henhaobio
Sign in

bioRxivBioinformaticsPreprint

A peptide-level multiple imputation strategy accounting for the different natures of missing values in proteomics data

Quentin Giai GianettoiD, Samuel WieczorekiD, Yohann CoutéiD, Thomas BürgeriD

PDFDOI
Structured summary available · Jump to summary ↓

Abstract

MotivationQuantitative mass spectrometry-based proteomics data are characterized by high rates of missing values, which may be of two kinds: missing completely-at-random (MCAR) and missing not-at-random (MNAR). Despite numerous imputation methods available in the literature, none account for this duality, for it would require to diagnose the missingness mechanism behind each missing value. ResultsA multiple imputation strategy is proposed by combining MCAR-devoted and MNAR-devoted imputation algorithms. First, we propose an estimator for the proportion of MCAR values and show it is asymptotically unbiased under assumptions adapted to label-free proteomics data. This allows us to estimate the number of MCAR values in each sample and to take into account the nature of missing values through an original multiple imputation method. We evaluate this approach on simulated data and shows it outperforms traditionally used imputation algorithms. AvailabilityThe proposed methods are implemented in the R package imp4p (available on the CRAN Giai Gianetto (2020)), which is itself accessible through Prostar software. Contactquentin.giaigianetto@pasteur.fr; thomas.burger@cea.fr

Abstract summary

Claude Opus 5.5 · xhigh effort · 18 Oct 2026 · from the abstract

Estimating what fraction of missing proteomics values are missing completely at random lets MCAR- and MNAR-specific imputers be combined; on simulations this beats standard imputation.

Quantitative mass spectrometry proteomics data carry many missing values of two different kinds, and available imputation methods treat them alike because deciding the mechanism behind each individual missing value is impractical. Estimating only the proportion of one kind avoids that per-value diagnosis and lets mechanism-specific imputers be mixed inside a single multiple imputation scheme.

Key results

  1. ClaimedAn estimator is proposed for the share of missing values that are missing completely at random, and is reported to be asymptotically unbiased under assumptions tailored to label-free proteomics.

    Stated as a theoretical property in the abstract; the assumptions are described as adapted to label-free data but are not listed, and no proof or simulation of the estimator's bias is given. · Abstract

  2. ClaimedThe estimated MCAR count per sample drives a multiple imputation procedure that applies MCAR-targeted and MNAR-targeted imputation algorithms to the appropriate share of missing values.

    Described as the design of the strategy; the abstract does not name the component algorithms or say how values are allocated between them. · Abstract

  3. SuggestedOn simulated data the combined strategy is reported to outperform imputation algorithms in common use.

    Simulation-only evaluation; no competing methods are named and no error metric, effect size or number of simulated datasets is reported. · Abstract

Methods, data and reproducibility

An estimator of the proportion of missing-completely-at-random values, argued to be asymptotically unbiased under label-free proteomics assumptions, is used to fix how many missing values per sample are treated as MCAR; a multiple imputation procedure then combines MCAR-targeted and MNAR-targeted imputation algorithms, and is evaluated on simulated data.

  • Simulated quantitative proteomics data used to evaluate the imputation strategy
Methods & tools
imp4p, Prostar, R, CRAN
Reproducibility gaps
The methods are said to be implemented in the R package imp4p on CRAN and reachable through the Prostar software, but no URL is printed in the text.
Limitations (2)
  • The evaluation uses simulated data only; no real mass spectrometry dataset, spike-in experiment or downstream differential analysis is reported in the abstract. · noted in this summary
  • The unbiasedness argument depends on assumptions chosen for label-free data, and the abstract neither states them nor reports how the estimator behaves when they fail or in labelled workflows. · noted in this summary

Read this if you impute missing values in label-free quantitative proteomics and want the two missingness mechanisms handled separately rather than with one imputer.

Citations
59

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISGardner '21Hediyeh-zad '20Harris '23Moon '25Dabke '21Egert '21
83% semantic matchBioinformatics
Miranda L. Gardner, Michael A. Freitas·International Journal of Molecular Sciences·Sep 6, 2021· 53 citations