Characterizing Human Protein Mass Density Distributions

Surface-enhanced laser desorption/ionization time- of-flight mass spectrometry (SELDI or SELDI-TOF MS) has yielded predictive protein profiles for a number of clinically relevant diseases. Yet, rather than identifying specific proteins, such studies have provided diagnostic information solely based on "black box" predictors that look at differential patterns of mass spectrometry peaks. This paper analyzes the number of proteins that could be represented by mass spectrometry peaks associated with corre- sponding masses. It proposes and compares three models to fit the probability density function (PDF) of such a distribution. These include the gamma, Poisson, and negative binomial models. The results yielded a nonuniform distribution of protein masses- particularly apparent near masses where proteins in- volved in somatic recombination are prevalent. This may be useful to consider when using protein databases for protein identification near such mass regions. In terms of PDF models, the distribution surprisingly does not follow a simple Poisson distribution of counts. Instead, it follows a negative binomial distribution.