Bioinformatics Seminar
Time: 11AM
Venue: Zoom Webinar
19 October 2021
This is a PhD Confirmation SeminarEstimating Detection Probability in Mass Spectrometry Proteomics Data
Mengbo LiWEHI Bioinformatics
Mass spectrometry (MS) is an essential technology for studying the proteome at system-wide scale. Recent improvements in MS technology promise dramatic improvements in throughput speed and accuracy. However, missing values widely exist in MS proteomics data due to peptides that are not detected. Such missing values are neither censored nor missing at random, which leads to great complication hindering the use of simple and powerful limma-style analyses. We start with examining the probability of detecting a peptide in a model-based fashion. We assume the detection probability in peptides to be a logit linear function of their expression values. We investigate two typical distributional assumptions in peptide expression levels and their implications in the estimation of the detection probability function. The proposed framework, which is also applicable on protein-level data, is tested on both simulated and real data sets. We also discuss a couple of possible paths toward formulating statistical tests for limma-style differential expression analyses following the estimation of the detection probability function.