HighRelax: Physics-Based Refinement of Deep Learning Protein Predictions with Noncanonical Amino Acids.

Dos aminoácidos existentes, vinte estão envolvidos diretamente na formação de proteínas. Os demais são conhecidos como aminoácidos não-proteinogênicos (AANP). Os AANP possuem diferentes funções, incluindo atividade anti-herbivoria, antimicrobiana, proteção contra o estresse, sinalização celular, armazenamento de nitrogênio, como toxinas contra invertebrados e vertebrados e, também, como aleloquímicos. As plantas são capazes de liberar esses metabólitos no solo, os quais podem influenciar o crescimento e desenvolvimento de plantas vizinhas, tanto positiva como negativamente. O objetivo dessa revisão é descrever as características químicas, o potencial alelopático e a seletividade para culturas dos aminoácidos não-proteinogênicos m-tirosina, L-DOPA, metionina sulfoximina e azetidino.

Cyclic peptides with backbone N-methylated amino acids (BNMeAAs) and D-amino acids (D-AAs) have gained increasing attention for their stability, membrane permeability, and other therapeutic potentials. Currently, Rosetta simple_cycpep_predict (SCP) can predict their structures through energy-based calculations, but this approach is computationally intensive and time-consuming. Moreover, the available crystal structures of such cyclic peptides remain highly limited, hindering the development of data-driven structure prediction models. To address these challenges, we propose HighFold-MeD, a deep learning-based framework that distills knowledge from Rosetta SCP by fine-tuning the AlphaFold model. First, a cyclic peptide structure dataset is constructed using Rosetta SCP by sampling massive conformations for cyclic peptides with BNMeAAs and D-AAs and evaluating their energy scores. The AlphaFold model is then fine-tuned to incorporate the extended 56 BNMeAAs and D-AAs. Besides, a relative position cyclic matrix is introduced to explicitly model head-to-tail cyclization. Finally, a force field is employed to minimize steric clashes in the predicted structures. Empirical experiments demonstrate that HighFold-MeD achieves accuracy comparable to that of Rosetta based on the sampled datasets by the SCP module of Rosetta, with the key parameters that nstruct = 500 and cyclic_peptide: genkic_closure_attempts = 1000, while accelerating structure prediction by 50-fold, thereby significantly expediting the development of cyclic peptide-based therapeutics.Scientific contributionWe propose HighFold-MeD, which provides a rapid and relatively accurate approach for predicting the structures of cyclic peptides containing backbone N-methylated amino acids and D-amino acids—key building blocks in peptide drug design. By distilling the knowledge of Rosetta SCP under specific parameters into a fine-tuned AlphaFold framework, our method achieves a 50-fold acceleration while maintaining relatively high accuracy, thereby enabling large-scale cyclic peptide drug design.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13321-025-01111-3.

Unnatural amino acids have gained significant attention in protein engineering and drug discovery as they allow the evolution of proteins with enhanced stability and activity. The incorporation of unnatural amino acids into proteins offers a rational approach to engineer enzymes for designing efficient biocatalysts that exhibit versatile physicochemical properties and biological functions. This review highlights the biological and synthetic routes of unnatural amino acids to yield a modified protein with altered functionality and their incorporation methods. Unnatural amino acids offer a wide array of applications such as antibody-drug conjugates, probes for change in protein conformation and structure–activity relationships, peptide-based imaging, antimicrobial activities, etc. Besides their emerging applications in fundamental and applied science, systemic research is necessary to explore unnatural amino acids with novel side chains that can address the limitations of natural amino acids.

Peptide drug discovery often benefits from the large structural diversity permitted by unnatural amino acids (UAAs). Indeed, numerous approved peptide drugs include UAAs in their sequences. Therefore, innovative chemical approaches either to synthesize UAAs or to allow late‐stage functionalization of peptides are emerging themes in peptide drug discovery. Thanks to the recent advances in deaminative strategies using alkylpyridiniums salts, often referred to as Katritzky salts, a variety of radical alkylation methods have been developed. In recent years the use of Katritzky salts have become popular in peptide chemistry due to their ease of preparation from a primary amine, which is a predominant functional group in amino acids. This review highlights the progress that has been made by using Katritzky salts in the synthesis of UAAs, late‐stage peptide functionalization, and peptide macrocyclization.

The utility of non-canonical amino acids in protein engineering has grown substantially over the past decade. Proteins containing these unnatural building blocks often have radically different biochemical or spectral characteristics than their wild-type counterparts. Furthermore, proteins may be endowed with chemical reactivity not found in the natural proteome upon the introduction of non-canonical amino acids. Successful incorporation of a non-canonical amino acid into recombinant proteins in E. coli is often dependent on engineering of the aminoacyl-tRNA synthetase (aaRS) activity of the cell. The bulk of the work described herein has focused on developing a system to rapidly screen libraries of mutant aaRS to identify clones capable of efficiently incorporating novel reactive non-canonical amino acids. The system is based on the display of reactive amino acid side chains on the surface of E. coli cells upon metabolic incorporation of the amino acid into recombinant outer membrane protein C (OmpC) and the subsequent covalent biotinylation of the reactive side chains. The cells are then stained with fluorescent avidin, thus rendering the cells incorporating the amino acid fluorescent and readily identifiable and sortable by flow cytometry. The feasibility of such a system was proven by incorporating the methionine surrogate azidohomoalanine (AHA) into OmpC and subsequently biotinylating the reactive azide groups via copper-catalyzed azide-alkyne ligation. Using an improved copper catalyst, low levels of incorporation of translationally inefficient amino acids azidoalanine, azidonorvaline, and azidonorleucine into OmpC were also detected. A saturation mutagenesis library of the methionyl-tRNA synthetase (MetRS) was designed, and cells transformed with this library were screened for the ability to incorporate the long chain amino acid azidonorleucine into recombinant proteins efficiently. Several MetRS mutants were identified with such activity using the cell surface display system. MetRS containing a single amino acid mutation, leucine 13 to glycine (L13G) that occurs in each of the three mutants discovered in the screen, is very efficient at incorporating azidonorleucine into proteins. In the last part of the work described in this thesis, azidohomoalanine was used to tag newly-synthesized proteins in mammalian cells, thus endowing the newly-synthesized proteins with unique bioorthogonal chemical reactivity. Following covalent biotinylation via the azide-alkyne ligation, these proteins could be selectively enriched for by avidin chromatography and identified using shotgun proteomic approaches. Nearly 200 newly-synthesized proteins were identified unequivocally in just a two-hour window. This technique promises to develop into a highly useful tool for the examination of proteome dynamics.

Genetic code expansion allows unnatural (non-canonical) amino acid incorporation into proteins of interest by repurposing the cellular translation machinery. The development of this technique has enabled site-specific incorporation of many structurally and chemically diverse amino acids, facilitating a plethora of applications, including protein imaging, engineering, mechanistic and structural investigations, and functional regulation. Particularly, genetic code expansion provides great tools to study mammalian proteins, of which dysregulations often have important implications in health. In recent years, a series of methods has been developed to modulate protein function through genetically incorporated unnatural amino acids. In this review, we will first discuss the basic concept of genetic code expansion and give an up-to-date list of amino acids that can be incorporated into proteins in mammalian cells. We then focus on the use of unnatural amino acids to activate, inhibit, or reversibly modulate protein function by translational, optical or chemical control. The features of each approach will also be highlighted.

Proteins mediate the bulk of biochemical functions within the cell. These biopolymers control processes utilizing specific arrangements of the natural twenty amino acids. Expanding the set of amino acids available could both aid in the study of these macromolecules as well as significantly increase their functional capabilities. A set of enzymes known as aminoacyl tRNA-synthetases lies at the heart of the fidelity of translation, the process by which genetic information gets decoded into proteins. These synthetases accurately charge a specific tRNA with its cognate amino acid in the presence of the other nineteen natural amino acids. Interestingly these enzymes demonstrate a much higher level of promiscuity with unnatural amino acids. However, acceptable amino acids are limited to those that bear steric and electronic resemblance to the natural analog. Our efforts to expand the substrate set of phenylalanyl-tRNA (PheRS) synthetase are described in Chapters 2-4. We redesigned the catalytic site of PheRS computationally. These results combined with an already known mutant allowed us to rationally create a third mutant. All three mutants were characterized for their ability to activate a large panel of unnatural amino acids in vitro. Further, we were able to confirm the in vivo incorporation of a number of these analogs. In vitro and in vivo results were consistent and defined an expanded substrate set for the described mutants. This substrate set includes a number of analogs that are dramatically different from phenylalanine both sterically and electronically, as well as a number which contain chemical moieties valuable to protein engineering efforts. One example is para-azidophenylalanine (pN3Phe), which provides access to photochemistry as well as modified Staudinger ligations and copper mediated electrocyclizations. In Chapter 5 we describe utilization pN3Phe as a photocrosslinking reagent. Our aim was to create photochemically crosslinkable artificial extracellular matrix proteins for the production of synthetic vascular grafts. These proteins, produced in E. coli, were diblocks of endothelial cell binding domains and structural domains including the pN3Phe site. Photochemical crosslinking of these constructs provided moduli well within the range presented by the natural vascular wall. Chapter 6 describes our ability to photopattern films composed of the above protein. Photopatterning provided a means to spatially array endothelial cells, based upon a number of controllable processing parameters of such films. The final chapter details the utilization of incorporated unnatural amino acids, particularly para-iodophenylalanine, para-acetylphenylalanine and homopropargylglycine, to access Pd(0) catalyzed cross-coupling chemistry. We demonstrated this chemistry exhibits the characteristics necessary for chemoselective ligations. Futher, we demonstrated the selective modification of proteins incorporating all of the above analogs.

The selective generation of covalent bonds between and within proteins would provide new avenues for studying protein function and engineering proteins with new properties. New covalent bonds were genetically introduced into proteins by enabling an unnatural amino acid (Uaa) to selectively react with a proximal natural residue. This proximity-enabled bioreactivity was expanded to a series of haloalkane Uaas. Orthogonal tRNA/synthetase pairs were evolved to incorporate these Uaas, which only form a covalent thioether bond with cysteine when positioned in close proximity. By using the Uaa and cysteine, spontaneous covalent bond formation was demonstrated between an affibody and its substrate Z protein, thereby leading to irreversible binding, and within the affibody to increase its thermostability. This strategy of proximity-enabled protein crosslinking (PEPC) may be generally expanded to target different natural amino acids, thus providing diversity and flexibility in covalent bond formation for protein research and protein engineering.

The complex between the N-methyl-d-aspartate receptor (NMDAR), neuronal nitric oxide synthase (nNOS), and the postsynaptic density protein-95 (PSD-95) is an attractive therapeutic target for the treatment of acute ischemic stroke. The complex is formed via the PDZ protein domains of PSD-95, and efforts to disrupt the complex have generally been based on C-terminal peptides derived from the NMDAR. However, nNOS binds PSD-95 through a β-hairpin motif, providing an alternative starting point for developing PSD-95 inhibitors. Here, we designed a cyclic nNOS β-hairpin mimetic peptide and generated cyclic nNOS β-hairpin peptide arrays with natural and unnatural amino acids (AAs), which provided molecular insights into this interaction. We then optimized cyclic peptides and identified a potent inhibitor of the nNOS/PSD-95 interaction, with the highest affinity reported thus far for a peptide macrocycle inhibitor of PDZ domains, which serves as a template for the development of treatment for acute ischemic stroke.

Naturally, proteins generally consist of twenty natural amino acids (NAAs) as building blocks, which can form a nearly unlimited number of combinations by random combination to realize structural and functional diversity. Unfortunately, it is not enough using these natural blocks to generate proteins with specific characteristics. Therefore, incorporating unnatural amino acids (UNAAs) with some novel groups to expand the repertoire of structures and functions of proteins is becoming more and more popular [1]. Diverse bioreactivity from novel functional groups are unreachable to natural proteins, which opens gates for new protein engineering and provides new ways for fundamental research, therapeutics, and synthetic biology [2]. Incorporation of UNAA at a specific site has been used for studying protein structure and dynamics, characterizing protein-protein interactions, mimicking post-translational modifications of proteins like eukaryote, and synthesize novel products hard to be created by other methods, such as enzymes, biomaterials, and therapeutics [3].

The studies discussed in this dissertation are aimed at the chemical-scale interactions involved in neuroreceptor structure and function. Unnatural amino acids were incorporated into several ligand-gated ion channels. Two different ionotropic glutamate receptors (iGluRs), the N-methyl-D-aspartate (NMDA) receptor and the α–amino-3-hydroxy-5-methyl- 4-isoxazolepropionic acid (AMPA) receptor were studied, along with an acetylcholine receptor - the nicotinic acetylcholine receptor (nAChR), and all were analyzed with electrophysiology–an assay of receptor function. In Chapter 2, a highly conserved tryptophan (Trp607) in the ion channel pore of the NMDA receptor was investigated for its role during extracellular Mg2+ block. Previous studies hypothesized that a cation-π interaction between NR2BW607 and Mg2+ contributed to the receptor blockade. However, our studies suggest that Trp607 is not involved in a cation-π interaction with Mg2+, instead it is a structural component of the pore. NR2B Trp607 acts as a steric “plug,” preventing Mg2+ permeation through the ion channel. These studies were the first to incorporate unnatural amino acids into a glutamate receptor, extending the scope of nonsense suppression methodology to a new class of neuroreceptors. Chapter 3 describes the incorporation of unnatural amino acids into the ligand binding domain (LBD) of NMDA and AMPA receptors. Previous structural studies of AMPA receptors established the overall topology of the LBD to be a clamshell, two domains clamp down around a central cleft. Further studies utilizing agonists that induce full receptor activation and partial receptor activation demonstrate a relationship between cleft closure and agonist efficacy, which is the ability to activate a receptor. Full agonists correlate with more cleft closure than partial agonists, which induce less cleft closure. To examine this relationship, we used unnatural amino acid mutagenesis to convert an NR2-conserved tyrosine to homotyrosine and an NR1 glutamine to homoglutamine, residues designed to disrupt clamshell closure by expanding the side chain without altering its functionality. The development of our functional probe demonstrates that the clamshell closure mechanism, previously shown for AMPA receptors, likely also applies to NMDA receptors, but to different degrees in the NR1 and NR2 subunits. Finally, in Chapter 4 we use unnatural amino acids, mutagenesis, and computational simulations to probe the binding interactions that are involved in agonist selectivity at the muscle-type nicotinic acetylcholine receptor. Acetylcholine (ACh) and nicotine, both agonists for nAChRs, have a high potency for neuronal receptors. However, nicotine is a weak agonist for the muscle-type nAChRs, yet the amino acids that contribute to the binding site remain the same between both types of receptors. These studies use mutagenesis and unnatural amino acids to introduce changes in the muscle-type receptor to increase nicotine potency. Although some of the mutations increase nicotine potency, none of the mutations result in a muscle-type receptor with nicotine potency as great as the neuronal receptors. A second set of studies generated a mouse muscle homology model and used molecular dynamics to simulate movements in the receptor with and without agonist bound. These structures demonstrate the importance of a hydrogen-bonding network that contributes to the pre-organization of the aromatic box.

Protein engineering, the process by which novel proteins with desired properties are developed, has grown by leaps and bounds since the first examples of protein mutagenesis were described over three decades ago. In the last 30 years, protein engineers have successfully created a wide range of proteins tailored to specific industrial, medical, and research applications. Yet challenges remain that prevent the engineering of complex protein functions on demand. With the aim of creating a forum for the protein engineering community, the inaugural Protein Engineering Canada Conference was held on June 20–22, 2014, in Ottawa, Canada. Over the course of the 2-day conference, 115 protein scientists from over 30 institutions representing 5 countries shared ideas, networked with colleagues, and learned about exciting new research covering a broad array of topics such as enzyme engineering, computational protein design, X-ray crystallography, protein NMR, molecular modeling, and protein evolution. This special issue highlights the breadth of research that was covered during the Protein Engineering Canada Conference by including previously unpublished research articles submitted by speakers as well as some selections that were submitted directly to Protein Science.

In ribosomal polypeptide synthesis the library of amino acid building blocks is limited by the manner in which codons are used. Of the proteinogenic amino acids, 18 are coded for by multiple codons and therefore many of the 61 sense codons can be considered redundant. Here we report a method to reduce the redundancy of codons by artificially dividing codon boxes to create vacant codons that can then be reassigned to non-proteinogenic amino acids and thereby expand the library of genetically encoded amino acids. To achieve this, we reconstituted a cell-free translation system with 32 in vitro transcripts of transfer RNASNN (tRNASNN) (S = G or C), assigning the initiator and 20 elongator amino acids. Reassignment of three redundant codons was achieved by replacing redundant tRNASNNs with tRNASNNs pre-charged with non-proteinogenic amino acids. As a demonstration, we expressed a 32-mer linear peptide that consists of 20 proteinogenic and three non-proteinogenic amino acids, and a 14-mer macrocyclic peptide that contains more than four non-proteinogenic amino acids.

In vitro genetic code reprogramming and expansion to study protein function and discover macrocyclic peptide ligands

Unnatural amino acids are widely utilized as biological active compounds and synthetic intermediates. Among them, thiazol-idine-2-carboxylic acid, which is an unnatural cyclic amino acid, was found in medicinal and organic chemistry area, many of which showed biological activity and used as useful building blocks in organic synthesis.