Methods
This page summarizes the algorithmic families included in spCellEval and the criteria used to assess method performance across datasets.
Supervised
Learns explicit labels and generally achieves strongest recovery when high-quality annotations are available.
Prior-Knowledge Based
Uses marker panels and biological priors for robust cell typing when labels are scarce.
Unsupervised
Identifies structure in large cohorts and supports exploratory phenotyping in novel tissues.
Pre-trained
Enables rapid cell typing in novel tissues using pre-existing models.
Method Papers
Open the primary paper or publication page for each method in the benchmark.
Need Full Quantitative Results?
Explore the full metric matrix and ranking tables on the homepage Results section.
Open Results