figshare
Browse
pcbi.1006093.g003.tif (2.16 MB)

Defining the glycosylation site determinants that shape bNAb binding profiles.

Download (2.16 MB)
figure
posted on 2018-04-20, 17:33 authored by Wen-Han Yu, Peng Zhao, Monia Draghi, Claudia Arevalo, Christina B. Karsten, Todd J. Suscovich, Bronwyn Gunn, Hendrik Streeck, Abraham L. Brass, Michael Tiemeyer, Michael Seaman, John R. Mascola, Lance Wells, Douglas A. Lauffenburger, Galit Alter

(a)-(d) Four different Bayesian MCMC-SVR models were evaluated for their respective abilities to predict PGT121 binding to the 94 proteins. The models include a Bayesian MCMC-SVR model based on: (a) sequon presence (Figure A in S1 File); (b) protein sequence; (c) glycosylation site occupancy; or (d) glycosylation site occupancy and sequence combined. Cross validation (100-iterative 10-fold) was used to evaluate model performance. Goodness-of-fit was assessed and is reported as the mean squared error (MSE) between predicted and ELISA-measured binding. (e) Heat map shows the binding signatures of individual Abs (rows), where the selected glycan sites (determinants) that mediate effects on Ab binding are highlighted. NAbs that share similar glycan determinants are grouped by hierarchical clustering. (f) The significant glycan site determinants for PGT121, PGT128, and VRC01 are plotted onto a 3-dimensional gp120 monomer structure using the same directional color coding as the heat-map. Additionally, the critical protein residues predicted by our model are shaded in yellow on the same 3D structure. Finally, broad Ab-binding sites were highlighted for each bNAb in hatched circles. (g) Agonistic and antagonistic glycan site determinants and critical protein residues for PGT122 are projected on the BG505 SOSIP.664-PGT122 co-crystal structure (PDB #: 4NCO) with the same color coding. The V3 loop is highlighted in light green shading.

History