De novo protein design by deep network hallucination

Anishchenko, Ivan; Pellock, S.J.; Chidyausiku, Tamuka M.; Ramelot, Theresa; Овчинников, С. Г.; Hao, Jingzhou; Bafna, Khushboo; Norn, Christoffer; Kang, Alex; Bera, Asim K.; DiMaio, Frank; Carter, Lauren; Chow, Cameron M.; Montelione, Gaetano T.; Baker, David

doi:10.1038/s41586-021-04184-w

Cited by 406 publications

(373 citation statements)

References 55 publications

Supporting

Mentioning

371

Contrasting

Unclassified

Order By: Relevance

“…Machine learning has revolutionized protein structure prediction, achieving an unprecedented level of accuracy that is transforming the field. 1,2 Deep neural network hallucination 3 has already served to design novel proteins based on the idealized version of proteins the network learns. The structural biology community has rapidly used these developments to test neural networks beyond their original application to problems including alternative conformational sampling, disordered protein identification, or complex structure prediction, with very promising results 4,5 .…”

Section: Introductionmentioning

confidence: 99%

AlphaFold encodes the principles to identify high affinity peptide binders

Chang

Pérez

2022

Preprint

View full text Add to dashboard Cite

Machine learning has revolutionized structural biology by solving the problem of predicting structures from sequence information. The community is pushing the limits of interpretability and application of these algorithms beyond their original objective. Already, AlphaFold's ability to predict bound conformations for complexes has surpassed the performance of docking methods, especially for protein-peptide binding. A key question is the ability of these methods to differentiate binding affinities between several peptides that bind the same receptor. We show a novel application of AlphaFold for competitive binding of different peptides to the same receptor. For systems in which the individual structures of the peptides are well predicted, predictions in which both peptides are introduced capture the stronger binder in the bound state, and the other peptide in the unbound form. The speed and robustness of the method will be a game changer to screen large libraries of peptide sequences to prioritize for detailed experimental characterization.

show abstract

Section: Introductionmentioning

confidence: 99%

AlphaFold encodes the principles to identify high affinity peptide binders

Chang

Pérez

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Lack of asparagine synthetase in some pancreatic or breast cancer cells, along with the extended half-life and improved pharmacological properties of new asparaginases, open up the opportunity to extend the therapeutic potential of new enzymes and use them for solid cancers’ treatment. Actively developing approaches for de novo protein design by deep-learning neural networks such as trRosetta [ 174 ] or AlphaFold 2 [ 175 ] become attractive for creating proteins with specified properties. These techniques should be used to create potent artificial L-asparaginases with reduced immunogenicity, low specificity to L-glutamine and increased stability in the blood.…”

Section: Discussionmentioning

confidence: 99%

Molecular Analysis of L-Asparaginases for Clarification of the Mechanism of Action and Optimization of Pharmacological Functions

et al. 2022

View full text Add to dashboard Cite

L-asparaginases (EC 3.5.1.1) are a family of enzymes that catalyze the hydrolysis of L-asparagine to L-aspartic acid and ammonia. These proteins with different biochemical, physicochemical and pharmacological properties are found in many organisms, including bacteria, fungi, algae, plants and mammals. To date, asparaginases from E. coli and Dickeya dadantii (formerly known as Erwinia chrysanthemi) are widely used in hematology for the treatment of lymphoblastic leukemias. However, their medical use is limited by side effects associated with the ability of these enzymes to hydrolyze L-glutamine, as well as the development of immune reactions. To solve these issues, gene-editing methods to introduce amino-acid substitutions of the enzyme are implemented. In this review, we focused on molecular analysis of the mechanism of enzyme action and to optimize the antitumor activity.

show abstract

“…It was also noted that predictions would be inaccurate for species-unique protein complexes and that there would be an overrepresentation of hydrophobic or amphipathic domains. As well as this, recent approaches involving deep network hallucination have investigated the generation of new proteins with entirely novel backbone structures and amino acid sequences [ 132 ].…”

Section: Towards a Top-down Approachmentioning

confidence: 99%

Next-Generation Molecular Discovery: From Bottom-Up In Vivo and In Vitro Approaches to In Silico Top-Down Approaches for Therapeutics Neogenesis

Kenny

Antaw

Locke

et al. 2022

Life

View full text Add to dashboard Cite

Protein and drug engineering comprises a major part of the medical and research industries, and yet approaches to discovering and understanding therapeutic molecular interactions in biological systems rely on trial and error. The general approach to molecular discovery involves screening large libraries of compounds, proteins, or antibodies, or in vivo antibody generation, which could be considered “bottom-up” approaches to therapeutic discovery. In these bottom-up approaches, a minimal amount is known about the therapeutics at the start of the process, but through meticulous and exhaustive laboratory work, the molecule is characterised in detail. In contrast, the advent of “big data” and access to extensive online databases and machine learning technologies offers promising new avenues to understanding molecular interactions. Artificial intelligence (AI) now has the potential to predict protein structure at an unprecedented accuracy using only the genetic sequence. This predictive approach to characterising molecular structure—when accompanied by high-quality experimental data for model training—has the capacity to invert the process of molecular discovery and characterisation. The process has potential to be transformed into a top-down approach, where new molecules can be designed directly based on the structure of a target and the desired function, rather than performing screening of large libraries of molecular variants. This paper will provide a brief evaluation of bottom-up approaches to discovering and characterising biological molecules and will discuss recent advances towards developing top-down approaches and the prospects of this.

show abstract

De novo protein design by deep network hallucination

Cited by 406 publications

References 55 publications

AlphaFold encodes the principles to identify high affinity peptide binders

AlphaFold encodes the principles to identify high affinity peptide binders

Molecular Analysis of L-Asparaginases for Clarification of the Mechanism of Action and Optimization of Pharmacological Functions

Next-Generation Molecular Discovery: From Bottom-Up In Vivo and In Vitro Approaches to In Silico Top-Down Approaches for Therapeutics Neogenesis

Contact Info

Product

Resources

About