文摘
A profile likelihood algorithm is proposed for quantitativeshotgun proteomics to infer the abundance ratios ofproteins from the abundance ratios of isotopically labeledpeptides derived from proteolysis. Previously, we haveshown that the estimation variability and bias of peptideabundance ratios can be predicted from their profilesignal-to-noise ratios. Given multiple quantified peptidesfor a protein, the profile likelihood algorithm probabilistically weighs the peptide abundance ratios by theirinferred estimation variability, accounts for their expectedestimation bias, and suppresses contribution from outliers. This algorithm yields maximum likelihood pointestimation and profile likelihood confidence intervalestimation of protein abundance ratios. This point estimator is more accurate than an estimator based on theaverage of peptide abundance ratios. The confidenceinterval estimation provides an "error bar" for eachprotein abundance ratio that reflects its estimation precision and statistical uncertainty. The accuracy of the pointestimation and the precision and confidence level of theinterval estimation were benchmarked with standardmixtures of isotopically labeled proteomes. The profilelikelihood algorithm was integrated into a quantitativeproteomics program, called ProRata, freely available atwww.MSProRata.org.