DR-BERT

DR-BERT webserver overview

DR-BERT predicts intrinsic protein disorder from amino acid sequences. ProteinIQ runs DR-BERT version 1.1.0 and returns continuous disorder scores from 0 to 1 for each scored residue, together with the native score file, a CSV table, a score profile, the prepared input, and a run log.

The scoring program evaluates the first 1022 residues of each submitted sequence. ProteinIQ does not add a disorder classification threshold to the continuous source scores.

Pricing

Each job costs 10 credits under the current limit of five sequences per job. Sequence count, sequence length, and the optional job name do not change this charge. The exact quote is shown before submission.

Inputs

Input	Accepted values	Limits and behavior
Protein sequences	Raw one-letter amino acid sequence or FASTA	Submit 1 to 5 non-empty sequences in one input. Raw input is converted to uppercase and has whitespace removed. Complete FASTA headers are retained.
File upload	.fasta, .fa, or .txt	Maximum file size is 10 MiB. Each FASTA record must contain sequence data.
RCSB fetch	PDB ID	Protein sequences can be fetched from RCSB and submitted as FASTA.

Sequences longer than 1022 residues are accepted, but only the first 1022 residues receive scores. The returned prepared FASTA and native pickle retain the complete submitted sequence.

Settings

Parameter	Type	Default	Description
Job name	string	optional	Adds a user-defined name to the job. It does not affect scoring or credit cost.

DR-BERT has no user-adjustable model or threshold settings.

Outputs

The Scores view contains one row per safely mapped residue.

Field	Description
sequence_id	FASTA header or submitted sequence label. Very long identifiers may be shortened in the table and CSV; the native pickle retains the complete identifier.
position	One-based residue position within the sequence.
residue	One-letter amino acid at that position.
disorder_score	Continuous DR-BERT disorder score from 0 to 1.

The Files view includes every returned artifact.

File	Format	Description
dr_bert_scores.pkl	Pickle	Native DR-BERT DataFrame with ID, sequence, and score columns. Each score value is a NumPy float32 array.
dr_bert_residue_scores.csv	CSV	Portable per-residue table derived from the native pickle without rounding scores.
dr_bert_score_profile.svg	SVG	Score profile by sequence position for every submitted sequence.
dr_bert_input.fasta	FASTA	Exact FASTA passed to the DR-BERT scoring program.
run.log	Log	Source and model revisions, runtime versions, input and scored residue counts, truncation count, mapping warnings, and returned artifact list.

If DR-BERT returns a score count that cannot be mapped reliably to a submitted sequence, ProteinIQ preserves the native pickle and score profile but omits that sequence from the derived table and CSV. The result includes a warning instead of assigning scores to the wrong residues.

Understanding results

For a worked explanation of the scores, long-sequence handling, and validation, read How to use DR-BERT online.

Higher disorder_score values indicate stronger model support for intrinsic disorder. The score remains a continuous model output rather than a binary classification. Use the profile alongside experimental evidence and other predictors when selecting construct boundaries or interpreting functional regions.