Learned protein embeddings for machine learning

Yang, Kevin; Wu, Zachary; Bedbrook, Claire N.; Arnold, Frances H.

doi:10.1093/bioinformatics/bty178

Cited by 271 publications

(235 citation statements)

References 21 publications

Supporting

Mentioning

234

Contrasting

Order By: Relevance

“…There has been growing interest in using neural network-based approaches for protein representation learning and design (19,(37)(38)(39)(40), with several new methods reported during the preparation of this manuscript (41,42). While most methods are accompanied by a variety of metrics which attempt to illustrate the accuracy of the predictions, it is inherently difficult to evaluate the quality of generative models, as their ultimate goal is to generate entirely novel sequences with no existing counterparts.…”

Section: Resultsmentioning

confidence: 99%

Fast and flexible design of novel proteins using graph neural networks

Strokach¹,

Becerra

Corbi‐Verge

et al. 2019

Preprint

View full text Add to dashboard Cite

Protein structure and function is determined by the arrangement of the linear sequence of amino acids in 3D space. Despite substantial advances, precisely designing sequences that fold into a predetermined shape (the "protein design" problem) remains difficult. We show that a deep graph neural network, ProteinSolver, can solve protein design by phrasing it as a constraint satisfaction problem (CSP). To sidestep the considerable issue of optimizing the network architecture, we first develop a network that is accurately able to solve the related and straightforward problem of Sudoku puzzles. Recognizing that each protein design CSP has many solutions, we train this network on millions of real protein sequences corresponding to thousands of protein structures. We show that our method rapidly designs novel protein sequences and perform a variety of in silico and in vitro validations suggesting that our designed proteins adopt the predetermined structures.One Sentence Summary: A neural network optimized using Sudoku puzzles designs protein sequences that adopt predetermined structures.

show abstract

Section: Resultsmentioning

confidence: 99%

Fast and flexible design of novel proteins using graph neural networks

Strokach¹,

Becerra

Corbi‐Verge

et al. 2019

Preprint

View full text Add to dashboard Cite

show abstract

“…[42] At hird challenge involves the use of machine learning models trained on polymer membrane formation parameters like concentration of polymer,s olvent and nonsolvent and their measured resultant pore size distributions and filtration results to infer optimal formations parameters without requiring an understanding of the underlying physical or chemical mechanisms. [43] 3. Mass Transfer Aspects 3.1.…”

Section: Unresolved Challengesmentioning

confidence: 99%

Membrane Filtration with Liquids: A Global Approach with Prior Successes, New Developments and Unresolved Challenges

Belfort

2018

Angew Chem Int Ed

View full text Add to dashboard Cite

After 70 years, modern pressure‐driven polymer membrane processes with liquids are mature and accepted in many industries due to their good performance, ease of scale‐up, low energy consumption, modular compact construction, and low operating costs compared with thermal systems. Successful isothermal operation of synthetic membranes with liquids requires consideration of three critical aspects or “legs” in order of relevance: selectivity, capacity (i.e. permeation flow rate per unit area) and transport of mass and momentum comprising concentration polarization (CP) and fouling (F). Major challenges remain with respect to increasing selectivity and controlling mass transport in, to and away from membranes. Thus, prediction and control of membrane morphology and a deep understanding of the mechanism of dissolved and suspended solute transport near and in the membrane (i.e. diffusional and convective mass transport) is essential. Here, we focus on materials development to address the relatively poor selectivity of liquid membrane filtration with polymers and discuss the critical aspects of transport limitations. Machine learning could help optimize membrane structure design and transport conditions for improved membrane filtration performance.

show abstract

“…Various methods to create embeddings for proteins are proposed (Asgari and Mofrad, 2015;Yang et al, 2018;Bepler and Berger, 2019;Asgari et al, 2019;Heinzinger et al, 2019;Alley et al, 2019). ProtVec fragmented the protein sequence in 3-mers for all possible starting shifts, then learned embeddings for each 3-mer and represented the respective protein as the average of its constituting 3-mer vectors (Asgari and Mofrad, 2015).…”

Section: Introductionmentioning

confidence: 99%

dom2vec: Unsupervised protein domain embeddings capture domains structure and function providing data-driven insights into collocations in domain architectures

Malone²,

Nejdl

2020

Preprint

View full text Add to dashboard Cite

Motivation: Word embedding approaches have revolutionized Natural Lange Processing (NLP) research. These approaches aim to map words to a low-dimensional vector space in which words with similar linguistic features are close in the vector space. These NLP approaches also preserve local linguistic features, such as analogy. Embedding-based approaches have also been developed for proteins. To date, such approaches treat amino acids as words, and proteins are treated as sentences of amino acids. These approaches have been evaluated either qualitatively, via visual inspection of the embedding space, or extrinsicaly, via performance on a downstream task. However, it is difficult to directly assess the intrinsic quality of the learned embeddings. Results: In this paper, we introduce dom2vec, an approach for learning protein domain embeddings. We also present four intrinsic evaluation strategies which directly assess the quality of protein domain embeddings. We leverage the hierarchy relationship of InterPro domains, known secondary structure classes, Enzyme Commision class information, and Gene Ontology annotations in these assessments. Importantly, these evaluations allow us to assess the quality of learned embeddings independently of a particular downstream task, similar to the local linguistic features used in traditional NLP. We also show that dom2vec embeddings are competitive with, or even outperform, state-of-the-art approaches on downstream tasks. Availability: The protein domain embeddings vectors and the entire code to reproduce the results will become available. Contact: melidis@l3s.uni-hannover.de

show abstract

Learned protein embeddings for machine learning

Abstract: Supplementary data are available at Bioinformatics online.

Cited by 271 publications

References 21 publications

Fast and flexible design of novel proteins using graph neural networks

Fast and flexible design of novel proteins using graph neural networks

Membrane Filtration with Liquids: A Global Approach with Prior Successes, New Developments and Unresolved Challenges

dom2vec: Unsupervised protein domain embeddings capture domains structure and function providing data-driven insights into collocations in domain architectures

Contact Info

Product

Resources

About