Machine learning optimization of candidate antibody yields highly diverse sub-nanomolar affinity antibody libraries.

Therapeutic antibodies are an important and rapidly growing drug modality. However, the design and discovery of early-stage antibody therapeutics remain a time and cost-intensive endeavor. Here we present an end-to-end Bayesian, language model-based method for designing large and diverse libraries of high-affinity single-chain variable fragments (scFvs) that are then empirically measured. In a head-to-head comparison with a directed evolution approach, we show that the best scFv generated from our method represents a 28.7-fold improvement in binding over the best scFv from the directed evolution. Additionally, 99% of designed scFvs in our most successful library are improvements over the initial candidate scFv. By comparing a library's predicted success to actual measurements, we demonstrate our method's ability to explore tradeoffs between library success and diversity. Results of our work highlight the significant impact machine learning models can have on scFv development. We expect our method to be broadly applicable and provide value to other protein engineering tasks.

A dataset comprised of binding interactions for 104,972 antibodies against a SARS-CoV-2 peptide.

The dataset presented here contains quantitative binding scores of scFv-format antibodies against a SARS-CoV-2 target peptide collected via an AlphaSeq assay that can be used in the development and benchmarking of machine learning models. Starting from three seed sequences identified from a phage display campaign using a human naïve library, four sets of 29,900 antibodies were designed in silico by creating all k = 1 mutations and random k = 2 and k = 3 mutations throughout the complementary-determining regions (CDRs). Of the 119,600 designs, 104,972 were successfully built in to the AlphaSeq library and target binding was subsequently measured with 71,384 designs resulting in a predicted affinity value for at least one of the triplicate measurements. Data include antibodies with predicted affinity measurements ranging from 37 pM to 22 mM. To our knowledge, this dataset is the largest, publicly available dataset that contains antibody sequences, antigen sequence and quantitative measurements of binding scores and provides an opportunity to serve as a benchmark to evaluate antibody-specific representation models for machine learning.

Massively multiplexed affinity characterization of therapeutic antibodies against SARS-CoV-2 variants.

Antibody therapies represent a valuable tool to reduce COVID-19 deaths and hospitalizations. Multiple antibody candidates have been granted emergency use authorization by the Food and Drug Administration and many more are in clinical trials. Most antibody therapies for COVID-19 are engineered to bind to the receptor-binding domain (RBD) of the SARS-CoV-2 Spike protein and disrupt its interaction with angiotensin-converting enzyme 2 (ACE2). Notably, several SARS-CoV-2 strains have accrued mutations throughout the RBD that improve ACE2 binding affinity, enhance viral transmission and escape some existing antibody therapies. Here, we measure the binding affinity of 33 therapeutic antibodies against a large panel of SARS-CoV-2 variants and related strains of clinical significance using AlphaSeq, a high-throughput yeast mating-based assay to determine epitopic residues, determine which mutations result in loss of binding and predict how future RBD variants may impact antibody efficacy.

Defective nucleotide-dependent assembly and membrane fusion in Mfn2 CMT2A variants improved by Bax.

Mitofusins are members of the dynamin-related protein family of large GTPases that harness the energy from nucleotide hydrolysis to remodel membranes. Mitofusins possess four structural domains, including a GTPase domain, two extended helical bundles (HB1 and HB2), and a transmembrane region. We have characterized four Charcot-Marie-Tooth type 2A-associated variants with amino acid substitutions in Mfn2 that are proximal to the hinge that connects HB1 and HB2. A functional defect was not apparent in cells as the mitochondrial morphology of Mfn2-null cells was restored by expression of any of these variants. However, a significant fusion deficiency was observed in vitro, which was improved by the addition of crude cytosol extract or soluble Bax. All four variants had reduced nucleotide-dependent assembly in cis, but not trans, and this was also improved by the addition of Bax. Together, our data demonstrate an important role for this region in Mfn2 GTP-dependent oligomerization and membrane fusion and is consistent with a model where cytosolic factors such as Bax are masking molecular defects associated with Mfn2 disease variants in cells.

A catalytic domain variant of mitofusin requiring a wildtype paralog for function uncouples mitochondrial outer-membrane tethering and fusion.

Mitofusins (Mfns) are dynamin-related GTPases that mediate mitochondrial outer-membrane fusion, a process that is required for mitochondrial and cellular health. In Mfn1 and Mfn2 paralogs, a conserved phenylalanine (Phe-202 (Mfn1) and Phe-223 (Mfn2)) located in the GTPase domain on a conserved ß strand is part of an aromatic network in the core of this domain. To gain insight into the poorly understood mechanism of Mfn-mediated membrane fusion, here we characterize a Mitofusin mutant variant etiologically linked to Charcot-Marie-Tooth syndrome. From analysis of mitochondrial structure in cells and mitochondrial fusion in vitro, we found that conversion of Phe-202 to leucine in either Mfn1 or Mfn2 diminishes the fusion activity of heterotypic complexes with both Mfn1 and Mfn2 and abolishes fusion activity of homotypic complexes. Using coimmunoprecipitation and native gel analysis, we further dissect the steps of mitochondrial fusion and demonstrate that the mutant variant has normal tethering activity but impaired higher-order nucleotide-dependent assembly. The defective coupling of tethering to membrane fusion observed here suggests that nucleotide-dependent self-assembly of Mitofusin is required after tethering to promote membrane fusion.