Batch-effect correction improves downstream imaging mass cytometry data quality and facilitates robust cell type identification in the liver and hepatocellular carcinoma microenvironment
Zeitschrift für Gastroenterologie, vol. 63
Abstract
Introduction Imaging mass cytometry (IMC) enables in-depth analyses of single cells in complex tissue architectures such as the liver and the hepatocellular carcinoma (HCC) microenvironment. However, batch effects can be a significant hurdle for data analysis due to difficulty to discriminate procedural and experimental variability from biological differences. Batch-effect correction strategies aim to reduce non-biological sample-specific signals and improve overall data quality. No formal comparison on IMC data between published methods exists. Methods We performed IMC and cell segmentation on 12 hepatocellular HCC patients. To reduce batch-effects between single patients, semi-automated background removal (SABR), percentile normalization GUI image deNoising (PENGUIN) and single-cell based correction tools (fastMNN, harmony and Seurat) were performed separately. After phonograph clustering, we compared the performance of these approaches based on the ability to identify expected cell types in the HCC and liver microenvironment, their numeric distribution and patient specificity of clusters. Results Application of batch-effect correction tools led to a reduction of sample-specific clusters. Most expected cell types were identified after fastMNN, harmony, PENGUIN and SABR. We observed varying numbers of CD8 T cells in some patients with dense immune infiltrates. Inferring test quality criteria from ground truth comparison showed a favorable balance between sensitivity and specificity after PENGUIN and harmony. Conclusion Batch-effect correction may enhance IMC data performance by limiting non-biological patient-specific variability and ensuring robust cell type detection using clustering algorithms in the liver and HCC microenvironment. Both, batch-effect correction on a primary data level and on a post-segmentation level can be successfully applied. Publication History Article published online: 20 January 2025 © 2025. Thieme. All rights reserved. Georg Thieme Verlag KG Oswald-Hesse-Straße 50, 70469 Stuttgart, Germany
Authors 10
-
University Medical Center Freiburg
Affiliation as printed
University Medical Center Freiburg
-
Ana Marta Sequeira Aachen
Leiden University Medical Center
Affiliation as printed
Leiden University Medical Center
-
Marieke E. Ijsselsteijn Aachen
Leiden University Medical Center
Affiliation as printed
Leiden University Medical Center
-
University Medical Center Freiburg
Affiliation as printed
University Medical Center Freiburg
-
University Medical Center Freiburg
Affiliation as printed
University Medical Center Freiburg
-
University Medical Center Freiburg
Affiliation as printed
University Medical Center Freiburg
-
University Medical Center Freiburg
Affiliation as printed
University Medical Center Freiburg
-
University Hospital Heidelberg
Affiliation as printed
University Hospital Heidelberg
-
Noel F.C.C. de Miranda Aachen
Leiden University Medical Center
Affiliation as printed
Leiden University Medical Center
-
University Medical Center Freiburg
Affiliation as printed
University Medical Center Freiburg
Cited by 0 stored of 0
No patents citing this paper on Lens.org (checked 2026-10-11).