BEELINE

This collection consists of over 400 single-cell gene expression datasets across four curated and six synthetic gene regulatory networks. It was created to benchmarking algorithms for gene regulatory network inference in Pratapa et al. (2020).

Task: The collection can be used to study causal inference algorithms.

Summary: 

• Size of collection: >400 datasets on 6 - 19 features of different size
• Task: Causal Inference Problem
• Data Type: Mixed Data
• Dataset Scope: Collection of Datasets
• Ground Truth: Known Graph
• Temporal Structure: Static Data
• License: CC BY-NC 4.0 (see 10.5281/zenodo.3701939)
• Missing Values: No Missing Values

Missingness Statement: There are no missing values.

Collection: (for a detailed description see Peng et al. (2024), for simulation details see Pratapa et al. (2020))

• Curated: There are experiments on four curated gene regulatory networks: mCAD (Mammalian Cortical Area Development, 14 edges and 5 nodes), VSC (Ventral Spinal Cord Development, 15 edges and 8 nodes), HSC (Hematopoietic Stem Cell Differentiation, 30 edges and 11 nodes), and GSD (Gonadal Sex Determination, 79 edges and 18 nodes).
  
• Synthetic: There are experiments six synthetic gene regulatory networks: dyn-BF (Bifurcating, 12 edges and 5 nodes), dyn-BFC (Bifurcating Converging, 18 edges and 9 nodes), dyn-CY (Cycle, 6 edges and 5 nodes), dyn-LI (Linear, 8 edges and 7 nodes), dyn-LL (Linear Long, 19 edges and 18 nodes), and dyn-TF (Trifurcating, 20 edges and 7 nodes).
  

Files per Experiment:

• GroundTruth.csv: This file represents the actual biological regulatory interactions between genes, typically derived from known databases, literature, or synthetic models. An edge weight of +1 represents activation, -1 represents inhibition.
• refNetwork.csv: This file is a processed version of the ground truth network, keeping only the sign (+ or -) of interactions.
• ExpressionData.csv: This file contains the RNAseq data, with genes as rows and cell IDs as columns.
• PseudoTime.csv: This file contains the Pseudotime. It is a computationally inferred measure that orders single cells along a trajectory to represent their progression through a biological process, such as differentiation or development.