mRNA levels of CHD-1. The binding of astaxanthin produced the DFG-out conformation, 7,7,8,8′-tetrahydro-β, -β-carotene, and β-carotene-15,15′-epoxide to the ATP binding domain, with interaction energies of 8.2, 10.3, and 10.5 kcal/mol, respectively (Metibemu et al, 2021). Synergistic dosing of doxorubicin, in tandem with dietary carotenoid fucoxanthin (FUC), drastically enhances DOX’s cytotoxicity prole and limits the dose of DOX (FR) in DOX-resistant cancer cell lines, including HepG-2/
212
https://t.me/medicina_free
ADR (HCC), MCF-7/ADR (BC) and SKOV-3/ADR (ovarian cancer) by 6.28, 8.42, and 4.56-fold, respectively, FUC elevates DOX accumulation in resistant cells, in con­trast to verapamil, FUC, and DOX together, increase Rho123 build-up, p53 caspases (CASP8, CASP3) activity, decrease the expressions of ABCB1, ABCC1, and ABCG2 along with PXR, GST, and CUP3A4 in resistant cancer cells, respectively (Eid et al,
2020). A low DOX dose with carotenoid substantially promoted anti-proliferation and
cytotoxicity in treated MCF-7 and MDA-MV-231 cell lines, compared with high-dose DOX treatment alone. It caused G0/G1 phase cell cycle arrest, promoted selective ROS-induced apoptosis, prevented mitochondrial dysfunctions, and downregulated protein expressions of p21, p27, p53, Bcl-2, Bax, and cyclin D1, thereby improving cancer therapy. In MDA-MB-231 cell lines, β-carotenes were reported to display sig­nicant antitumoral actions by suppressing cell proliferation, migration, and inva­sion ability. β-carotene promotes apoptosis by interfering with S-phase cell arrest and upregulates 3H-deoxy-D-glucose uptake, while it does not affect either 3H-folic acid or 3H-glutamine uptake and oxidative stress. Β-carotene’s anti-proliferative effect involves JNK intracellular signaling and did not impact cell viability, cell cycle, pro­liferation, and migration capabilities of MCF12-A non-tumoral cell lines, asserting that β-carotene exhibits cancer cell-selective actions (Antunes et al, 2022).
NeuroPhytomedicine
11.5.4 flAvonoiDs
Flavonoids are a group of polyphenolic secondary bioactive metabolites that are found in plants and are commonly consumed as dietary supplements due to their immense therapeutic benets. A citrus avonoid, diosmetin displays a range of therapeutic impacts for its antioxidant, anti-bacterial, and anti-inammatory effects. In DNCB-mediated atopic dermatitis (AD) mouse models, diosmetin remarkably lowers epidermis thickness, dermatitis score, and mast cell population compared with untreated groups. It mitigates the macrophage inltration into AD lesions, with the observed reduction in IL-4, IL-1β, and TNF-α. Further, diosmetin ham- pers nitric oxide production, downregulates iNOS expression, suppresses MAPKs (JNK, p38, ERK1/2), and JAK/STAT3 signaling phosphorylation and activation, respectively, in raw 264.7 mouse macrophage cell line (Lee et al, 2020). Oral dos­ing of setin (FIS) in HFD-fed mice ameliorates HFD-induced cardiac dysregula­tions and metabolic disorders by reducing insulin levels, insulin resistance, body weight, and fasting blood glucose. FIS supplementation substantially prevents the pathology of dyslipidemia in both mouse cardiomyocytes and the heart, triggered by metabolic stress. Moreover, FIS treatment represses HFD-mediated inammatory responses in cardiac tissues via lowered Tnfr-1/Tnfr-2 signaling and downstream target expressions. It promotes a robust reduction in brosis-associated genes, attrib­uting to attenuation of brosis by inactivating TGF-β1/Smads/ERK1/2 signaling (Hu et al, 2020). Dietary avonoid galangin prevents metastatic features such as TPA­mediated HepG2 cell migration and invasion via inhibiting TPA-mediated PKC-δ, PKC-α, phospho-IkBα, c-Jun, C-Fos, and NF-κB, thus showcasing its applications in anti-metastatic clinical therapy. Myricetin portrays potent anti-cancer, antioxi­dant, and anti-inammatory activities, serving as a potential therapeutic agent. In U-86 MG cell lines, myricetin treatment depicts anti-glioblastoma, anti-proliferative
Deciphering the Deep Learning and Machine Learning Tactics
https://t.me/medicina_free
effects, hampering cellular migration and invasion, and PTEN status independently. The cytotoxic actions of myricetin are cell-sensitive, wherein it shows minimal cytotoxicity to normal astrocytes, in contrast to the GBM cell line. Additionally, it suppresses the synthesis of focal adhesions, membrane rufes, lamellipodia, and vasculogenic mimicry and blocks the ROCK2 phosphorylation, cortactin, paxillin, JNK, and PI3K/AKT signaling. Myricetin binds to various kinases and scaffold pro­teins such as 3-phosphoinositide-dependent kinase 1 (PDK1), c-Jun, JNK, vinculin, VE-cadherin, and PI3K catalytical isoforms (p110α-γ) (Zhao et al, 2018).
213
11.6 MACHINE LEARNING AND DEEP LEARNING TACTICS IN NATURAL PRODUCT DRUG DISCOVERY AND DEVELOPMENT
11.6.1 mAchine leArning APProAches
ML involves algorithms analyzing curated data to make decisions and build ef­cient models for different functions. So far, several ML tools and development kits have been utilized, including TensorFlow, Random Forest, Waikato Environment for Knowledge Analysis, Scikit-Learn, and Support vector machine. The benets of using ML tools in a quantitative structure activity relationship (QSAR) study include an accurate prediction of the inuence of a compound’s chemical structure on its bioactivity and the modeling of drug toxicity and metabolism (Stitou et al, 2019). Therefore, ML is considered an essential resource for discovering new drugs (Dara et al, 2022).
Yang et al reported developing a neuroprotective ML model for identify­ing potent neuroprotective compounds in Xiaxuming decoction (XXMD) against H202-mediated and Hypoxia-induced brain cell damage. The TCM prescription Xiaoxuming decoction (XXMD) (made up of 12 herbs) has been a benecial remedy for the therapy of stroke in clinical settings, as documented in the Tang Dynasty’s Beiji Qianjin Yaofang. They constructed stacked naïve Bayesian models based on molecular ngerprint descriptors and four distinct single classiers (AB, CT, kNN, and RF). The nalized models were used for the virtual screening of neuroprotective agents in XXMD and further selected for cell-based assay. Among the compounds, two compounds substantially inhibited Na2S2O4-induced and H2O2-mediated neuro­toxicity (Yang et al, 2019). Baicalein was discovered to be a potent neuroprotective drug in this work, with effects on hypoxia and oxidative stress phenotypes. Baicalein has gained signicant focus for its antioxidant and anti-inammatory effects as a bioactive phenolic avonoid molecule. The validation of baicalein against two dam­age phenotypes suggests that it might be used as a promising neuroprotective agent (Dinda et al, 2017).
Another study by Fang J and colleagues reported 28 compounds exhibiting neu­roprotective properties against H2O2-mediated and monosodium glutamate-induced neurotoxicity by ML approaches. The results suggested that techniques that inte­grated single classiers into combined Bayesian models could be a viable approach to predicting neuroprotective compounds. On monosodium glutamate-induced and H2O2-induced PC12 cells, three selected molecules (J14572, J27152, and J27114)
214
https://t.me/medicina_free
demonstrate a favorable dose-response relation and cell viability. Cell survival for model groups affected by 40 mM monosodium glutamate or 300 mM H2O2 was signicantly lower than in the control group (P 0.01) (Fang et al, 2016).
NeuroPhytomedicine
11.6.2 DeeP leArning APProAches
Deep learning is a sub-eld of ML that utilizes articial neural cells to process data in decision-making. It also applies to the drug discovery process besides computa­tional elds such as pattern recognition. It is highly implemented in QSAR stud­ies; virtual screening; absorption, distribution, metabolism, excretion, and toxicity properties; and lead optimization. Earlier deep learning implementations propose a wide range of disease detection and prediction screening models, including a Densely connected convolutional neural network (CNN)-based automated COVID screening model, drug-target interaction CNN models, and CNN RF models to ana­lyze drug functions from chemical structure (Chakravarti and Alla, 2019; Kumari and Subbarao, 2021). This novel approach has led to the discovery of several impact­ful techniques that could be effectively utilized in phytochemical drug screening and their therapeutic and pharmacological targets in several chronic states, such as neuroprotection, neurotoxicity, apoptosis, oxidative stress, and inammation. Hence, several studies surrounding such have been described, showing a vast diversity of models throughout its development phases (Shanmuganathan, 2016).
The algorithm for all the naturally derived drugs and compounds functions in three steps: (A) Collection of natural compounds and drug information from the public database. (B). Producing molecular interaction and chemical features from collected data by text mining, network analysis, and chemical property analysis. (C) Training the deep learning models on the elements of approved drugs and predicting the medicinal properties based on the trained model. Deep learning can increase prediction performance when the input characteristics are extensive and diversi­ed by extracting high-level interpretation using low-level features. The mentioned approach has four consecutive levels: Input, partially linked hidden layers, fully coupled hidden layers, and output. The models are developed to forecast the likely effects list utilizing the input attributes. Theoretical knowledge, molecular interac­tion, and chemical property data are generated for each drug or natural component and used as inputs to the model. Hidden layers extended their outputs by detecting nonlinear connections between low- and high-level data to provide a higher-level representation than the last layer (Yoo et al, 2020).
Rodríguez FR and colleagues developed a computational strategy involving deep learning, structural bioinformatics, and signaling pathway manual reconstruction to predict eight novel nicotine analogs’ neuroprotective activity based on the PI3KAKT pathway in the context of PD therapeutics. The model predicted the potential neu­roprotective efcacy of seven novel nicotine analogs based on the binomial Bcl-2 response regulated by PI3K/AKT activation (Rojas-Rodríguez et al, 2020).
Further studies by Wang H and co-workers identied sclareol as a naturally derived Cav1.3-neuroprotective antagonist in PD. While L-type voltage-gated cal­cium channel blockers treatment selectively alleviates Cav1.3 for PD, drug develop­ment is impeded due to a shortfall in high-throughput screening techniques allowing
215Deciphering the Deep Learning and Machine Learning Tactics
https://t.me/medicina_free
FIGURE 11.2 The above illustration depicts a contrasting feature between ML approaches and deep learning techniques, utilized in drug discovery and development, virtual screening, and SAR study of phytochemicals and natural products.
isoform-specic assessment of Cav-antagonistic activities. Integrated in silico vir­tual screening and deep learning models enabled the discovery of (6)-Gingerol and sclareol as novel Cav1.3 antagonists. In silico analysis of 198 candidate molecules produced 14 hits as the most promising Cav1.3 inhibitors.
Moreover, structure clustering analysis enabled a selection of ve phytoconstitu­ents as potential structures. In addition, parallel articial intelligence–based valida­tion showcased similar results, with a receiver operating characteristic curve–area under curve value of 97.78%. Experimentations on the ve compounds with CaB-A assay conrmed that sclareol and (6)-gingerol portrayed robust inhibition on Cav1.3. Both molecules showed a more substantial antagonistic effect on Cav1.3-induced reporter gene expression than the Cav1.3-dependent CaB system. Sclareol-treated PD mouse models observed minimal loss in DA neurons, exhibited in vivo neu­roprotective effects toward 6-ODHA-mediated neurodegeneration, prevented over­synchronization and locomotion decits of striatal neurons (the contrast between ML and deep learning approaches are described in Figu re 11.2) (Wang et al, 2022).
11.7 CONCLUSION
The emergence of expanding disease conditions and the shortage of disease­modulating or disease-alleviating therapies pose alarming challenges to research institutions. Although diagnosing and treating the symptomatic events of chronic