Explore the vast range of titles available.

The optimization of therapeutic antibodies is time-intensive and resource-demanding, largely because of the low-throughput screening of full-length antibodies (approximately 1 × 10 3 variants) expressed in mammalian cells, which typically results in few optimized leads. Here we show that optimized antibody variants can be identified by predicting antigen specificity via deep learning from a massively diverse space of antibody sequences. To produce data for training deep neural networks, we deep-sequenced libraries of the therapeutic antibody trastuzumab (about 1 × 10 4 variants), expressed in a mammalian cell line through site-directed mutagenesis via CRISPR–Cas9-mediated homology-directed repair, and screened the libraries for specificity to human epidermal growth factor receptor 2 (HER2). We then used the trained neural networks to screen a computational library of approximately 1 × 10 8 trastuzumab variants and predict the HER2-specific subset (approximately 1 × 10 6 variants), which can then be filtered for viscosity, clearance, solubility and immunogenicity to generate thousands of highly optimized lead candidates. Recombinant expression and experimental testing of 30 randomly selected variants from the unfiltered library showed that all 30 retained specificity for HER2. Deep learning may facilitate antibody engineering and optimization. Therapeutic antibodies can be optimized using deep-learning models trained on antibody-mutagenesis libraries to generate antibody variants and predict their antigen specificity.

Calcitonin gene-related peptide (CGRP) is crucial in the pathophysiology of migraine. We assessed the safety, tolerability, and efficacy of ALD403, a genetically engineered humanised anti-CGRP antibody, for migraine prevention. In this randomised, double-blind, placebo-controlled, exploratory, proof-of-concept phase 2 trial, patients aged 18–55 years with five to 14 migraine days per 28-day period were randomly assigned (1:1) via an interactive web response system to receive an intravenous dose of ALD403 1000 mg or placebo. Site investigators, patients, and the sponsor were masked to treatment allocation during the study. The primary objective was to assess safety at 12 weeks after infusion. The primary efficacy endpoint was the change from baseline to weeks 5–8 in the frequency of migraine days, as recorded in patient electronic diaries. Patients were followed up until 24 weeks for exploratory safety and efficacy analyses. Safety and efficacy analyses were done by intention to treat. This study is registered with ClinicalTrials.gov, NCT01772524. Between Jan 28, 2013, and Dec 23, 2013, of 174 patients randomly assigned at 26 centres in the USA, 163 received either ALD403 (n=81) or placebo (n=82). Adverse events were experienced by 46 (57%) of 81 patients in the ALD403 group and 43 (52%) of 82 in the placebo group. The most frequent adverse events were upper respiratory tract infection (placebo 6 [7%] patients vs ALD403 7 [9%] patients), urinary tract infection (4 [5%] vs 1 [1%]), fatigue (3 [4%] vs 3 [4%]), back pain (4 [5%] vs 3 [4%]), arthralgia (4 [5%] vs 1 [1%]), and nausea and vomiting (2 [2%] vs 3 [4%]). Six serious adverse events were reported by three patients and were judged to be unrelated to study drug: in the ALD403 group, one patient had four serious adverse events and one had one serious adverse event, and in the placebo group, one patient had one serious adverse event. There were no differences in vital signs or laboratory safety data between the two treatment groups. The mean change in migraine days between baseline and weeks 5–8 was −5·6 (SD 3·0) for the ALD403 group compared with −4·6 (3·6) for the placebo group (difference −1·0, 95% CI −2·0 to 0·1; one-sided p=0·0306). No safety concerns were noted with an intravenous dose of ALD403 1000 mg. This study also provides preliminary evidence for the efficacy of ALD403 in the preventive treatment of migraine in patients with a high monthly frequency of migraine days. Alder Biopharmaceuticals.

Antisense peptide technology (APT) is based on a useful heuristic algorithm for rational peptide design. It was deduced from empirical observations that peptides consisting of complementary (sense and antisense) amino acids interact with higher probability and affinity than the randomly selected ones. This phenomenon is closely related to the structure of the standard genetic code table, and at the same time, is unrelated to the direction of its codon sequence translation. The concept of complementary peptide interaction is discussed, and its possible applications to diagnostic tests and bioengineering research are summarized. Problems and difficulties that may arise using APT are discussed, and possible solutions are proposed. The methodology was tested on the example of SARS-CoV-2. It is shown that the CABS-dock server accurately predicts the binding of antisense peptides to the SARS-CoV-2 receptor binding domain without requiring predefinition of the binding site. It is concluded that the benefits of APT outweigh the costs of random peptide screening and could lead to considerable savings in time and resources, especially if combined with other computational and immunochemical methods.

3-Hydroxypropionic acid (3-HP) can be produced via the biological route involving two enzymatic reactions: dehydration of glycerol to 3-hydroxypropanal (3-HPA) and then oxidation to 3-HP. However, commercial production of 3-HP using recombinant microorganisms has been hampered with several problems, some of which are associated with the toxicity of 3-HPA and the efficiency of NAD + regeneration. We engineered α-ketoglutaric semialdehyde dehydrogenase (KGSADH) from Azospirillum brasilense for the second reaction to address these issues. The residues in the binding sites for the substrates, 3-HPA and NAD + , were randomized, and the resulting libraries were screened for higher activity. Isolated KGSADH variants had significantly lower K m values for both the substrates. The enzymes also showed higher substrate specificities for aldehyde and NAD + , less inhibition by NADH, and greater resistance to inactivation by 3-HPA than the wild-type enzyme. A recombinant Pseudomonas denitrificans strain with one of the engineered KGSADH variants exhibited less accumulation of 3-HPA, decreased levels of inactivation of the enzymes, and higher cell growth than that with the wild-type KGSADH. The flask culture of the P. denitrificans strain with the mutant KGSADH resulted in about 40% increase of 3-HP titer (53 mM) compared with that using the wild-type enzyme (37 mM).

The continuous search for novel enzyme backbones and the engineering of already well studied enzymes for biotechnological applications has become an increasing challenge, especially by the increasing potential diversity space provided by directed enzyme evolution approaches and the demands of experimental data generated by rational design of enzymes. In this work, we propose a semi-rational mutational strategy focused on introducing diversity in structurally variable regions in enzymes. The identified sequences are subjected to a progressive deletion of two amino acids and the joining residues are subjected to saturation mutagenesis using NNK degenerate codons. This strategy offers a novel library diversity approach while simultaneously decreasing enzyme size in the variable regions. In this way, we intend to identify and reduce variable regions found in enzymes, probably resulting from neutral drift evolution, and simultaneously studying the functional effect of said regions. This strategy was applied to Bacillus. subtilis lipase A (BSLA), by selecting and deleting six variable enzyme regions (named regions 1 to 6) by the deletion of two amino acids and additionally randomizing the joining amino acid residues. After screening, no active variants were found in libraries 1% and 4%, 15% active variants were found in libraries 2% and 3%, and 25% for libraries 5 and 6 (n = 3000 per library, activity detected using tributyrin agar plates). Active variants were assessed for activity in microtiter plate assay (pNP-butyrate), thermal stability, substrate preference (pNP-butyrate, -palmitate), and compared to wildtype BSLA. From these analyses, variant P5F3 (F41L-ΔW42-ΔD43-K44P), from library 3 was identified, showing increased activity towards longer chain p-nitrophenyl fatty acid esters, when compared to BSLA. This study allowed to propose the targeted region 3 (positions 40–46) as a potential modulator for substrate specificity (fatty acid chain length) in BSLA, which can be further studied to increase its substrate spectrum and selectivity. Additionally, this variant showed a decreased thermal resistance but interestingly, higher isopropanol and Triton X-100 resistance. This deletion-randomization strategy could help to expand and explore sequence diversity, even in already well studied and characterized enzyme backbones such as BSLA. In addition, this strategy can contribute to investigate and identify important non-conserved regions in classic and novel enzymes, as well as generating novel biocatalysts with increased performance in specific processes, such as enzyme immobilization.

The ability to design efficient enzymes from scratch would have a profound effect on chemistry, biotechnology and medicine. Rapid progress in protein engineering over the past decade makes us optimistic that this ambition is within reach. The development of artificial enzymes containing metal cofactors and noncanonical organocatalytic groups shows how protein structure can be optimized to harness the reactivity of nonproteinogenic elements. In parallel, computational methods have been used to design protein catalysts for diverse reactions on the basis of fundamental principles of transition state stabilization. Although the activities of designed catalysts have been quite low, extensive laboratory evolution has been used to generate efficient enzymes. Structural analysis of these systems has revealed the high degree of precision that will be needed to design catalysts with greater activity. To this end, emerging protein design methods, including deep learning, hold particular promise for improving model accuracy. Here we take stock of key developments in the field and highlight new opportunities for innovation that should allow us to transition beyond the current state of the art and enable the robust design of biocatalysts to address societal needs. Recent progress in computational enzyme design, active site engineering and directed evolution are reviewed, highlighting methodological innovations needed to deliver improved designer biocatalysts.

Rational protein engineering requires a holistic understanding of protein function. Here, we apply deep learning to unlabeled amino-acid sequences to distill the fundamental features of a protein into a statistical representation that is semantically rich and structurally, evolutionarily and biophysically grounded. We show that the simplest models built on top of this unified representation (UniRep) are broadly applicable and generalize to unseen regions of sequence space. Our data-driven approach predicts the stability of natural and de novo designed proteins, and the quantitative function of molecularly diverse mutants, competitively with the state-of-the-art methods. UniRep further enables two orders of magnitude efficiency improvement in a protein engineering task. UniRep is a versatile summary of fundamental protein features that can be applied across protein engineering informatics.

G-protein-coupled receptors (GPCRs) remain major drug targets, despite our incomplete understanding of how they signal through 16 non-visual G-protein signal transducers (collectively named the transducerome) to exert their actions. To address this gap, we have developed an open-source suite of 14 optimized bioluminescence resonance energy transfer (BRET) Gαβγ biosensors (named TRUPATH) to interrogate the transducerome with single pathway resolution in cells. Generated through exhaustive protein engineering and empirical testing, the TRUPATH suite of Gαβγ biosensors includes the first Gα15 and GαGustducin probes. In head-to-head studies, TRUPATH biosensors outperformed first-generation sensors at multiple GPCRs and in different cell lines. Benchmarking studies with TRUPATH biosensors recapitulated previously documented signaling bias and revealed new coupling preferences for prototypic and understudied GPCRs with potential in vivo relevance. To enable a greater understanding of GPCR molecular pharmacology by the scientific community, we have made TRUPATH biosensors easily accessible as a kit through Addgene. Development of BRET sensors for nearly all major G proteins show that GPCR–G-protein coupling ranges from promiscuous to extremely specific, Switch III is a novel site for G-protein engineering, and optimal donor–acceptor positioning is non-obvious.

Protein engineering has enormous academic and industrial potential. However, it is limited by the lack of experimental assays that are consistent with the design goal and sufficiently high throughput to find rare, enhanced variants. Here we introduce a machine learning-guided paradigm that can use as few as 24 functionally assayed mutant sequences to build an accurate virtual fitness landscape and screen ten million sequences via in silico directed evolution. As demonstrated in two dissimilar proteins, GFP from Aequorea victoria (avGFP) and E. coli strain TEM-1 β-lactamase, top candidates from a single round are diverse and as active as engineered mutants obtained from previous high-throughput efforts. By distilling information from natural protein sequence landscapes, our model learns a latent representation of ‘unnaturalness’, which helps to guide search away from nonfunctional sequence neighborhoods. Subsequent low-N supervision then identifies improvements to the activity of interest. In sum, our approach enables efficient use of resource-intensive high-fidelity assays without sacrificing throughput, and helps to accelerate engineered proteins into the fermenter, field and clinic.A deep-learning-guided approach enables protein engineering using only a small number (‘low N’) of functionally characterized variants of target proteins.

Advances in DNA sequencing and machine learning are providing insights into protein sequences and structures on an enormous scale 1 . However, the energetics driving folding are invisible in these structures and remain largely unknown 2 . The hidden thermodynamics of folding can drive disease 3 , 4 , shape protein evolution 5 – 7 and guide protein engineering 8 – 10 , and new approaches are needed to reveal these thermodynamics for every sequence and structure. Here we present cDNA display proteolysis, a method for measuring thermodynamic folding stability for up to 900,000 protein domains in a one-week experiment. From 1.8 million measurements in total, we curated a set of around 776,000 high-quality folding stabilities covering all single amino acid variants and selected double mutants of 331 natural and 148 de novo designed protein domains 40–72 amino acids in length. Using this extensive dataset, we quantified (1) environmental factors influencing amino acid fitness, (2) thermodynamic couplings (including unexpected interactions) between protein sites, and (3) the global divergence between evolutionary amino acid usage and protein folding stability. We also examined how our approach could identify stability determinants in designed proteins and evaluate design methods. The cDNA display proteolysis method is fast, accurate and uniquely scalable, and promises to reveal the quantitative rules for how amino acid sequences encode folding stability. Large-scale assays using cDNA display proteolysis are used to measure the folding stabilities of protein domains, providing a method to quantify the effects of mutations on protein folding, with applications in protein design.