The next Workshop on Quantitative Methods in Education, Health and the Social Sciences (QMeHSS) will be on Friday January 23rd from 11:00-12:30pm and will be led by Dr. Lin Chen from the University of Chicago. As with previous workshops, the seminar will be held in the NORC conference room 344. NORC is located at 1155 E. 60th Street. We look forward to seeing you there.
Mixed effects models for proteomics data with cluster-level non-ignorable missingness
Lin Chen, University of Chicago
In recent years, iTRAQ (Isobaric tag for relative and absolute quantitation) technique has become the workhorse in mass spectrometry based quantitative proteomics research. The iTRAQ method enables one to process multiple samples in a single experiment, and thus greatly enhances the throughput of protein profiling while reducing the cost. Due to the data-generating process of iTRAQ experiments, measurements of a peptide/protein in all the samples of one iTRAQ multiplex tend to be observed or missing altogether with the missing data probability depending on the combined intensity of the peptide/protein from all the samples in the iTRAQ multiplex experiment. We term this cluster-level non-ignorable missing-data mechanism as the Experiment-level Abundance-Dependent Missing-data mechanism (EADMM). Motivated by the EADMM observed in iTRAQ proteomics data, we introduce a new method —mixEMM—for analyzing cluster-level non-ignorable incomplete data. The proposed mixEMM method employs a linear mixed effects model and explicitly models the EADMM. Compared with mixed-effects models under the missing-at-random assumption, the proposed mixEMM method achieves more accurate parameter estimation and inference.
In the second half of this talk, we further extend the mixEMM model to detect protein-quantitative-trait-loci (pQTL). I will introduce a penalized mixed effects model with a novel penalty to encourage the detection of trans-hubs (pQTLs with many trans-regulations) accounting for natural grouping structures among single nucleotide variants.