Building computational benchmarks: an Omnibenchmark reimplementation of a single-cell preprocessing pipeline evaluation

Omnibenchmark standardizes comparisons of five single-cell RNA-seq preprocessing pipelines across three datasets, enabling reproducible, extensible benchmarking and revealing differences in clustering resolution parameter implementation.

• Why it matters: Single-cell benchmarking is fragmented by incompatible design choices, including differing datasets, metrics, and parameter sets, limiting consistent method comparisons.
• What they did: The framework evaluates five pipelines spanning filtering, normalization, dimensionality reduction, and clustering on three datasets, and supports standardized benchmark construction for extensible comparisons.
• The result: Harmonizing the pipelines showed that clustering resolution parameters are implemented differently by tools, and the standardized setup enables extensions like parameter sweeps and step-level isolation.