Missing value imputation
Helping researchers work with incomplete omics data through reproducible methods, realistic benchmarks and accessible software.
What does a missing measurement mean?
An empty value in an omics dataset can have several explanations. A signal may fall below a measurement threshold, an instrument may not detect a feature consistently, or a value may be absent for another reason. Those differences matter when deciding how to analyse the data.
Imputation replaces missing entries with estimates. Different methods make different assumptions about the data and its missingness, so replacing every gap with the same simple value can change relationships among measurements. A useful evaluation considers both reconstruction and the consequences for subsequent analysis.
Questions we explore
- How do different patterns of missingness influence imputation performance?
- How can researchers compare algorithms within a reproducible workflow?
- What happens to downstream analysis after missing values are replaced?
Our approach
We develop software that lets researchers compare imputation methods within a common workflow. The imputomics web application and R package bring multiple algorithms together and support benchmarking under different missingness scenarios.
Our focus is on making the comparison reproducible and accessible. Instead of treating one method as universally best, we examine performance in the context of a dataset, the simulated missingness and the question being asked. The linked publication story introduces the software and its evaluation framework.
Tools, projects & people
Read about the imputomics2 project and find the application in our software catalogue.
Publications & research stories
Introducing imputomics
A web server and R package for comparing and applying missing value imputation methods in metabolomics.