Biophysics-based protein language models for protein engineering

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: March 17, 2024

Protein language models trained on evolutionary data have emerged as powerful tools for predictive problems involving protein sequence, structure, and function. However, these overlook decades of research into biophysical factors governing We propose Mutational Effect Transfer Learning (METL), a model framework that unites advanced machine learning modeling. Using the METL framework, we pretrain transformer-based neural networks simulation to capture fundamental relationships between energetics. finetune experimental sequence-function harness signals apply them when predicting properties like thermostability, catalytic activity, fluorescence. excels in challenging engineering tasks generalizing from small training sets position extrapolation, although existing methods train remain many types assays. demonstrate METL's ability design functional green fluorescent variants only 64 examples, showcasing potential biophysics-based engineering.

Language: Английский

---

Engineering strategies and challenges of endolysin as an antibacterial agent against Gram‐negative bacteria

Microbial Biotechnology, Journal Year: 2024, Volume and Issue: 17(4)

Published: April 1, 2024

Bacteriophage endolysin is a novel antibacterial agent that has attracted much attention in the prevention and control of drug-resistant bacteria due to its unique mechanism hydrolysing peptidoglycans. Although exhibits excellent bactericidal effects on Gram-positive bacteria, presence outer membrane Gram-negative makes it difficult lyse them extracellularly, thus limiting their application field. To enhance extracellular activity facilitate crossing through researchers have adopted physical, chemical, molecular methods. This review summarizes characterization targeting strategies for modification, challenges future engineering against clinical applications, promote bacteria.

Language: Английский

---

Open-endedness in synthetic biology: A route to continual innovation for biological design

Science Advances, Journal Year: 2024, Volume and Issue: 10(3)

Published: Jan. 19, 2024

Design in synthetic biology is typically goal oriented, aiming to repurpose or optimize existing biological functions, augmenting with new-to-nature capabilities, creating life-like systems from scratch. While the field has seen many advances, bottlenecks complexity of built are emerging and designs that function lab often fail when used real-world contexts. Here, we propose an open-ended approach design, novelty designed being at least as important how well it fulfils its goal. Rather than solely focusing on optimization toward a single best designing mind may allow us move beyond diminishing returns see performance for most engineered biology. Research artificial life community demonstrated embracing can automatically generate innovative unexpected solutions challenging problems local optima. Synthetic offers ideal playground explore more creative approaches design.

Language: Английский

---

Artificial Intelligence Technology Assists Enzyme Prediction and Rational Design

Journal of Agricultural and Food Chemistry, Journal Year: 2025, Volume and Issue: unknown

Published: March 11, 2025

Since the structure of enzymes determines their function, elucidating lays a solid foundation for deciphering catalytic mechanism and enabling rational design. The development artificial intelligence (AI) has sparked technological revolution, infusing new vitality into theoretical studies enzymology advancement enzyme engineering techniques. This Review outlines process main methods AI applied in structural elucidation functional prediction enzymes. Furthermore, it emphasizes AI-based design provides detailed exposition representative algorithms case studies. With support technology, comprehension function relationship will become deeper more efficient, thereby promoting widespread application various fields.

Language: Английский

---

Synthetic biology: Learning the way toward high-precision biological design

PLoS Biology, Journal Year: 2023, Volume and Issue: 21(4), P. e3002116 - e3002116

Published: April 26, 2023

Since its inception, synthetic biology has overcome many technical barriers but is at a crossroads for high-precision biological design. Devising ways to fully utilize big data may be the key achieving greater heights in biology.

Language: Английский

---

Activity‐Based Protein Profiling – Finding General Solutions to Specific Problems

Israel Journal of Chemistry, Journal Year: 2023, Volume and Issue: 63(3-4)

Published: March 1, 2023

In this retrospective/perspective, I will share thoughts on developing and applying the activity-based protein profiling (ABPP) technology, an endeavor that has consumed much of our lab's attention over 25+ year existence. Before doing so, first wish to thank colleagues who so kindly contributed Special Issue. am appreciative humbled they were willing their innovative impactful science in format.

Language: Английский

---

Improved structure-related prediction for insufficient homologous proteins using MSA enhancement and pre-trained language model

Briefings in Bioinformatics, Journal Year: 2023, Volume and Issue: 24(4)

Published: June 15, 2023

Abstract In recent years, protein structure problems have become a hotspot for understanding folding and function mechanisms. It has been observed that most of the works rely on benefit from co-evolutionary information obtained by multiple sequence alignment (MSA). As an example, AlphaFold2 (AF2) is typical MSA-based tool which famous its high accuracy. consequence, these methods are limited quality MSAs. Especially orphan proteins no homologous sequence, performs unsatisfactorily as MSA depth decreases, may pose barrier to widespread application in mutation design there rich sequences rapid prediction needed. this paper, we constructed two standard datasets de novo insufficient/none homology information, called Orphan62 Design204, respectively, fairly evaluate performance various case. Then, depending whether or not utilizing scarce summarized approaches, MSA-enhanced MSA-free methods, effectively solve issue without sufficient model aims improve poor data source knowledge distillation generation models. directly learns relationship between residues enormous pre-trained models, bypassing step extracting residue pair representation MSA. Next, evaluated four (trRosettaX-Single, TRFold, ESMFold ProtT5) (Bagging MSA) method compared with traditional AlphaFold2, structure-related tasks, respectively. Comparison analyses show trRosettaX-Single belong can achieve fast ($\sim\! 40$s) comparable AF2 tertiary prediction, especially short peptides, $\alpha $-helical segments targets few sequences. Bagging enhancement improves accuracy our trained base when exists secondary prediction. Our study provides biologists insight how select appropriate tools enzyme engineering peptide drug development. Contact [email protected], [email protected]

Language: Английский

---

Ongoing shuffling of protein fragments diversifies core viral functions linked to interactions with bacterial hosts

Nature Communications, Journal Year: 2023, Volume and Issue: 14(1)

Published: Nov. 28, 2023

Biological modularity enhances evolutionary adaptability. This principle is vividly exemplified by bacterial viruses (phages), which display extensive genomic modularity. Phage genomes are composed of independent functional modules that evolve separately and recombine in various configurations. While phages has been extensively studied, less attention paid to protein modularity-proteins consisting distinct building blocks can recombine, enhancing genetic diversity. Here, we use a set 133,574 representative phage proteins highly sensitive homology detection capture instances domain mosaicism, defined as fragment sharing between two otherwise unrelated proteins, understand its relationship with diversity genomes. We discover from diverse classes frequently share homologous domains. phenomenon particularly pronounced within receptor-binding endolysins, DNA polymerases. also identify multiple recent diversification via shuffling neck passage structures, endolysins some members the core replication machinery, often transcending distant taxonomic ecological boundaries. Our findings suggest ongoing reflective co-evolutionary arms race, driven need overcome resistance mechanisms against phages.

Language: Английский

---

Functionally Diverse Peroxygenases by AlphaFold2, Design, and Signal Peptide Shuffling

ACS Catalysis, Journal Year: 2024, Volume and Issue: 14(7), P. 4738 - 4748

Published: March 13, 2024

Unspecific peroxygenases (UPOs) are fungal enzymes that attract significant attention for their ability to perform versatile oxyfunctionalization reactions using H2O2. Unlike other oxygenases, UPOs do not require additional reductive equivalents or electron transfer chains complicate basic and applied research. Nevertheless, generally exhibit low no heterologous production levels only four UPO structures have been determined date by crystallography limiting usefulness obstructing To overcome this bottleneck, we implemented a workflow applies PROSS stability design AlphaFold2 model of 10 unique diverse followed signal peptide shuffling enable production. Nine were functionally produced in Pichia pastoris, including the recalcitrant CciUPO three derived from oomycetes─the first nonfungal be experimentally characterized. We conclude high accuracy reliability new modeling workflows dramatically expand pool

Language: Английский

---

Computational design of highly active de novo enzymes

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Aug. 3, 2024

Custom designed enzymes can further enhance the use of biocatalysts in industrial biotransformations, thereby helping to tackle biotechnological challenges 21st century. We present rotamer inverted fragment finder - diffusion (Riff-Diff) a hybrid machine learning and atomistic modeling strategy for scaffolding catalytic arrays de novo protein backbones with custom substrate pockets. used Riff-Diff scaffold tetrad capable efficiently catalyzing retro-aldol reaction. Functional designs exhibit high fold diversity, pockets similar natural enzymes. Some thus generated show activities rivaling those optimized by in-vitro evolution. The design can, principle, be applied any catalytically competent amino acid constellation. These findings are paving way address factors practical applicability catalysts processes shed light on fundamental principles enzyme catalysis.

Language: Английский

---