SPECK

SPECK estimates cell surface receptor abundance from single-cell RNA-sequencing (scRNA-seq) data using thresholded reduced rank reconstruction and CKmeans-based clustered thresholding to provide unsupervised receptor abundance estimates for studies of cellular communication and tissue composition.

Key Features:

• Unsupervised estimation: Provides receptor abundance estimation without requiring labeled training datasets.
• Thresholded reduced rank reconstruction: Applies thresholded reduced rank reconstruction to denoise and reconstruct scRNA-seq expression signals for receptors.
• CKmeans-based clustered thresholding: Uses CKmeans-based clustered thresholding to determine thresholds for quantifying reconstructed receptor expression distributions.
• scRNA-seq input: Operates directly on single-cell RNA-sequencing data to infer surface protein abundance.
• Validated performance: Demonstrated superior performance versus other unsupervised approaches for estimating the abundance of at least 25 human receptors across multiple tissue types.
• Indirect protein estimation: Enables receptor abundance inference when direct surface protein measurements (e.g., antibody-based assays) are unavailable.

Scientific Applications:

• Surface protein profiling from scRNA-seq: Estimating cell-surface receptor abundance to profile surface protein distribution across cell types and tissues.
• Cellular communication and interaction studies: Inferring receptor expression patterns relevant to cell–cell signaling and interactions.
• Comparative tissue receptor analysis: Comparing receptor abundance across multiple tissue types and benchmarking unsupervised receptor estimation methods.

Methodology:

Performs thresholded reduced rank reconstruction of scRNA-seq data followed by CKmeans-based clustered thresholding to generate unsupervised receptor abundance estimates.