E
EWCL

EWCL data release

Publication downloads

Versioned benchmark archives, full database tables, ground-truth tracks, grouped EWCL model scores, EWCL-Raw interpretability files, and external comparator scores.

Release manifest

Database tables

Residue-level data and GT tracks

EWCL database full residue table

One Parquet table with binary DisProt/MobiDB GT, observed-order GT, PDB-missing GT, EWCL publication scores, EWCL-Raw, AlphaFold pLDDT, and external predictors where present.

5,068,184 residue rows

Download Parquet

EWCL database full residue CSV

Gzipped CSV mirror of the full residue table for workflows that do not read Parquet.

Download CSV.gz

Ground-truth tracks

GT-only table with DisProt disorder, MobiDB disorder, observed-order labels, PDB-missing labels, and PDB support counts where available.

Download GT CSV

Current protein index

Search/index table for the current EWCL protein pages, including evidence availability flags used by the database filters.

9,199 proteins

Download CSV

Release upload manifest

Object map for the current database release prefix, used to verify upload coverage before switching production links.

Current release prefix: publication/ewcl_database_downloads/v01

Open manifest

PDB-missing benchmark labels

PDB-missing val/test label frame used for the structural-support benchmark.

8,991 PDB records

Parquet

Benchmark archives

External archive records

CAID benchmark archive

Curated disorder and binding benchmark tables used for CAID-style disorder, binding, binding-IDR, and linker evaluation.

DOI 10.5281/zenodo.17352525

View Zenodo

MobiDB benchmark archive

DisProt, IDEAL, MobiDB merge, priority, D-to-D binding, and UniProt disorder sources used in EWCL evaluation.

DOI 10.5281/zenodo.17409146

View Zenodo

EWCL models

Main publication model scores

EWCL-Sequence scores

Main sequence-only EWCL publication model residue scores.

9,199 proteins · 3,611,619 scored residues

Parquet

EWCL-Structure scores

Main structure-aware EWCL publication model residue scores where structure features are available.

8,450 proteins · 3,121,171 scored residues

Parquet

PDB-missing support

EWCL PDB-missing model scores

EWCL-PDB-Missing-Sequence scores

Sequence-only PDB-missing support model residue scores.

5,645 proteins · 1,423,329 scored residues

Parquet

EWCL-PDB-Missing-Structure scores

Structure-aware PDB-missing support model residue scores.

5,645 proteins · 1,423,329 scored residues

Parquet

CDC support

Sequence and structure AlphaFold consistency

CDC publication bundle

Canonical residue states and contiguous regions for sequence-only and structure-integrated CDC, including the frozen 11-residue persistent-conflict rule.

Multi-state CDC release; binary high/high masks are included only as derived summaries

ZIP

Conflict primitive summary

Derived high-EWCL/high-pLDDT conflict totals for EWCL-Sequence and EWCL-Structure across the public GT overlap frames.

DisProt/MobiDB pLDDT overlap: 3,306 proteins · 1,763,505 residues

CSV

CDC state distribution

By-model CDC state counts for EWCL-Sequence and EWCL-Structure against the public GT/pLDDT overlap.

CSV

Derived conflict segments

Derived high-EWCL/high-pLDDT segments for EWCL-Sequence and EWCL-Structure with GT-frame support; these do not replace the canonical CDC states.

CSV

EWCL-Raw

Raw feature-family interpretability

EWCL-Raw interpretability scores

Frozen EWCL-Raw residue table with combined raw-physics support and separated family evidence tracks.

9,199 source proteins · 5,068,920 source-coordinate residues; fixed-coordinate release maps 9,196 proteins

Parquet

EWCL-Raw family summaries

Per-protein summary statistics for each raw feature-family evidence track.

CSV

EWCL-Raw region calls

Compact intervals called from the EWCL-Raw and family evidence tracks.

Parquet