Entropy-Weighted Collapse Likelihood
EWCL
Interpretable residue-level prediction of intrinsic disorder.
EWCL connects explicit sequence physicochemistry, AlphaFold confidence, and local structural geometry to produce inspectable residue-level disorder profiles.
Example analyses
EWCL disorder likelihood0–1 residue score
0.0 · order-supporting0.5 · transition1.0 · disorder-supporting
Why disorder matters
Proteins do not always function through one stable structure.
Intrinsically disordered proteins and regions sample ensembles rather than settling into a single fold. Their sequence-encoded flexibility enables signaling, molecular recognition, regulation, and biomolecular condensation.
01 · compact globular stateUbiquitinPDB 1UBQ · 76 residuesExperimental X-ray structure · 1.8 Å
02 · structure-aware disorder exampleBRCA1UniProt P38398 · 1,863 residuesFull-length AlphaFold DB v6 model
AF-P38398-F1Full-length model · EWCL-Structure residue coloringRender publication endpoint · 1,863 residue scores
Ubiquitin is shown from the experimental PDB 1UBQ coordinates and colored with a 76-residue EWCL-Structure score vector. BRCA1 is a mixed-architecture protein—not a wholly disordered chain. Its full-length AlphaFold model is shown as a confidence-aware structural reference colored by EWCL output; it is not presented as an experimental structure or as one validated conformation of the central IDR.
What EWCL measures
Sequence physics becomes residue-level evidence.
EWCL transforms explicit sequence-derived physicochemistry into a continuous prediction, then aligns that profile with AlphaFold confidence and local structural geometry. Every output remains connected to interpretable evidence.
01 · input
Sequence
Primary amino-acid context
PRMPEAAPPVAPAPAAPTPAAPAPAPSWPP04637 · residues 64–9202 · representation
Physicochemical representation
Windowed, multichannel descriptors
03 · inference
Residue-level model
Explicit features at i, w25, w50, w100
xif(X)ŷi
Model · EWCL-Structure · publication release 2026.0704 · output
Disorder profile
EWCL disorder likelihood
Range · 0–1 / threshold · 0.5005 · comparison
AlphaFold confidence
AlphaFold pLDDT
Range · 0–100Analysis access
Choose your analysis workflow
Four routes into the same residue-resolved EWCL framework.
01
SequenceIndividual sequence-only prediction
Submit a protein sequence or UniProt accession for an individual residue-level EWCL-Sequence prediction. No structure file is required.
02StructureIndividual structure-aware analysis
Analyze an individual protein using its sequence together with a compatible PDB or AlphaFold structure. This enables EWCL-Structure, pLDDT comparison, local geometry, and Confidence–Disorder Conflict analysis.
03BatchUp to 50 proteins per submission
Submit a multi-FASTA file containing up to 50 proteins. Batch processing returns residue-level predictions and downloadable result files for each sequence.
04APIDirect programmatic integration
Integrate EWCL into computational pipelines through direct API requests to the deployed prediction service.
01Explicit sequence features
Inspectable physicochemical descriptors across multiple residue windows.
02Residue-level prediction
A continuous disorder likelihood for every position in the sequence.
03Feature-family attribution
Model evidence organized by physical and compositional feature family.
04AlphaFold conflict analysis
Localizes agreement and confidence–disorder conflict regions.
Analysis preview
From score to evidence.
Inspect the continuous prediction against structural confidence, curated disorder, experimentally observed structure, and model-level feature evidence—without losing residue coordinates.
Preview data are loaded from the bundled EWCL example analyses. Evidence tracks use the corresponding accession’s DisProt, MobiDB, observed-structure, and AlphaFold records.