Camelid single-domain antibody fragments (“nanobodies”) provide the remarkable specificity of antibodies within a single 15 kDa immunoglobulin VHH domain. This unique feature has enabled applications ranging from use as biochemical tools to therapeutic agents. Nanobodies have emerged as especially useful tools in protein structural biology, facilitating studies of conformationally dynamic proteins such as G protein-coupled receptors (GPCRs). Nearly all nanobodies available to date have been obtained by animal immunization, a bottleneck restricting many applications of this technology. To solve this problem, we report a fully in vitro platform for nanobody discovery based on yeast surface display. We provide a blueprint for identifying nanobodies, demonstrate the utility of the library by crystallizing a nanobody with its antigen, and most importantly, we utilize the platform to discover conformationally-selective nanobodies to two distinct human GPCRs. To facilitate broad deployment of this platform, the library and associated protocols are freely available for non-profit research.

The ability to design functional sequences and predict effects of variation is central to protein engineering and biotherapeutics. State-of-art computational methods rely on models that leverage evolutionary information but are inadequate for important applications where multiple sequence alignments are not robust. Such applications include the prediction of variant effects of indels, disordered proteins, and the design of proteins such as antibodies due to the highly variable complementarity determining regions. We introduce a deep generative model adapted from natural language processing for prediction and design of diverse functional sequences without the need for alignments. The model performs state-of-art prediction of missense and indel effects and we successfully design and test a diverse 105-nanobody library that shows better expression than a 1000-fold larger synthetic library. Our results demonstrate the power of the alignment-free autoregressive model in generalizing to regions of sequence space traditionally considered beyond the reach of prediction and design.