Deep sequencing in library selection projects: what insight does it bring?

Abstract

High throughput sequencing is poised to change all aspects of the way antibodies and other binders are discovered and engineered. Millions of available sequence reads provide an unprecedented sampling depth able to guide the design and construction of effective, high quality naïve libraries containing tens of billions of unique molecules. Furthermore, during selections, high throughput sequencing enables quantitative tracing of enriched clones and position-specific guidance to amino acid variation under positive selection during antibody engineering. Successful application of the technologies relies on specific PCR reagent design, correct sequencing platform selection, and effective use of computational tools and statistical measures to remove error, identify antibodies, estimate diversity, and extract signatures of selection from the clone down to individual structural positions. Here we review these considerations and discuss some of the remaining challenges to the widespread adoption of the technology.

Figure 1

NGS sequencing on scFv genes. Variability plots for representative VL and VH genes are shown, with the CDRs shaded in grey. Length coverage for the most popular NGS platforms and scFv-based libraries targeted regions are shown. For each platform, single or double directional arrows indicate single or paired-end sequencing, respectively.

Figure 2

A) Schematic representation of NGS barcoded primers. Primers on conserved regions flanking the diversity carry barcode sequencing and NGS-specific adapters. The PCR product contains a unique identifier (barcode) for a specific library. Multiple amplicon libraries can be pooled in a unique sample and sequenced together. Each single sequenced DNA fragment is associated to a specific library based on its barcode. B) An amplicon library can also be generated with the inclusion of a Unique Molecular Identifier (UMI) sequence, which are added at an initial step (e.g., first-strand cDNA synthesis) resulting in each molecule of a library being tagged with a UMI. Similar to A), amplicons can be sequenced in a multiplex fashion but following NGS, sequences with identical UMIs are grouped together for consensus building-based error correction.

Figure 3

Estimating upper and lower diversity bounds as a function of sequencing depth. Maximum theoretical diversity is the total number of unique molecules that could exist in a library of this design if the number of transformants were infinite. Maximum transformant diversity is the maximum library size if every molecule in the library was non-redundant. Minimum observed diversity is the accumulated diversity observed from sequencing: the number of different clones actually seen. Minimum diversity estimated by capture-recapture methods more rapidly approaches the true diversity of the library, by anticipating library diversity from subsample overlap. Maximum diversity can be calculated by extracting known library redundancy from the transformation size. Actual diversity is the number of unique clones in the library. All measures convergence on true diversity with increasing sampling depth, although libraries with “long tails” of rare clones will converge slowly.

Figure 4

Relative abundance and affinity. Panel A: the experimentally measured kDs (nM) of selected clones are plotted in relation to their ranking position in the sequenced selection output (i.e. the clone in ranking position 1 has the highest relative abundance in the selection output). Average affinity of the clones (black) and antigen concentration used in the selection process (red) are shown as dotted lines. Data collected for three different antigen selections (ESAT6, Antigen85, and Ubiquitin) are reported. Panel B: ranking and affinity plots are shown for anti CDK2 selection at different antigen concentrations. The affinities of identical clones (identified by the same symbol in the 2 plots) found in sequenced populations selected at different antigen concentrations are shown in relation to their ranking position. P-values for significant correlation are reported.