ADTnorm: Robust Integration of Single-cell Protein Measurement across CITE-seq Datasets.

Area of Science:

• Single-cell multi-omics analysis
• Immunogenomics
• Computational biology

Background:

• CITE-seq (Cellular Indexing of Transcriptomes and Epitopes sequencing) measures both surface protein and mRNA expression in single cells.
• Antibody-derived tags (ADTs) from CITE-seq offer robust surface protein detection, enhancing cell-type identification.
• Variability in antibody staining causes batch effects in ADT expression, hindering data interpretation and cross-study comparisons.

Purpose of the Study:

• To introduce ADTnorm, a novel normalization and integration method specifically designed for Antibody-Derived Tag (ADT) abundance in CITE-seq data.
• To address and mitigate batch effects caused by antibody staining variability in CITE-seq experiments.
• To improve the accuracy of cell-type identification and enable integration of diverse CITE-seq datasets.

Main Methods:

• ADTnorm was developed as a normalization and integration method for ADT abundance.
• The method was benchmarked against 14 existing scaling and normalization techniques using 13 public CITE-seq datasets.
• Performance was evaluated based on the alignment of cell populations, removal of technical variation, and improvement in cell-type separation.

Main Results:

• ADTnorm demonstrated superior performance in aligning cell populations with varying surface protein expression levels across multiple datasets.
• The method effectively removed technical variation across different batches, improving the consistency of ADT expression data.
• ADTnorm enhanced cell-type separation and facilitated the integration of public CITE-seq datasets with diverse experimental designs.

Conclusions:

• ADTnorm provides an effective solution for normalizing and integrating CITE-seq ADT data, overcoming batch effects and improving biological signal.
• The method supports atlas-level analyses by enabling the integration of heterogeneous public datasets.
• ADTnorm includes utilities for automated threshold-gating and antibody quality assessment, aiding experimental optimization and discovery of novel biomarkers, as demonstrated in a COVID-19 dataset.