Near-Lossless Compression of Mass Spectra for Proteomics

Recent improvements in mass spectrometry (MS) technology led to an explosive amount of MS data collected and shared. A typical liquid chromatography/mass spectrometry (LC/MS) "image" from the instrument used in this study consists of 4 GB of data. To reduce the bit rate required to code the MS data below that of the authors' previous (lossless) algorithm, we introduce a technique for near-lossless compression. It guarantees that each decompressed sample differs from its original value by no more than a user-specified quantity defined as the target maximum absolute distortion (MAD). We evaluate the proposed method by introducing feature-based metrics applied to the decompressed MS data and show that the MAD-based compression outperforms a traditional coding algorithm aimed at minimizing the mean squared error.