Campaign · NOBBLE

Pair ledger

Every (receptor, peptide, ligand) triple in the campaign. Sort by any column. Click a row to pin it into every other tab. Hover any number for its raw value, units and where it came from. The composite is re-weighted in the browser from components the pipeline already measured: moving a slider re-ranks this table without a server round trip.

Modality Tier bridging seeds only pinned only

Pair	E3	Target	Ligand	Mode	Composite	Bridge %ile	Bridge rate	Δ-ipTM	Converge	PLIP	Flags

Composite weights

Spec 4.8 fits these on the benchmark panel and the Sievers matched set, never by hand. Until that fit has the data, these sliders explore the weighting: they do not publish one. The pipeline's own nobble_score stays null for the same reason.

Bridge percentile 0.35

Buried surface of the glue against both chains, as a percentile of the experimentally observed distribution. Compares a prediction to real interfaces rather than to the model's own confidence.

Delta-ipTM 0.25

Ternary minus binary interface confidence on the receptor-to-peptide chain pair, minimum across seeds. A collapsed seed can dominate it: the median is shown alongside.

Pose convergence 0.25

Pairwise peptide backbone RMSD across seeds after superposing on the receptor. Tracks reality better than confidence for peptides.

PLIP fingerprint match 0.15

Fraction of the canonical degron contacts for that E3 class that the prediction reproduces.

Lysine geometry 0.0

BINMAN's reach window is deliberately unfitted: Cb-Cb AUC 0.42 to 0.48 on 369 assayed lysines, at or below chance and slightly inverted. Reported as a measurement only.

Throughput

s / token 0.4557 median, measured

Jobs timed 75

Affordable pairs 175 in 48.0 h

Jobs per pair 15 computed, not quoted