Machine learning to predict continuous protein properties from binary cell sorting data and map unseen sequence space.

Proteins are a diverse class of biomolecules responsible for wide-ranging cellular functions, from catalyzing reactions to recognizing pathogens. The ability to evolve proteins rapidly and inexpensively toward improved properties is a common objective for protein engineers. Powerful high-throughput methods like fluorescent activated cell sorting and next-generation sequencing have dramatically improved directed evolution experiments. However, it is unclear how to best leverage these data to characterize protein fitness landscapes more completely and identify lead candidates. In this work, we develop a simple yet powerful framework to improve protein optimization by predicting continuous protein properties from simple directed evolution experiments using interpretable, linear machine learning models. Importantly, we find that these models, which use data from simple but imprecise experimental estimates of protein fitness, have predictive capabilities that approach more precise but expensive data. Evaluated across five diverse protein engineering tasks, continuous properties are consistently predicted from readily available deep sequencing data, demonstrating that protein fitness space can be reasonably well modeled by linear relationships among sequence mutations. To prospectively test the utility of this approach, we generated a library of stapled peptides and applied the framework to predict affinity and specificity from simple cell sorting data. We then coupled integer linear programming, a method to optimize protein fitness from linear weights, with mutation scores from machine learning to identify variants in unseen sequence space that have improved and co-optimal properties. This approach represents a versatile tool for improved analysis and identification of protein variants across many domains of protein engineering.

Position-Specific Enrichment Ratio Matrix scores predict antibody variant properties from deep sequencing data.

Motivation: Deep sequencing of antibody and related protein libraries after phage or yeast-surface display sorting is widely used to identify variants with increased affinity, specificity and/or improvements in key biophysical properties. Conventional approaches for identifying optimal variants typically use the frequencies of observation in enriched libraries or the corresponding enrichment ratios. However, these approaches disregard the vast majority of deep sequencing data and often fail to identify the best variants in the libraries. Results: Here, we present a method, Position-Specific Enrichment Ratio Matrix (PSERM) scoring, that uses entire deep sequencing datasets from pre- and post-selections to score each observed protein variant. The PSERM scores are the sum of the site-specific enrichment ratios observed at each mutated position. We find that PSERM scores are much more reproducible and correlate more strongly with experimentally measured properties than frequencies or enrichment ratios, including for multiple antibody properties (affinity and non-specific binding) for a clinical-stage antibody (emibetuzumab). We expect that this method will be broadly applicable to diverse protein engineering campaigns. Availability: All deep sequencing datasets and code to do the analyses presented within are available via GitHub. Contact: Peter Tessier, ptessier@umich.edu. Supplementary information: Supplementary data are available at Bioinformatics online.

Position-Specific Enrichment Ratio Matrix scores predict antibody variant properties from deep sequencing data.

MOTIVATION: Deep sequencing of antibody and related protein libraries after phage or yeast-surface display sorting is widely used to identify variants with increased affinity, specificity, and/or improvements in key biophysical properties. Conventional approaches for identifying optimal variants typically use the frequencies of observation in enriched libraries or the corresponding enrichment ratios. However, these approaches disregard the vast majority of deep sequencing data and often fail to identify the best variants in the libraries. RESULTS: Here, we present a method, Position-Specific Enrichment Ratio Matrix (PSERM) scoring, that uses entire deep sequencing datasets from pre- and post-selections to score each observed protein variant. The PSERM scores are the sum of the site-specific enrichment ratios observed at each mutated position. We find that PSERM scores are much more reproducible and correlate more strongly with experimentally measured properties than frequencies or enrichment ratios, including for multiple antibody properties (affinity and non-specific binding) for a clinical-stage antibody (emibetuzumab). We expect that this method will be broadly applicable to diverse protein engineering campaigns. AVAILABILITY AND IMPLEMENTATION: All deep sequencing datasets and code to perform the analyses presented within are available via https://github.com/Tessier-Lab-UMich/PSERM_paper.

Rapid Evaluation of Staple Placement in Stabilized α Helices Using Bacterial Surface Display.

There are a wealth of proteins involved in disease that cannot be targeted by current therapeutics because they are inside cells, inaccessible to most macromolecules, and lack small-molecule binding pockets. Stapled peptides, where two amino acids are covalently linked, form a class of macrocycles that have the potential to penetrate cell membranes and disrupt intracellular protein-protein interactions. However, their discovery relies on solid-phase synthesis, greatly limiting queries into their complex design space involving amino acid sequence, staple location, and staple chemistry. Here, we use stabilized peptide engineering by Escherichia coli display (SPEED), which utilizes noncanonical amino acids and click chemistry for stabilization, to rapidly screen staple location and linker structure to accelerate peptide design. After using SPEED to confirm hotspots in the mdm2-p53 interaction, we evaluated different staple locations and staple chemistry to identify several novel nanomolar and sub-nanomolar antagonists. Next, we evaluated SPEED in the B cell lymphoma 2 (Bcl-2) protein family, which is responsible for regulating apoptosis. We report that novel staple locations modified in the context of BIM, a high affinity but nonspecific naturally occurring peptide, improve its specificity against the highly homologous proteins in the Bcl-2 family. These compounds demonstrate the importance of screening linker location and chemistry in identifying high affinity and specific peptide antagonists. Therefore, SPEED can be used as a versatile platform to evaluate multiple design criteria for stabilized peptide engineering.

Directed Evolution Using Stabilized Bacterial Peptide Display.

Chemically stabilized peptides have attracted intense interest by academics and pharmaceutical companies due to their potential to hit currently "undruggable" targets. However, engineering an optimal sequence, stabilizing linker location, and physicochemical properties is a slow and arduous process. By pairing non-natural amino acid incorporation and cell surface click chemistry in bacteria with high-throughput sorting, we developed a method to quantitatively select high affinity ligands and applied the Stabilized Peptide Evolution by E. coli Display technique to develop disrupters of the therapeutically relevant MDM2-p53 interface. Through in situ stabilization on the bacterial surface, we demonstrate rapid isolation of stabilized peptides with improved affinity and novel structures. Several peptides evolved a second loop including one sequence (Kd = 1.8 nM) containing an i, i+4 disulfide bond. NMR structural determination indicated a bent helix in solution and bound to MDM2. The bicyclic peptide had improved protease stability, and we demonstrated that protease resistance could be measured both on the bacterial surface and in solution, enabling the method to test and/or screen for additional drug-like properties critical for biologically active compounds.