A new take on missing value imputation for bottom-up label-free LC-MS/MS proteomics

Abstract

AbstractLabel-free bottom-up proteomics using mass spectrometry and liquid chromatography has long established as one of the most popular high-throughput analysis workflow for proteome characterization. However, it produces data hindered by complex and heterogeneous missing values, which imputation has long remained problematic. To cope with this, we introduce Pirat, an algorithm that harnesses this challenge following an unprecedented approach. Notably, it models the instrument limit by estimating a global censoring mechanism from the data available. Moreover, it leverages the correlations between enzymatic cleavage products (i.e., peptides or precursor ions), while offering a natural way to integrate complementary transcriptomic information, when available. Our benchmarking on several datasets covering a variety of experimental designs (number of samples, acquisition mode, missingness patterns, etc.) and using a variety of metrics (differential analysis ground truth or imputation errors) shows that Pirat outperforms all pre-existing imputation methods. These results pinpoint the potential of Pirat as an advanced tool for imputation in proteomic data analysis, and more generally underscore the worthiness of improving imputation by explicitly modeling the correlation structures either grounded to the analytical pipeline or to the molecular biology central dogma governing multiple omic approaches.