Figure 1.
Comparison of human and murine sequences.
Histograms of statistical sequence scores calculated for human (red) and murine (grey) sequences. Left: Heavy chains (F). Right: Light chains (F
). Scores for Rituximab, a chimeric antibody, and Avastin(Bevacizumab), a humanized antibody are indicated by green lines/stars.
Figure 2.
Sequence analysis of Rituximab-VH.
Contributions to the total score are mapped onto individual residues. Yellow and red colors indicate that this amino acid is uncommon in this position and/or shows unfavourable couplings with other positions. Green shaded areas indicate CDRs of which some residues are not taken into account. Upper graph: Ritxumab heavy chain with human heavy chain potential (F). Lower graph: Rituximab heavy chain with murine heavy chain potential (F
).
Figure 3.
Stochastic Humanization of Rituximab-VH.
Humanization of a murine heavy chain can be treated as an optimization of the objective function F. The Monte Carlo algorithm starts from the sequence of Rituximab-VH and optimizes the scoring function and the sequence identity to the parental sequence simultaneously.
Figure 4.
Distribution of Epivax Scores for human and murine heavy chain sequences of 500 randomly picked sequences from the Abysis data set are shown. Low Scores indicate a lower risk of immunogenicity.
Figure 5.
Development of Epivax Scores during stochastic humanization of Rituximab.
A) Light Chain. B) Heavy Chain. Color codes indicate progress of the simulation, starting from blue to red. Both trajectories sample sequences with low immunogenic potential as predicted by the Epivax score.
Figure 6.
Couplings are required to generate non-immunogenic sequences.
The plot shows the development of the Epivax score over the course of a Monte Carlo optimization of Rituximab-VH. The scoring function used here does not consider couplings of sequence positions and as a consequence, no sequences with low immunogenic potential are sampled.
Figure 7.
Influence of couplings on generated sequences.
The histograms show the distributions of Epivax scores for ensembles of sequences generated with different influence of positional couplings (). The green shaded area marks the desired region for non-immunogenic sequences.
Figure 8.
Sequence processing is shown for heavy (A) and light chain (B) of Rituximab as an example. CDR residues marked in green belong to structurally varying parts and are not considered in the analysis.