ProteoAnalyst treats strict zero values as missing data (NA). This is a standard practice in proteomics because zeros typically represent signals that were present but fell below the mass spectrometer’s limit of detection (LoD), rather than a total absence of the protein.
Missing values (NA or zeros) will be handled by missing value imputation step in data processing. ProteoAnalyst offers methods supporting both missing not at random (MNAR) behavior, where low‑abundance proteins are undetected rather than absent, the default option replaces these entries with 1/5 of the minimum observed intensity. For more advanced handling of such left-censored data, users can choose methods like MinDet, MinProb, and quantile regression imputation of left-censored data (QRILC). To support broader data types where values might be missing at random (MAR), ProteoAnalyst also offers established methods including replace by mean, median, K-Nearest Neighbors (KNN) or Bayesian PCA