Regulatory approved monoclonal antibodies contain framework mutations predicted from human antibody repertoires

Abstract

Monoclonal antibodies (mAbs) are an important class of therapeutics used to treat cancer, inflammation, and infectious diseases. Identifying highly developable mAb sequences in silico could greatly reduce the time and cost required for therapeutic mAb development. Here, we present position-specific scoring matrices (PSSMs) for antibody framework mutations developed using natural human antibody repertoire sequences. Our analysis shows that natural human antibody repertoire-based PSSMs are consistent across individuals and demonstrate high correlations between related germlines. We show that mutations in existing therapeutic antibodies can be accurately predicted solely from natural human antibody sequence data. mAbs developed using humanized mice had more human-like FR mutations than mAbs originally developed by hybridoma technology. A quantitative assessment of entire framework regions of therapeutic antibodies revealed that there may be potential for improving the properties of existing therapeutic antibodies by incorporating additional mutations of high frequency in natural human antibody repertoires. In addition, high frequency mutations in natural human antibody repertoires were predicted in silico to reduce immunogenicity in therapeutic mAbs due to the removal of T cell epitopes. Several therapeutic mAbs were identified to have common, universally high-scoring framework mutations, and molecular dynamics simulations revealed the mechanistic basis for the evolutionary selection of these mutations. Our results suggest that natural human antibody repertoires may be useful as predictive tools to guide mAb development in the future.

Hallucinating structure-conditioned antibody libraries for target-specific binders

“…Figure 2E shows the distribution of net FR scores per design (summed over 12 design positions) for 40 designs and the net FR score for the wildtype for comparison. These data show that hallucinated sequences have mutations that are preferentially accumulated in human repertoires during antibody maturation, suggesting that hallucinated antibodies are more human-like and less immunogenic than the starting antibody (37).…”

“…To evaluate the human-likeness of the hallucinated designs, we turned to the work of Petersen et al (37), which investigated the amino acid preference for framework mutations in human repertoires and FDA approved therapeutic antibodies. Petersen et al (37) calculated the frequency of amino acids in human repertoires at framework positions for 25 VH genes that represent the precursors of several FDA approved antibodies and many of the most common VH genes. They converted these frequencies into "FR scores", a measure of amino acid enrichment over the germline residue at each framework position.…”

“…They converted these frequencies into "FR scores", a measure of amino acid enrichment over the germline residue at each framework position. They found that mutations with higher FR scores result in lower immunogenicity and are highly enriched in FDA approved antibodies (37). Hence, we calculated the FR scores to evaluate the humanness of hallucination-designed mutations at the V H -V L interface.…”

Progress and challenges for the machine learning-based design of fit-for-purpose monoclonal antibodies

“…309 A recent study envisaged that human antibody repertoires can be a useful predictive tool for mAb development. 310 They analyzed the AA substitutions in mAbs using position-specific scoring matrices (PSSMs) and observed that positions with high frequency of AA alteration may potentially reduce immunogenicity and improve other developability parameters.…”