Statistics or biology: the zero-inflation controversy about scRNA-seq data

Area of Science:

• Bioinformatics
• Genomics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) data analysis is challenged by the interpretation of zero counts.
• Zeros in scRNA-seq data are controversially viewed as either true biological signals or technical artifacts (missing data).

Purpose of the Study:

• To address the controversy surrounding zeros in scRNA-seq data.
• To differentiate between biological and non-biological zeros and their computational origins.
• To evaluate the impact of non-biological zeros on downstream data analysis.

Main Methods:

• Discussion of biological and non-biological zero sources.
• Introduction of five mechanisms for introducing non-biological zeros in computational benchmarking.
• Benchmarking analysis using three input data types: observed counts, imputed counts, and binarized counts.

Main Results:

• Non-biological zeros can significantly impact scRNA-seq data analysis outcomes.
• Benchmarking reveals differential effects of observed, imputed, and binarized count data on analysis.
• Identification of key areas requiring further investigation regarding non-biological zeros.

Conclusions:

• Understanding the origin and impact of non-biological zeros is crucial for accurate scRNA-seq data interpretation.
• Transparent analysis practices are essential to mitigate potential biases introduced by non-biological zeros.
• Further research is needed to resolve open questions surrounding non-biological zeros in scRNA-seq.