spillR: Causal Modelling of Spillover in Mass Cytometry

Marco Guazzini, Alexander Reisach, Sebastian Weichwald, Christof Seiler*

*Corresponding author for this work

Research output: Contribution to conferenceAbstractAcademic


In mass cytometry marker interference called spillover, can cause markers to have higher abundances than their true abundances. Chevrier and Crowell et al. 2018 introduced an experimental and computational procedure to estimate spillover and compensate for it in downstream analyses. Their R package CATALYST implements this in two steps: estimate spillover and remove spillover from data. We propose a method that combines estimation and correction in one step, in order to quantify uncertainty and improve data efficiency. Building on a hierarchical causal model that represents the spillover from one marker to another, we extend the usual negative binomial model by modifying the link function using information from the spillover matrix. Our simulations show that an explicit causal model has two advantages: We control false positive errors even in noisy settings and improve statistical power when the causal model is correctly specified. We will present our new R package spillR and vignettes for reproducibility.
Original languageEnglish
Publication statusPublished - 15 Sept 2022
EventEuroBioC 2022 - BioQuant, Im Neuenheimer Feld 267, Heidelberg, Germany
Duration: 12 Sept 202213 Sept 2022


ConferenceEuroBioC 2022
Internet address


Dive into the research topics of 'spillR: Causal Modelling of Spillover in Mass Cytometry'. Together they form a unique fingerprint.

Cite this