Relative Quantification in Mass Spectrometry Based Proteomics Studies: Understanding Bias and Variability in an iTRAQ Spike-in Study

摘要

1. Peptide detection was found to be a function of abundance and protein mass. 2. Between-run and between-tag biases exist in this iTRAQ spike-in study as has been observed in other mass spectrometry platforms. Statistical normalization algorithms properly address linear biases. 3. Protein abundance is measured with unequal precision, therefore statistical analyses should utilize weighting. The most abundant proteins have the smallest abundance CV. 4. Observed fold changes are biased towards the null. 5. The variance associated with the fold change between two groups decreases as mass (and number of data points) increases. 6. Small, low abundance proteins have the largest variance, making their biological fold changes the most difficult to measure with accuracy and precision. Thus, marker discovery studies focused on this class of proteins should be powered accordingly.