From literature to manuscript.
Discover, understand, connect, create: six modules that carry a study from its first lead to finished work.
01DiscoverDaily Research
Start each day with what is worth reading.
Follow the journals and preprint servers in your field and find each day's new papers gathered in one place. Recommended papers come with a note on why they matter to your work.
Daily ResearchToday
New today
Accurate structure prediction of biomolecular interactions with AlphaFold 3
Why it matters
Extends structure prediction to how proteins interact with DNA, RNA, ligands and other molecules; worth reading against the AlphaFold paper in your library.
De novo design of protein structure and function with RFdiffusion
Accurate proteome-wide missense variant effect prediction with AlphaMissense
Sources
- arXiv · q-bio.BM
- bioRxiv · Bioinformatics
- Nature · Structural biology
- Science · Computational biology
02DiscoverLibrary
Every paper in its place.
Papers, metadata and projects live together, with type, journal, volume, pages and DOI at hand. Bring a Zotero library across, or import a single paper or a whole folder.
Protein structure
ZOne-Time Zotero ImportTitleAuthorsYearHighly accurate protein structure prediction with AlphaFoldJumper et al. · Nature2021Highly accurate protein structure prediction for the human proteomeTunyasuvunakool et al. · Nature2021Accurate prediction of protein structures and interactions using a three-track neural networkBaek et al. · Science2021Evolutionary-scale prediction of atomic-level protein structure with a language modelLin et al. · Science2023Improved protein structure prediction using potentials from deep learningSenior et al. · Nature2020Highly accurate protein structure prediction with AlphaFoldJournal Article
- Type
- Journal Article
- Authors
- Jumper, JohnEvans, RichardPritzel, AlexanderGreen, Tim…
- Journal
- Nature
- Published
- 2021
- Volume
- 596
- Pages
- 583–589
- DOI
- 10.1038/s41586-021-03819-2
03UnderstandInsight
See the structure before the detail.
Before the close reading, grasp a paper's research question, method, contribution and limitations. Once you know what it solves, what it rests on and where it stops, the details have somewhere to go.
Journal ArticleHighly accurate protein structure prediction with AlphaFold
Article
Highly accurate protein structure prediction with AlphaFold
Proteins are essential to life, and understanding their structure can facilitate a mechanistic understanding of their function. Through an enormous experimental effort, the structures of around 100,000 unique proteins have been determined, but this represents a small fraction of the billions of known protein sequences. Here we provide the first computational method that can regularly predict protein structures with atomic accuracy even in cases in which no similar structure is known.
Insight
Research question
Can protein structure be predicted from sequence with atomic accuracy, even when no similar structure is known?
Method
A deep learning model that builds physical and biological knowledge of protein structure into its design and draws on multiple sequence alignments.
Contribution
In CASP14, accuracy was competitive with experimental structures in a majority of cases, well ahead of other methods.
Limitations
Accuracy drops markedly when few related sequences are available (shallow alignments).
04UnderstandMentor
Every answer has a source.
Ask about the paper in front of you: Mentor answers from the text and marks the sentences it relies on. Explanation and evidence stay matched, so you can always check against the original.
Journal ArticleHighly accurate protein structure prediction with AlphaFold
Article
Highly accurate protein structure prediction with AlphaFold
Proteins are essential to life, and understanding their structure can facilitate a mechanistic understanding of their function. Through an enormous experimental effort, the structures of around 100,000 unique proteins have been determined, but this represents a small fraction of the billions of known protein sequences. Here we provide the first computational method that can regularly predict protein structures with atomic accuracy even in cases in which no similar structure is known.
Paper Mentor
Explain the paper, trace the evidence, open the field.
Current Paper
Highly accurate protein structure prediction with AlphaFold
What gap does this paper address?
Response
Experiments have determined the structures of about 100,000 unique proteins, a small fraction of the billions of known sequences 1. The paper presents a computational method that predicts structures with atomic accuracy even when no similar structure is known 2.
Evidence
1“Through an enormous experimental effort, the structures of around 100,000 unique proteins have been determined, but this represents a small fraction of the billions of known protein sequences.”
2“Here we provide the first computational method that can regularly predict protein structures with atomic accuracy even in cases in which no similar structure is known.”
Ask next
How is accuracy measured?
Which evidence supports the claim?
Ask about this paper…
05ConnectRecall
What you read stays within reach.
Concepts, methods, findings and research questions settle out of your reading, each tied to its exact evidence and source. AI-generated candidates become knowledge only after your review, and return whenever you need them.
Knowledge
RecallRelated to the paper you are reading
Evolutionary-scale prediction of atomic-level protein structure with a language model
- FindingExperimental structures cover only a small fraction of known protein sequences.Jumper et al. · 2021
- QuestionWhat gap does this paper address?Jumper et al. · 2021
- FindingPrediction reaches atomic accuracy even without a similar known structure.Jumper et al. · 2021
FindingExperimental structures cover only a small fraction of known protein sequences.
Exact Evidence
“Through an enormous experimental effort, the structures of around 100,000 unique proteins have been determined, but this represents a small fraction of the billions of known protein sequences.”
Jumper et al. · Nature · 2021
Related Knowledge
06CreateResearch Workbench
Write what you understand into research of your own.
Organize manuscripts by project, write in the editor, and build and preview the PDF as you go. The research assistant checks and suggests alongside you; when you are done, export a single file or the whole project.
Research Workbench
main.texreferences.bib1\documentclass{article} 2\usepackage{natbib} 3\title{Predicted structures at proteome scale} 4\begin{document} 5\maketitle 6 7\section{Introduction} 8Experiments have resolved around 100,000 unique 9protein structures, a small fraction of the billions 10of known sequences~\cite{jumper2021}. Predicted 11structures now cover nearly the entire human 12proteome~\cite{jumper2021}. 13 14\section{Methods} 15We compare predicted and experimental structures 16for every entry in the benchmark set. 17 18\bibliographystyle{unsrtnat} 19\bibliography{references} 20\end{document}
Preview · main.pdf
Predicted structures at proteome scale
1 Introduction
Experiments have resolved around 100,000 unique protein structures, a small fraction of the billions of known sequences [1]. Predicted structures now cover nearly the entire human proteome [1].
2 Methods
We compare predicted and experimental structures for every entry in the benchmark set.
References
[1] J. Jumper et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589 (2021).
Research Assistant
The proteome-scale claim cites Jumper et al. (2021). That result is reported by Tunyasuvunakool et al. (2021); consider citing it instead.