Multistate and functional protein design using RoseTTAFold sequence space diffusion

Nature Biotechnology, Год журнала: 2024, Номер unknown

Опубликована: Сен. 25, 2024

Protein denoising diffusion probabilistic models are used for the de novo generation of protein backbones but limited in their ability to guide proteins with sequence-specific attributes and functional properties. To overcome this limitation, we developed ProteinGenerator (PG), a sequence space model based on RoseTTAFold that simultaneously generates sequences structures. Beginning from noised representation, PG structure pairs by iterative denoising, guided desired structural attributes. We designed thermostable varying amino acid compositions internal repeats cage bioactive peptides, such as melittin. By averaging logits between trajectories distinct constraints, multistate parent-child triples which same folds different supersecondary structures when intact parent versus split into two child domains. design can be experimental sequence-activity data, providing general approach integrated computational optimization function.

Язык: Английский

---

Easy and accurate protein structure prediction using ColabFold

Nature Protocols, Год журнала: 2024, Номер unknown

Опубликована: Окт. 14, 2024

Язык: Английский

---

Proteins with alternative folds reveal blind spots in AlphaFold-based protein structure prediction

Current Opinion in Structural Biology, Год журнала: 2025, Номер 90, С. 102973 - 102973

Опубликована: Янв. 5, 2025

In recent years, advances in artificial intelligence (AI) have transformed structural biology, particularly protein structure prediction. Though AI-based methods, such as AlphaFold (AF), often predict single conformations of proteins with high accuracy and confidence, predictions alternative folds are inaccurate, low-confidence, or simply not predicted at all. Here, we review three blind spots that reveal about AF-based First, assume distinct from their training-set homologs can be mispredicted. Second, AF overrelies on its training set to conformations. Third, degeneracies pairwise representations lead high-confidence inconsistent experiment. These weaknesses suggest approaches more reliably.

Язык: Английский

---

Protein folds vs. protein folding: Differing questions, different challenges

Proceedings of the National Academy of Sciences, Год журнала: 2022, Номер 120(1)

Опубликована: Дек. 29, 2022

Microbial communities are found throughout the biosphere, from human guts to glaciers, soil activated sludge. Understanding statistical properties of such diverse can pave way elucidate common mechanisms ...Multiple ecological forces act together shape composition microbial communities. Phyloecology approaches—which combine phylogenetic relationships between species with community ecology—have potential disentangle but often ...

Язык: Английский

---

Evolutionary selection of proteins with two folds

Nature Communications, Год журнала: 2023, Номер 14(1)

Опубликована: Сен. 6, 2023

Although most globular proteins fold into a single stable structure, an increasing number have been shown to remodel their secondary and tertiary structures in response cellular stimuli. State-of-the-art algorithms predict that these fold-switching adopt only one missing functionally critical alternative folds. Why is unclear, but all of them infer protein structure from coevolved amino acid pairs. Here, we hypothesize coevolutionary signatures are being missed. Suspecting single-fold variants could be masking signatures, developed approach, called Alternative Contact Enhancement (ACE), search both highly diverse superfamilies-composed variants-and subfamilies with more variants. ACE successfully revealed coevolution pairs uniquely corresponding conformations 56/56 distinct families. Then, used ACE-derived contacts (1) two experimentally consistent candidate unsolved (2) develop blind prediction pipeline for proteins. The discovery widespread dual-fold indicates sequences preserved by natural selection, implying functionalities provide evolutionary advantage paving the way predictions sequences.

Язык: Английский

---

3D-equivariant graph neural networks for protein model quality assessment

Bioinformatics, Год журнала: 2023, Номер 39(1)

Опубликована: Янв. 1, 2023

Abstract Motivation Quality assessment (QA) of predicted protein tertiary structure models plays an important role in ranking and using them. With the recent development deep learning end-to-end prediction techniques for generating highly confident structures most proteins, it is to explore corresponding QA strategies evaluate select structural by them since these have better quality different properties than traditional methods. Results We develop EnQA, a novel graph-based 3D-equivariant neural network method that equivariant rotation translation 3D objects estimate accuracy leveraging features acquired from state-of-the-art method—AlphaFold2. train test on both model datasets (e.g. Critical Assessment Techniques Protein Structure Prediction) new dataset high-quality only AlphaFold2 proteins whose experimental were released recently. Our approach achieves performance methods latest It performs even scores provided itself. The results illustrate graph promising evaluation models. Integrating with other complementary sequence improving QA. Availability implementation source code available at https://github.com/BioinfoMachineLearning/EnQA. Supplementary information data are Bioinformatics online.

Язык: Английский

---

Exploring and Learning the Universe of Protein Allostery Using Artificial Intelligence Augmented Biophysical and Computational Approaches

Journal of Chemical Information and Modeling, Год журнала: 2023, Номер 63(5), С. 1413 - 1428

Опубликована: Фев. 24, 2023

Allosteric mechanisms are commonly employed regulatory tools used by proteins to orchestrate complex biochemical processes and control communications in cells. The quantitative understanding characterization of allosteric molecular events among major challenges modern biology require integration innovative computational experimental approaches obtain atomistic-level knowledge the states, interactions, dynamic conformational landscapes. growing body studies empowered emerging artificial intelligence (AI) technologies has opened up new paradigms for exploring learning universe protein allostery from first principles. In this review we analyze recent developments high-throughput deep mutational scanning functions; applications latest adaptations Alpha-fold structural prediction methods dynamics allostery; frontiers integrating machine enhanced sampling techniques advances systems. We also highlight SARS-CoV-2 spike (S) revealing an important often hidden role regulation driving functional changes, binding interactions with host receptor, escape S which critical viral infection. conclude a summary outlook future directions suggesting that AI-augmented biophysical computer simulation beginning transform toward systematic landscapes, may bring about revolution drug discovery.

Язык: Английский

---

Modeling Flexible Protein Structure With AlphaFold2 and Crosslinking Mass Spectrometry

Molecular & Cellular Proteomics, Год журнала: 2024, Номер 23(3), С. 100724 - 100724

Опубликована: Янв. 22, 2024

We propose a pipeline that combines AlphaFold2 (AF2) and crosslinking mass spectrometry (XL-MS) to model the structure of proteins with multiple conformations. The consists two main steps: ensemble generation using AF2 conformer selection XL-MS data. For selection, we developed scores—the monolink probability score (MP) crosslink (XLP)—both which are based on residue depth from protein surface. benchmarked MP XLP large dataset decoy structures showed our scores outperform previously scores. then tested methodology three having an open closed conformation in Protein Data Bank: Complement component 3 (C3), luciferase, glutamine-binding periplasmic protein, first generating ensembles AF2, were screened for conformations experimental In five out six cases, most accurate within ensembles—or 1 Å this model—was identified crosslinks, as assessed through score. remaining case, only monolinks (assessed score) successfully these results further improved by including "occupancy" monolinks. This serves compelling proof-of-concept effectiveness contrast, assessment was able identify cases. Our highlight complementarity methods like XL-MS, providing reliable metrics assess quality predicted models. scoring functions mentioned above available at https://gitlab.com/topf-lab/xlms-tools.

Язык: Английский

---

Metamorphic proteins and how to find them

Current Opinion in Structural Biology, Год журнала: 2024, Номер 86, С. 102807 - 102807

Опубликована: Март 26, 2024

In the last two decades, our existing notion that most foldable proteins have a unique native state has been challenged by discovery of metamorphic proteins, which reversibly interconvert between multiple, sometimes highly dissimilar, states. As number known increases, several computational and experimental strategies emerged for gaining insights about their refolding processes identifying unknown amongst proteome. this review, we describe current advances in biophysically functionally ascertaining structural interconversions how coevolution can be harnessed to identify novel from sequence information. We also discuss challenges ongoing efforts using artificial intelligence-based protein structure prediction methods discover predict corresponding three-dimensional structures.

Язык: Английский

---

AFsample2: Predicting multiple conformations and ensembles with AlphaFold2

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2024, Номер unknown

Опубликована: Июнь 2, 2024

Abstract Understanding protein dynamics and conformational states carries profound scientific practical implications for several areas of research, ranging from a general understanding biological processes at the molecular level to detailed disease mechanisms, which in turn can open up new avenues drug development. Multiple solutions have been recently developed widen landscape predictions made by Alphafold2 (AF2). Here, we introduce AFsample2, method employing random MSA column masking reduce influence co-evolutionary signals enhance structural diversity models generated AF2 neural network. AFsample2 improves prediction alternative broad range proteins, yielding high-quality end diverse ensembles. In data set open-closed conformations (OC23), alternate state improved 17 out 23 cases without compromising generation preferred state. Consistent results were observed 16 membrane transporters, with improvements 12 targets. TM-score experimental substantial, sometimes exceeding 50%, elevating mediocre scores 0.58 nearly perfect 0.98. Furthermore, increased intermediate 70% compared standard system, producing highly confident models, that could potentially be on-path between two states. addition, also propose way selecting end-states model These identification conformations, thereby providing more comprehensive function dynamics. Future work will focus on validating accuracy these exploring their relevance functional transitions proteins.

Язык: Английский

---