Huge thanks to our collaborators —
Britta Velten (@brittavelten.bsky.social)
, the Klingmüller lab at DKFZ (@klingmuelab.bsky.social), and the Winter team at Thoraxklink Heidelberg
— for their contributions 🙏
We invite you to try out msBayesImpute and welcome any feedback or suggestions! 💬

msBayesImpute works for both small and large datasets, requires no parameter tuning, and is available in R and Python. A Shiny app will be available soon!
👉 Python version: github.com/Lu-Group-UKH...
👉 R version: github.com/Lu-Group-UKH...

Across synthetic, semi-synthetic, and experimental serial dilution datasets, msBayesImpute consistently:
✅ Achieved the lowest imputation error
✅ Improved sample-wise normalization
✅ Delivered the highest accuracy in DE analysis across 9 methods

Missing values in proteomics are often not missing at random (MNAR). Existing methods either assume MAR or oversimplify MNAR.
💡 msBayesImpute learns protein-specific dropout curves directly from the data using Bayesian matrix factorization + probabilistic dropout models.