Computational and experimental characterization of estrogenic activities of 20(S, R)-protopanaxadiol and 20(S, R)-protopanaxatriol

Abstract

Background As the main metabolites of ginsenosides, 20( S , R )-protopanaxadiol [PPD( S , R )] and 20( S , R )-protopanaxatriol [PPT( S , R )] are the structural basis response to a series of pharmacological effects of their parent components. Although the estrogenicity of several ginsenosides has been confirmed, however, the underlying mechanisms of their estrogenic effects are still largely unclear. In this work, PPD( S , R ) and PPT( S , R ) were assessed for their ability to bind and activate human estrogen receptor α (hERα) by a combination of in vitro and in silico analysis. Methods The recombinant hERα ligand-binding domain (hERα-LBD) was expressed in E. coli strain. The direct binding interactions of ginsenosides with hERα-LBD and their ERα agonistic potency were investigated by fluorescence polarization and reporter gene assays, respectively. Then, molecular dynamics simulations were carried out to simulate the binding modes between ginsenosides and hERα-LBD to reveal the structural basis for their agonist activities toward receptor. Results Fluorescence polarization assay revealed that PPD( S , R ) and PPT( S , R ) could bind to hERα-LBD with moderate affinities. In the dual luciferase reporter assay using transiently transfected MCF-7 cells, PPD( S , R ) and PPT( S , R ) acted as agonists of hERα. Molecular docking results showed that these ginsenosides adopted an agonist conformation in the flexible hydrophobic ligand-binding pocket. The stereostructure of C-20 hydroxyl group and the presence of C-6 hydroxyl group exerted significant influence on the hydrogen bond network and steric hindrance, respectively. Conclusion This work may provide insight into the chemical and pharmacological screening of novel therapeutic agents from ginsenosides.

Taking a Look at Small-Scale Pedestrians and Occluded Pedestrians

“…In this section, some experiments on two famous pedestrian datasets (i.e., the Citypersons dataset [54] and the Caltech pedestrian dataset [15]) are conducted to demonstrate the effectiveness of proposed methods and compare with some state-of-the-art methods.…”

“…With the success of Faster RCNN [43] on general object detection, many end-to-end CNN based methods have been also proposed on pedestrian detection. For example, Zhang et al [54] proposed AdaptedRCNN by modifying some network settings to better detect pedestrians. Mao et al [37] proposed to add semantic segmentation task to help improve pedestrian detection.…”

“…In fact, the visual information on both small-scale pedestrians and occluded pedestrians is not as adequate as the normal pedestrians, so that it is difficult to accurately detect them. By loosening the evaluation metric between bounding boxes and ground-truth from 0.5 to 0.3, the detection performance of many small-scale pedestrians and occluded pedestrians on Citypersons [54] both have about 8% improvement. It means that to some degree many small-scale pedestrians and occluded pedestrians can be found but not accurately located.…”

“…Meanwhile, small-scale pedestrians and occluded pedestrians both lack visual information in some degree such that accurately localizing them becomes problematic. When the evaluation IoU threshold on Citypersons [54] varies from 0.5 to 0.3, we find that the miss rates of FPN on the small-scale subset and occlusion subset of Citypersons have 8.8% and 7.6% drops. It means that many small-scale pedestrians and occluded pedestrians can be found but not accurately located, which is called inaccurate location problem.…”

“…(3) Experiments on the Citypersons [54] and Caltech [15] datasets show the effectiveness of proposed methods. Meanwhile, the proposed method does not add much computational costs.…”