Table 6 List of studies showing enhanced affinity maturation strategies based on deep sequencing big data assessments
Study | Findings | Limitations |
|---|---|---|
Hu et al. [60] | The deep sequencing of affinity maturation Ab libraries allowed for an in-depth evaluation of the enrichment landscapes in CDR sequences and amino acid substitutions. Here, potent affinity candidates were identified according to their high frequencies, by-passing the need for experimental primary screenings | ▪ Uses mutagenesis libraries of single CDRs (either L1, L3, and H1–H3) ▪ For generating high affinity clones it requires the development of combinatorial sub-libraries to explore the synergistic effects of the different CDRs |
Reddy et al. [61] | Deep sequencing and bioinformatic analyses were used to mine Ab variable region repertoires from plasma cells of immunized mice. Following, the pairing of the most abundant variable heavy (VH) and variable light (VL) genes, based on their relative frequencies, produced nanomolar affinity Abs | â–ª Based on natural Ab repertoires derived from animal immunizations â–ª Requires further Ab reconstructions, based on pairings of VH and VL genes and subsequent in vitro screenings |
Pan et al. [62] | Mature B cell repertoires from immunized mice were used to generate yeast display Ab libraries. After in vitro selections followed by deep sequencing analysis the authors were able to elucidate the affinity and molecular features of antigen-specific clonal lineages | â–ª Based on natural Abs from bone marrow and spleen â–ª Uses Ab transgenic immunized mice â–ª Ab lineages based only on CDR-H3 clustering |
Fujino et al. [70] | A mutational scanning Ab library was subjected to in vitro selections followed by deep sequencing analysis. This allowed the identification of all the critical residues associated with antigen binding. Subsequently, based on the identified CDR hotspots, a combinatorial sub-library was constructed to generate high affinity clones | ▪ Uses the Ab sequence of a previously identified Ab ▪ Mutagenesis only focuses on 50 amino acid positions comprising all CDR loops ▪ Requires the generation of sub-libraries to identify high affinity Abs |
Saka et al. [71] | A machine learning model was capable to decipher the distinct CDR patterns associated with high-affinity antigen binding interactions in deep sequencing datasets. This model was able to sample virtual sequences and avoid combinatorial features commonly encountered in the deep sequencing space. The in silico predicted clones had unique sequences with exceptionally improved affinities | â–ª Uses a synthetic heavy chain library with limited sequence diversity â–ª Requires Ab-antigen in vitro experimentations â–ª The deep sequencing training data contained non-specific binder sequences, and true binder sequences were never determined |
Liu et al. [72] | A deep generative model, based on Ab sequence generation and prioritization, was trained on deep sequencing big data to effectively generate Ab sequences with higher affinity interactions. Its predictions can be generalized to produce de novo Ab sequences with improved affinities | ▪ Clone predictions were based on computing the minimal sets of specific CDR-H3 amino acids required for binding ▪ Requires primary Ab campaigns and subsequent affinity maturation steps to achieve desired affinities |