McWhite Lab

Research previewPreliminary · work in progress

A blinded language-model screen for repeats in human disordered proteins

Repeated motifs in intrinsically disordered regions are hard to find by alignment: the copies are short, differ from one another, and are spread out along the sequence. We gave a language model anonymized human protein sequences, with names and annotations removed, and asked it to describe any recurring pattern as a consensus together with the copies it was based on. A script then located each reported copy in the sequence and discarded any that could not be found. Set against UniProt modification sites and ELM motif annotations, the screen places known sites and motifs at a fixed position in the repeat unit, finds the same motifs in arrays that have not been curated, and turns up positional modification patterns with no description.

proteins screened, each with ≥100 aa of predicted disorder
with at least one repeat family
repeat families, every copy located in the sequence
located repeat copies

Highlights

Highlights are grouped by novelty: repeated patterns with no description, new occurrences of known motifs, and known repeats the screen recovers blind. Within each group, families with a UniProt modification at a fixed unit position lead, because a modification at the same position in most copies is independent support that the unit is real. Each family gets an evidence badge for the modification it is about, a literature badge from a PubMed abstract sweep, and one of three ELM calls: curated in the array, curated elsewhere in the protein, or a protein ELM has never curated (87% of families, where ELM's silence says nothing). Repeat architectures follow as a separate set of examples. Select any family to see its AlphaFold model, its copies against UniProt annotations, and the modified unit position marked in each copy.

Color
Sequence logos and per-position modification frequency for the ten headline repeat families
Figure 2. The ten headline families. Each panel shows a sequence logo of the located copies (bits) and, above it, the percentage of copies carrying a UniProt modification at each unit position (phosphorylation, O-glycosylation, methylation, sumoylation) or, for ELM cases, a class regex match and curated instances. Logos are computed without a pseudocount; information content is biased upwards when there are few copies, so n is given on each panel. Click to enlarge. Panel labels reflect the categories when the figure was made; ZNHIT6, for example, is now classed as a new occurrence of the SUMO consensus (MOD_SUMO_for_1).
Sequence logos for sixteen further highlighted repeat families
Figure 3. Sixteen further families: eight with a UniProt modification at a fixed unit position (phosphorylation, N-glycosylation, methylation, hydroxylation) and eight with no annotated modification anywhere in the array. Panels are drawn as in Figure 2. Click to enlarge.
Repeat families by recurrence across proteins and ELM curation status
Figure 4. How often the screen's repeats are already curated, from an analysis snapshot of 2,194 clustered families (numbers differ from the live counts above). a, Families recurring in two or more proteins (383) and families seen in one protein (1,811), split by whether ELM curates an instance inside the repeat (17 and 82; about 4.5% in both), curated the protein but not the repeat, or never curated the protein. b, The 383 recurring families grouped into 95 classes: 32 classes match a known motif, 60 have no annotation match and 3 are structured or low-complexity; colored segments are classes that cross gene families. c, The 9 unmatched classes that cross gene families, by the number of proteins carrying the repeat, colored by whether the carrier proteins share an interaction partner.

The screen at a glance

Structured contextDisordered contextMixed
Figure 1. Each row is one protein (gray), sorted by length; colored segments are located repeat copies, colored by the sequence context the model assigned. Proteins longer than 2,500 aa are truncated. Hover to identify a protein; click to open it in the explorer.
A. Sequence context (per family)
B. Repeat unit length (aa; ranges at midpoint)
C. Copies per family

Method

Scripts handle cohort selection, blinding and coordinate lookup and contain no motif-finding logic. The model sees only an anonymous sequence. Throughout, a repeat family is one repeat array inside one protein: counts of families are not counts of distinct motifs, and every count on this page says whether it counts proteins, families or copies.

1

Cohort

Human proteins with ≥100 aa of disorder, from UniProt / MobiDB-lite regions or a TOP-IDP propensity window.

2

Blinding

Names and annotations are removed and sequences get shuffled anonymous IDs. The key is kept separately.

3

Two-pass reading

One pass for long structural repeats, then an independent pass over disordered stretches for short anchored motifs (≥5 aa, usually ≥3 copies).

4

Locating copies

The model quotes each copy it reports. A script finds each quote in the sequence and records its coordinates; quotes that aren't in the sequence are dropped. Copies need not match each other.

5

Unblinding

Identities are attached only afterwards, then results are compared with UniProt features, ELM and AlphaFold models.

Consensus notation. One symbol per residue position; unit positions in the highlights count these symbols from 1.

W fixed residue[ST] one of thesex any residue(x){4,13} variable loop φ hydrophobic (AVILMFWY)ψ large aliphatic (VILM)Ω aromatic (FWY) + basic (KR)- acidic (DE)o polar (STNQ)

Catalog

Every protein with at least one repeat family. Search by gene, protein name or consensus; click a row to open it in the explorer.

GeneProteinLengthRepeat families (consensus, copies)
Loading catalog…

Citation

This is a preview of work in progress; the screen is still running and the numbers will change. Please get in touch before citing.

@misc{mcwhite_repeats_preview,
  title  = {A blinded language-model screen for repeats in human disordered proteins},
  author = {{McWhite Lab}},
  year   = {2026},
  note   = {Research preview},
  url    = {https://clairemcwhite.github.io/repeats/highlights.html}
}