Antibody Abnormal Site Detection (Ab-normal)

This tool evaluates abnormal sites in antibody sequences. We use an antibody MLM model to compute NLL (Negative Log-Likelihood) for each position, then apply z-score normalization to identify amino acid positions that are relatively "less natural / more abnormal."

Two input modes are supported: ① single-chain sequence (VH or VL, no |); ② paired sequence (VH|VL, exactly one |, with heavy and light chains on either side), e.g.: QVQL...|DIQM....

This tool supports only a single sequence input. You may select the species (Human / Camelid) to match the corresponding reference statistics and scoring framework.

1. Antibody Amino Acid Sequence (supports only 1 FASTA entry):

Parsed sequences: 0, Total residues: 0

Species:



Model Performance (Current Version)

Antibody MLM Model Performance:
  PPL = 1.39
  Acc = 0.9162

Antibody Autoregressive (GPT) Model Performance:
  PPL = 1.47
  Acc = 0.8974

Metrics Description

NLL (Negative Log-Likelihood) reflects how "surprising" a given amino acid is to the model at a particular position. The z-score is a standardized NLL score relative to a reference distribution, enabling more intuitive cross-position comparison of abnormality. Higher z-scores generally indicate more "abnormal" positions.

Last updated: 2026-05-26