MetaNeighbor

MetaNeighbor quantifies the replicability of cell types across single-cell RNA-sequencing (scRNA-seq) datasets using a neighbor voting mechanism to assess cluster similarity and identify robust marker genes.

Key Features:

• Replicability Quantification: Evaluates how consistently specific cell types are identified across multiple scRNA-seq datasets and addresses technical biases and analytic variability.
• Neighbor Voting Mechanism: Utilizes a neighbor voting approach to assess similarity between clusters from different datasets.
• Rapid Identification of Similar Clusters: Compares cluster results across datasets to identify clusters with high similarity.
• Application to Diverse Datasets: Has been tested on eight technically and biologically diverse scRNA-seq datasets.
• Identification of Robust Marker Genes: Aids identification of candidate marker genes for novel cell types by evaluating marker replicability across datasets, including application to 45 interneuron subtypes where 24 showed evidence of replication.
• Use of Variably Expressed Genes: Demonstrates that sets of variably expressed genes can accurately identify replicable cell types.

Scientific Applications:

• Defining Best Practices: Contributes to establishing best practices for complex assessments of scRNA-seq data through cross-dataset evaluation.
• Large-Scale Evaluation: Supports large-scale meta-analyses by enabling assessment of cell-type replicability using variably expressed gene sets.
• Marker Gene Discovery: Facilitates discovery and cross-dataset validation of robust candidate marker genes for novel cell types and interneuron subtypes.

Methodology:

Measures cell-type replicability by applying a neighbor voting mechanism to compute cluster similarity across datasets, analyzes sets of variably expressed genes, and compares results across datasets to evaluate neuronal identity and replication of interneuron subtypes.