scValue

scValue ranks individual cells by data value using out-of-bag estimates from a random forest model to prioritize high-value cells and preserve key biological signals within subsamples of large scRNA-seq datasets for machine learning and deep learning workflows.

Key Features:

• Value-based cell ranking: Ranks cells using out-of-bag estimates from a random forest model to quantify data value.
• Variable cell-type representation: Allocates greater representation to cell types with higher variability in data value.
• Balanced subsampling: Achieves well-balanced cell-type representation with distributional properties akin to uniform sampling.
• Scalable execution: Demonstrates fast execution on datasets ranging from tens of thousands to millions of cells.
• Subsampling quality metrics: Evaluates performance using computational time, Gini coefficient, and Hausdorff distance.

Scientific Applications:

• Automatic cell-type annotation: Benchmarked on annotation tasks across four large scRNA-seq datasets paired with distinct ML/DL models, closely matching full-data performance.
• Label transfer learning: More effectively preserved T-cell annotations across human gut-colon datasets.
• Cross-study label harmonization: More accurately reproduced T-cell subtype relationships in a human spleen dataset.
• Bulk RNA-seq deconvolution: Constructed a more reliable single-cell immune reference for cell-type deconvolution in simulated bulk tissue samples.

Methodology:

scValue uses out-of-bag estimates from a random forest model to rank individual cells by data value, then prioritizes high-value cells for subsampling.