Structure Prediction Benchmark · AlQuraishi Lab

ProteinNet

Sequences, structures, alignments and time-reset train/validation/test splits built around the CASP experiments.

Use it with the research agent Project page Paper

What ProteinNet is

ProteinNet exists because structure-prediction results were not comparable: every group built its own splits, and a model trained on structures released after a benchmark was set had already seen the answers. It uses the CASP targets from CASP7 to CASP12 as test sets and reconstructs the sequence and structure databases as they stood before each experiment began, so a method evaluated on it is evaluated under the conditions the original entrants faced. Each release carries sequences, secondary and tertiary structure, multiple sequence alignments and position-specific scoring matrices, with validation subsets graded from easy to very hard by sequence identity. Records come as human-readable text and as TensorFlow TFRecords; the raw alignment data behind ProteinNet 12 is 4 TB and the authors release it on request rather than as a download.

Get the data

Its authors publish ProteinNet outside the Hugging Face Hub, so the dataset explorer has no table to read and there are no rows on this page. The files are linked below, and the research agent can fetch them onto its machine and work through them there.

Try asking

“Download one ProteinNet CASP12 text record archive on the machine and describe the fields in a single entry.”

Open in the research agent

Licence

Commercial use permitted · MIT

Read 2026-10-24: the repository that publishes and documents the records declares MIT. The underlying coordinates come from the Protein Data Bank, which places no restriction on reuse; the CASP target sets belong to the Prediction Center.

Questions

Can I browse ProteinNet in the browser?

No. Its authors publish it outside the Hugging Face Hub, so the dataset explorer has no table to read and there are no rows to show here (format: Text records and TensorFlow TFRecords, one archive per CASP release). This page describes what it holds and links to the files; the research agent can fetch and read them on its machine.

What licence is ProteinNet under?

MIT. Read 2026-10-24: the repository that publishes and documents the records declares MIT. The underlying coordinates come from the Protein Data Bank, which places no restriction on reuse; the CASP target sets belong to the Prediction Center.

How to cite it

AlQuraishi, M. ProteinNet: a standardized data set for machine learning of protein structure. BMC Bioinformatics 20, 311 (2019).