Fig. 4: Characterizing a single stem antibody within head + stem mixtures. | Nature Computational Science

Fig. 4: Characterizing a single stem antibody within head + stem mixtures.

From: Harnessing low dimensionality to visualize the antibody–virus landscape for influenza

Fig. 4

a, We created mixtures of two or three antibodies targeting the HA head and stem. Using a mixture’s neutralization titers, we predict the behavior of the stem antibody within. b, The virus–virus distance on the landscape constrains how differently any stem antibody can neutralize two viruses. c, Left: the neutralization from a head + stem mixture (#7 in Supplementary Fig. 10) cannot be described by any point on the landscape due to the head antibody. The gold disk represents a lower bound on neutralization (for example, IC50 > 10−7 M, d > 3); an antibody should lie outside all gold disks but on every red circle (representing exact IC50 values). Right: after using equation (2) to remove the neutralization from the head antibody, the IC50 values of all blue viruses become lower bounds (gold disks), and we can infer the position of the stem antibody (red). The insets at the top left show the number of head (brown) and stem (gray) antibodies in each mixture, and whether the head signal has been removed (antibody icon crossed out). d, Examples showing additional mixtures combining one stem antibody with one or two head antibodies (additional decompositions are shown in Supplementary Fig. 10). Average error quantifies the fold-difference between the predicted antibody’s IC50 values and measurements.

Source data

Back to article page