Combining machine learning with structure-based protein design to predict and engineer post-translational modifications of proteins

In silico design of foldable lasso peptides

Lasso peptides are a unique class of natural products with distinctively threaded structures, conferring exceptional stability against thermal and proteolytic degradation. Despite their promising biotechnological and pharmaceutical applications, reported attempts to prepare them by chemical synthesis result in forming the nonthreaded branched-cyclic isomer, rather than the desired lassoed structure. This is likely due to the entropic challenge of folding a short, threaded motif before chemically mediated cyclization. Accordingly, this study aims to better understand and enhance the relative stability of pre-lasso conformations-the essential precursor to lasso peptide formation-through sequence optimization, chemical modification, and disulfide incorporation. Using Rosetta fixed backbone design, optimal sequences for several class II lasso peptides are identified. Enhanced sampling with well-tempered metadynamics confirmed that designed sequences derived from the lasso structures of rubrivinodin and microcin J25 exhibit a notable improvement in pre-lasso stability relative to the competing nonthreaded conformations. Chemical modifications to the isopeptide bond-forming residues of microcin J25 further increase the probability of pre-lasso formation, highlighting the beneficial role of noncanonical amino acid residues. Counterintuitively, the introduction of a disulfide cross-link decreased pre-lasso stability. Although cross-linking inherently constrains the peptide structure, decreasing the entropic dominance of unfolded phase space, it hinders the requisite wrapping of the N-terminal end around the tail to adopt the pre-lasso conformation. However, combining chemical modifications with the disulfide cross-link results in further pre-lasso stabilization, indicating that the ring modifications counteract the constraints and provide a cooperative benefit with cross-linking. These findings lay the groundwork for further design efforts to enable synthetic access to the lasso peptide scaffold.

Engineering nitrogen and carbon fixation for next-generation plants

Improving plant nitrogen (N) and carbon (C) acquisition and assimilation is a major challenge for global agriculture, food security, and ecological sustainability. Emerging synthetic biology techniques, including directed evolution, artificial intelligence (AI)-guided enzyme design, and metabolic engineering, have opened new avenues for optimizing nitrogenase to fix atmospheric N2 in plants, engineering Rhizobia or other nitrogen-fixing bacteria for symbiotic associations with both legume and nonlegume crops, and enhancing carbon fixation to improve photosynthetic efficiency and source-to-sink assimilate fluxes. Here, we discuss the potential for engineering nitrogen fixation and carbon fixation mechanisms in plants, from rational and AI-driven optimization of nitrogen and carbon fixation cycles. Furthermore, we discuss strategies for modifying source-to-sink relationships to promote robust growth in extreme conditions, such as arid deserts, saline-alkaline soils, or even extraterrestrial environments like Mars. The combined engineering of N and C pathways promises a new generation of crops with enhanced productivity, resource-use efficiency, and resilience. Finally, we explore future perspectives, focusing on the integration of enzyme engineering via directed evolution and computational design to accelerate metabolic innovation in plants.

Self-supervised machine learning methods for protein design improve sampling but not the identification of high-fitness variants

Machine learning (ML) is changing the world of computational protein design, with data-driven methods surpassing biophysical-based methods in experimental success. However, they are most often reported as case studies, lack integration and standardization, and are therefore hard to objectively compare. In this study, we established a streamlined and diverse toolbox for methods that predict amino acid probabilities inside the Rosetta software framework that allows for the side-by-side comparison of these models. Subsequently, existing protein fitness landscapes were used to benchmark novel ML methods in realistic protein design settings. We focused on the traditional problems of protein design: sampling and scoring. A major finding of our study is that ML approaches are better at purging the sampling space from deleterious mutations. Nevertheless, scoring resulting mutations without model fine-tuning showed no clear improvement over scoring with Rosetta. We conclude that ML now complements, rather than replaces, biophysical methods in protein design.

Tying the Knot: In Silico Design of Foldable Lasso Peptides

Lasso peptides are a unique class of natural products with distinctively threaded structures, conferring exceptional stability against thermal and proteolytic degradation. Despite their promising biotechnological and pharmaceutical applications, reported attempts to prepare them by chemical synthesis result in forming the nonthreaded branched-cyclic isomer, rather than the desired lassoed structure. This is likely due to the entropic challenge of folding a short, threaded motif prior to chemically mediated cyclization. Accordingly, this study aims to better understand and enhance the relative stability of pre-lasso conformations-the essential precursor to lasso peptide formation-through sequence optimization, chemical modification, and disulfide incorporation. Using Rosetta fixed backbone design, optimal sequences for several class II lasso peptides are identified. Enhanced sampling with well-tempered metadynamics confirmed that designed sequences derived from the lasso structures of rubrivinodin and microcin J25 exhibit a notable improvement in pre-lasso stability relative to the competing nonthreaded conformations. Chemical modifications to the isopeptide bond-forming residues of microcin J25 further increase the probability of pre-lasso formation, highlighting the beneficial role of non-canonical amino acid residues. Counterintuitively, the introduction of a disulfide cross-link decreased pre-lasso stability. Although cross-linking inherently constrains the peptide structure, decreasing the entropic dominance of unfolded phase space, it hinders the requisite wrapping of the N-terminal end around the tail to adopt the pre-lasso conformation. However, combining chemical modifications with the disulfide cross-link results in further pre-lasso stabilization, indicating that the ring modifications counteract the constraints and provide a cooperative benefit with cross-linking. These findings lay the groundwork for further design efforts to enable synthetic access to the lasso peptide scaffold.

SIGNIFICANCE

Lasso peptides are a unique class of ribosomally synthesized and post-translationally modified natural products with diverse biological activities and potential for therapeutic applications. Although direct synthesis would facilitate therapeutic design, it has not yet been possible to fold these short sequences to their threaded architecture without the help of biosynthetic enzyme stabilization. Our work explores strategies to enhance the stability of the pre-lasso structure, the essential precursor to de novo lasso peptide formation. We find that sequence design, incorporating non-canonical amino acid residues, and design-guided cross-linking can augment stability to increase the likelihood of lasso motif accessibility. This work presents several strategies for the continued design of foldable lasso peptides.

DLBWE-Cys: a deep-learning-based tool for identifying cysteine S-carboxyethylation sites using binary-weight encoding

Cysteine S-carboxyethylation, a novel post-translational modification (PTM), plays a critical role in the pathogenesis of autoimmune diseases, particularly ankylosing spondylitis. Accurate identification of S-carboxyethylation modification sites is essential for elucidating their functional mechanisms. Unfortunately, there are currently no computational tools that can accurately predict these sites, posing a significant challenge to this area of research. In this study, we developed a new deep learning model, DLBWE-Cys, which integrates CNN, BiLSTM, Bahdanau attention mechanisms, and a fully connected neural network (FNN), using Binary-Weight encoding specifically designed for the accurate identification of cysteine S-carboxyethylation sites. Our experimental results show that our model architecture outperforms other machine learning and deep learning models in 5-fold cross-validation and independent testing. Feature comparison experiments confirmed the superiority of our proposed Binary-Weight encoding method over other encoding techniques. t-SNE visualization further validated the model's effective classification capabilities. Additionally, we confirmed the similarity between the distribution of positional weights in our Binary-Weight encoding and the allocation of weights in attentional mechanisms. Further experiments proved the effectiveness of our Binary-Weight encoding approach. Thus, this model paves the way for predicting cysteine S-carboxyethylation modification sites in protein sequences. The source code of DLBWE-Cys and experiments data are available at: https://github.com/ztLuo-bioinfo/DLBWE-Cys.

Artificial Intelligence Transforming Post-Translational Modification Research

Post-Translational Modifications (PTMs) are covalent changes to amino acids that occur after protein synthesis, including covalent modifications on side chains and peptide backbones. Many PTMs profoundly impact cellular and molecular functions and structures, and their significance extends to evolutionary studies as well. In light of these implications, we have explored how artificial intelligence (AI) can be utilized in researching PTMs. Initially, rationales for adopting AI and its advantages in understanding the functions of PTMs are discussed. Then, various deep learning architectures and programs, including recent applications of language models, for predicting PTM sites on proteins and the regulatory functions of these PTMs are compared. Finally, our high-throughput PTM-data-generation pipeline, which formats data suitably for AI training and predictions is described. We hope this review illuminates areas where future AI models on PTMs can be improved, thereby contributing to the field of PTM bioengineering.

Integrative Multi-PTM Proteomics Reveals Dynamic Global, Redox, Phosphorylation, and Acetylation Regulation in Cytokine-Treated Pancreatic Beta Cells

Studying regulation of protein function at a systems level necessitates an understanding of the interplay among diverse posttranslational modifications (PTMs). A variety of proteomics sample processing workflows are currently used to study specific PTMs but rarely characterize multiple types of PTMs from the same sample inputs. Method incompatibilities and laborious sample preparation steps complicate large-scale physiological investigations and can lead to variations in results. The single-pot, solid-phase-enhanced sample preparation (SP3) method for sample cleanup is compatible with different lysis buffers and amenable to automation, making it attractive for high-throughput multi-PTM profiling. Herein, we describe an integrative SP3 workflow for multiplexed quantification of protein abundance, cysteine thiol oxidation, phosphorylation, and acetylation. The broad applicability of this approach is demonstrated using cell and tissue samples, and its utility for studying interacting regulatory networks is highlighted in a time-course experiment of cytokine-treated β-cells. We observed a swift response in the global regulation of protein abundances consistent with rapid activation of JAK-STAT and NF-κB signaling pathways. Regulators of these pathways as well as proteins involved in their target processes displayed multi-PTM dynamics indicative of complex cellular response stages: acute, adaptation, and chronic (prolonged stress). PARP14, a negative regulator of JAK-STAT, had multiple colocalized PTMs that may be involved in intraprotein regulatory crosstalk. Our workflow provides a high-throughput platform that can profile multi-PTMomes from the same sample set, which is valuable in unraveling the functional roles of PTMs and their co-regulation.

Rational Design for Enhancing Cellobiose Dehydrogenase Activity and Its Synergistic Role in Straw Degradation

Addressing the demand for efficient biological degradation of straw, this study employed rational design coupled with structural biology and enzyme engineering techniques to enhance the catalytic activity of cellobiose dehydrogenase (PsCDH, CDH form Pycnoporus sanguineus). By predicting and modifying the active site and key amino acids of PsCDH, several CDH immobilized enzyme preparations with higher catalytic activities were successfully obtained. The excellent mutant T1 (C286Y/A461H/F464R) exhibited a 2.7-fold increase in enzyme activity compared to the wild type. Simulated calculations indicated that the enhancement of catalytic activity was primarily due to the formation of additional intermolecular interactions between CDH and the substrate, as well as the enlargement of the substrate pocket to reduce steric hindrance effects. Additionally, molecular dynamics simulation analysis revealed a potential correlation between structural stability and enzyme activity. When PsCDH was added to a multienzyme synergistic straw degradation system, the cellulose degradation rate increased by 1.84-fold. Moreover, mutant T1 further increased the degradation of lignocellulose in the mixed system. This study provides efficient enzyme sources and modification strategies for the high-efficiency biological conversion of straw and unconventional feedstock degradation, thereby possessing significant academic value and application prospects.

Current computational tools for protein lysine acylation site prediction

As a main subtype of post-translational modification (PTM), protein lysine acylations (PLAs) play crucial roles in regulating diverse functions of proteins. With recent advancements in proteomics technology, the identification of PTM is becoming a data-rich field. A large amount of experimentally verified data is urgently required to be translated into valuable biological insights. With computational approaches, PLA can be accurately detected across the whole proteome, even for organisms with small-scale datasets. Herein, a comprehensive summary of 166 in silico PLA prediction methods is presented, including a single type of PLA site and multiple types of PLA sites. This recapitulation covers important aspects that are critical for the development of a robust predictor, including data collection and preparation, sample selection, feature representation, classification algorithm design, model evaluation, and method availability. Notably, we discuss the application of protein language models and transfer learning to solve the small-sample learning issue. We also highlight the prediction methods developed for functionally relevant PLA sites and species/substrate/cell-type-specific PLA sites. In conclusion, this systematic review could potentially facilitate the development of novel PLA predictors and offer useful insights to researchers from various disciplines.

Combining Rosetta Sequence Design with Protein Language Model Predictions Using Evolutionary Scale Modeling (ESM) as Restraint

Computational protein sequence design has the ambitious goal of modifying existing or creating new proteins; however, designing stable and functional proteins is challenging without predictability of protein dynamics and allostery. Informing protein design methods with evolutionary information limits the mutational space to more native-like sequences and results in increased stability while maintaining functions. Recently, language models, trained on millions of protein sequences, have shown impressive performance in predicting the effects of mutations. Assessing Rosetta-designed sequences with a language model showed scores that were worse than those of their original sequence. To inform Rosetta design protocols with language model predictions, we added a new metric to restrain the energy function during design using the Evolutionary Scale Modeling (ESM) model. The resulting sequences have better language model scores and similar sequence recovery, with only a minor decrease in the fitness as assessed by Rosetta energy. In conclusion, our work combines the strength of recent machine learning approaches with the Rosetta protein design toolbox.