Normalization and Missing Value Imputation for Proteomics Analysis

• Activity Number: 285
• Type: Topic Contributed
• Date/Time: Tuesday, August 3, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Yuliya V. Karpievitch*+
• Companies: Pacific Northwest National Laboratory
• Keywords: normalization ; bias removal ; mass spectrometry ; missing value ; significance analysis

Abstract

In mass "shotgun" proteomics, translation of peptide to protein abundances is complicated by many factors. Intensities can vary greatly across peptides from the same protein, due to, for example, differing ionization. Many peptides that are observed in some samples are not observed in others, resulting in widespread missing values. Furthermore, the fact that a peak was not observed for a peptide is often due to that peptide's presence at a lower abundance than the instrument can detect. Further analysis is complicated by the presence of systematic biases. Normalization models need to be flexible enough to capture biases of arbitrary complexity, while avoiding overfitting that would invalidate downstream statistical inference. I will discuss these issues in view of analysis pipeline utilized at the Pacific Northwest National Lab (PNNL).