Development of Scoring Functions for Antibody Sequence Assessment and Optimization

Abstract

Antibody development is still associated with substantial risks and difficulties as single mutations can radically change molecule properties like thermodynamic stability, solubility or viscosity. Since antibody generation methodologies cannot select and optimize for molecule properties which are important for biotechnological applications, careful sequence analysis and optimization is necessary to develop antibodies that fulfil the ambitious requirements of future drugs. While efforts to grab the physical principles of undesired molecule properties from the very bottom are becoming increasingly powerful, the wealth of publically available antibody sequences provides an alternative way to develop early assessment strategies for antibodies using a statistical approach which is the objective of this paper. Here, publically available sequences were used to develop heuristic potentials for the framework regions of heavy and light chains of antibodies of human and murine origin. The potentials take into account position dependent probabilities of individual amino acids but also conditional probabilities which are inevitable for sequence assessment and optimization. It is shown that the potentials derived from human sequences clearly distinguish between human sequences and sequences from mice and, hence, can be used as a measure of humaness which compares a given sequence with the phenotypic pool of human sequences instead of comparing sequence identities to germline genes. Following this line, it is demonstrated that, using the developed potentials, humanization of an antibody can be described as a simple mathematical optimization problem and that the in-silico generated framework variants closely resemble native sequences in terms of predicted immunogenicity.

Antibody Structure and Function: The Basis for Engineering Therapeutics

“…The humanness scores that are based on pairwise sequence identity between the sample and a set of germline human sequences may consider the average similarity [156], or the average among the top 20 sequences [157], or the highest similarity over windows of 9 residues [98,155]. In a different approach [158], the score function accounts both for local preferences and for pair correlations between residues at different positions. The method does not distinguish CDRs from FRs, which may be a plus since the latter may contain antigen-binding residues.…”

Engineering Stability, Viscosity, and Immunogenicity of Antibodies by Computational Design

“…Seeliger 192 has expanded on the use of simple pairwise sequence comparisons to derive sequence-based statistical potentials using 11,849 antibody sequences of humans and mice obtained from the abYsis database. 202 Instead of simply computing sequence identities between a given sequence and sequences of human antibodies, the author incorporated position-specific probabilities of individual amino acids derived from a multiple-sequence alignment of each chain type (i.e., heavy, k-light, or l-light chains of humans and mice); the resulting potentials were able to distinguish between human and mouse antibodies.…”

Boosting antibody developability through rational sequence optimization

“…23 Subsequently, we applied a statistical sequence analysis method described in the literature. 30 In brief, the antibody sequence was evaluated by scoring functions developed from the entire pool of antibody sequences available in public databases. This allows for a statistical assessment of each individual position in the sequence.…”