학술논문
The Performance of UMAP plus Linkage Compared with Daura-Clustering of Molecular Dynamics of the PD-1 Checkpoint Receptor
Document Type
Conference
Author
Source
2022 IEEE International Conference on Bioinformatics and Biomedicine (BIBM) Bioinformatics and Biomedicine (BIBM), 2022 IEEE International Conference on. :3569-3573 Dec, 2022
Subject
Language
Abstract
Leucocytes check if unknown peptides are loaded in MHC-receptors, and if so, kill the respective cell. However, in some cases, attacks occur erroneously. Cells may guard against such attacks by expressing the ligand (PD-LI) of PD-1, the leucocyte’s self-destroy receptor (checkpoint receptor). This important protection mechanism against autoimmunity is, however, abused by cancer cells, which also express PD-LI and thereby evade appropriate immune clearance. Scrutinizing the molecular binding mechanisms between PD-1 and PD-LI as well as respective drugs (checkpoint receptor blockers) is hence of utmost importance in the development of drugs for cancer therapy. We performed all-atom molecular dynamics simulations for PD 1 to characterize the atomic movements of the CC’-loop without a ligand. Trajectories were analyzed via two clustering algorithms: Daura-clustering, based on RMSD distances was compared with dimension reduction via UMAP (Uniform Manifold Approximation and Projection), followed by linkage clustering (UMAP-Lnk). Daura-clustering with several cutoffs and the correspondence with UMAP-Lnk was evaluated. While the size of Daura-clusters resulted merely from a skillful choice of the cutoff, the number of linkage-clusters was evaluated and optimally selected via the silhouette criterion. Accordance between Daura and UMAP-Lnk was evaluated for several choices of Daura-cutoffs.