DreamAI: algorithm for the imputation of proteomics data
Weiping MaiD, Sunkyu KimiD, Shrabanti ChowdhuryiD, Zhi LiiD, Mi YangiD, Seungyeul YooiD, Francesca PetraliaiD, Jeremy R. JacobseniD, Jingyi Jessica LiiD, Xinzhou GeiD, Kexin LiiD, Thomas Von YuiD, Anna Pamela CalinawaniD, Nathan J EdwardsiD, Samuel PayneiD, Paul Christopher BoutrosiD, Henry RodrigueziD, Gustavo A. StolovitzkyiD, Jun ZhuiD, Jaewoo KangiD, David FenyöiD, Julio Sáez-RodrígueziD, Pei WangiD
Abstract
Deep proteomics profiling using labeled LC-MS/MS experiments has been proven to be powerful to study complex diseases. However, due to the dynamic nature of the discovery mass spectrometry, the generated data contain a substantial fraction of missing values. This poses great challenges for data analyses, as many tools, especially those for high dimensional data, cannot deal with missing values directly. To address this problem, the NCI-CPTAC Proteogenomics DREAM Challenge was carried out to develop effective imputation algorithms for labeled LC-MS/MS proteomics data through crowd learning. The final resulting algorithm, DreamAI, is based on an ensemble of six different imputation methods. The imputation accuracy of DreamAI, as measured by Pearson correlation, is about 15%-50% greater than existing tools among less abundant proteins, which are more vulnerable to be missed in proteomics data sets. This new tool notably enhances data analysis capabilities in proteomics research.
02 /Specification & Classification
Click any tag to browse matching papers- Citations
- 51
- Assay
- Research Focus
- Approach
Abstract summary
Claude Opus 5.5 · xhigh effort · 18 Oct 2026 · from the abstract · CC BY-NC-ND 4.0
An ensemble of six imputation methods, built through the NCI-CPTAC Proteogenomics DREAM Challenge, fills missing values in labeled LC-MS/MS proteomics data, helping most for low-abundance proteins.
Discovery mass spectrometry leaves a substantial fraction of protein measurements missing, and many high-dimensional analysis tools cannot run on incomplete matrices. Framing the problem as a crowd-learning competition yielded an ensemble imputer that the authors report beats existing tools where missingness bites hardest, among less abundant proteins.
Key results
SuggestedImputation accuracy for the less abundant proteins, which are the ones most often missing, is reported as higher than that of existing tools.
Pearson correlation is given as roughly 15%-50% higher for DreamAI than for prior imputation tools in the low-abundance range; neither the competing tools nor the evaluation datasets are named. · Abstract
ClaimedThe delivered algorithm is a combination of six separate imputation methods rather than a single model, assembled from entries to a community challenge.
DreamAI is described as an ensemble of six different imputation methods, produced by crowd learning in the NCI-CPTAC Proteogenomics DREAM Challenge. · Abstract
Methods, data and reproducibility
A crowd-learning competition, the NCI-CPTAC Proteogenomics DREAM Challenge, was run to develop imputation algorithms for labeled LC-MS/MS proteomics data; the resulting algorithm, DreamAI, combines six imputation methods, and its accuracy was assessed by Pearson correlation against existing tools, stratified by protein abundance.
- Methods & tools
- DreamAI, LC-MS/MS
Limitations (2)
- The accuracy gain is given only as a range for less abundant proteins; the competing tools, the benchmark datasets and the number of samples or proteins are not identified in the abstract. · noted in this summary
- The statement that the tool enhances proteomics data analysis is asserted without any downstream analysis result in the abstract. · noted in this summary
Who should read:Read this if you analyse labeled LC-MS/MS proteomics matrices and need an imputation method aimed at missing values among low-abundance proteins.
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
- Imputation of label-free quantitative mass spectrometry-based proteomics data using self-supervised deep learning
Henry Emanuel Webel, Lili Niu, … Simon Rasmussen·Nature CommunicationsPeer-reviewed·Jun 26, 2024
- DIMA: Data-driven selection of a suitable imputation algorithm
Janine Egert, Eva Brombacher, … Clemens Kreutz·Journal of Proteome ResearchPeer-reviewed·Jun 1, 2021
- Imputation of cancer proteomics data with a deep model that learns from many datasets
Lincoln Harris, William Stafford Noble·Molecular & Cellular ProteomicsPeer-reviewed·Nov 20, 2025
- A Simple Optimization Workflow to Enable Precise and Accurate Imputation of Missing Values in Proteomic Datasets
Kruttika Dabke, Simion D. Kreimer, … Sarah J. Parker·Journal of Proteome ResearchPeer-reviewed·May 3, 2021
- MSImpute: Imputation of label-free mass spectrometry peptides by low-rank approximation
Soroor Hediyeh-zadeh, Andrew Ian Webb, Melissa J. Davis·bioRxiv·Aug 12, 2020
- msBayesImpute: A Versatile Framework for Addressing Missing Values in Biomedical Mass Spectrometry Proteomics Data
Jiaojiao He, Barbara Helm, … Junyan Lu·Communications ChemistryPeer-reviewed·Jul 7, 2026