CASP14 - 14th Community Wide Experiment on Critical Assessment of Protein Structure Prediction
webCASP14 is the 14th biennial protein structure prediction competition where AlphaFold2 by DeepMind achieved a breakthrough, demonstrating transformative AI capabilities in biology and illustrating how AI benchmarks can signal major capability jumps relevant to AI safety discussions.
Metadata
Importance: 55/100homepage
Summary
CASP14 is the 2020 edition of the biennial Critical Assessment of Structure Prediction competition, where ~100 research groups worldwide submitted over 67,000 protein structure models across 90 targets. This round became historically significant as DeepMind's AlphaFold2 achieved near-experimental accuracy, representing a landmark AI capability breakthrough. Results were published in a special issue of the journal Proteins in 2021.
Key Points
- •AlphaFold2 achieved a landmark breakthrough in protein structure prediction at CASP14, solving a 50-year-old grand challenge in biology.
- •Nearly 100 research groups submitted over 67,000 models on 90 protein targets, making it a large-scale community benchmark.
- •CASP14 received widespread media coverage (Nature, Science, NYT, BBC) highlighting the significance of the AI capability jump.
- •The competition runs every two years and provides independent, blind assessment of protein modeling methods against experimental structures.
- •Results demonstrate how AI benchmarks can serve as clear signals of transformative capability advances with real-world scientific impact.
1 FactBase fact citing this source
Cached Content Preview
HTTP 200Fetched Jun 15, 202615 KB
Home - CASP14
14th Community Wide Experiment on the
Critical Assessment of Techniques for Protein Structure Prediction
-->
Menu
CASP14
CASP provides an independent mechanism for the assessment of methods of protein structure modeling. From May through August 2020, CASP organizers have been posting on this website sequences of unknown protein structures for modeling. Protein models have been collected from May through mid-September, and evaluated as the experimental coordinates become available. In the summer and fall, the tens of thousands of models submitted by approximately 100 research groups worldwide are processed and evaluated. Independent assessors in each of the prediction categories bring independent insight into their assessment. Tools for viewing, comparison, and analysis of submitted models are available from this website.
Targets
Predictors
Conference
Results
CASP14 in news
Target List
Domain Definition
Target Submission
Submit a prediction
Model Viewer
Archive: Server Predictions
Archive: All Predictions
Groups Info
Register [external link]
Abstracts
Abstracts
Abstract book will be available here in the end of October.
Abstract submission
Abstract submission deadline - September 15, 2020
Program
Meeting program will be available here in the end of November.
Snapshot of the Conference Program
Program
Presentations
Recordings
CASP14 Conference Platforms
AUTOMATIC EVALUATION
CASP14 results will be published
in a special edition of Proteins in 2021.
Parseable Data
Ranking tables
will be available here on Nov 30.
Rankings: Regular targets (T)
Rankings: Multimeric targets (H,To)
Rankings: Inter-domain prediction
Rankings: Refinement targets (R)
Rankings: Contact predictions
CASP Press Release
Nature
Science
New York Times
BBC news
Fortune
CNBC news
Bloomberg
Financial Post
MIT Technology Review
The Guardian
The Telegraph
Daily Mail
Tech Crunch
Venture Beat
New Scientist
SciTech Daily
Eureka Alert
News Medical
MedCity News
The Verge
DeepMind blog
CASP14 in numbers
Detailed description of the experiment
CASP (Critical Assessment of Structure Prediction) is a community wide experiment to determine and advance the state of the art in modeling protein structure from amino acid sequence. Every two years, participants are invited to submit models for a set of proteins for which the exper
... (truncated, 15 KB total)Resource ID:
86785cb10fb47641 | Stable ID: sid_4k7LTJMVYw