Digital Cell Sorter (DCS): a cell type identification, anomaly detection, and Hopfield landscapes toolkit for...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) generates vast amounts of data requiring sophisticated analytical tools.
• Existing scRNA-seq analysis pipelines often involve multiple discrete steps, necessitating integrated solutions.
• Advancements in algorithmic approaches are crucial for maximizing biological insights from scRNA-seq data.

Purpose of the Study:

• To introduce a comprehensive software platform, Digital Cell Sorter (DCS), for streamlined scRNA-seq data analysis.
• To present novel algorithms for automated cell type identification, cell anomaly quantification, and phenotypic landscape visualization.
• To provide a self-contained toolkit integrating state-of-the-art methodologies for scRNA-seq analysis.

Main Methods:

• Development of two automatic cell type identification methods: a voting algorithm and a Hopfield classifier.
• Implementation of an isolation forest-based method for quantifying cell anomalies.
• Creation of a visualization tool for cell phenotypic landscapes using Hopfield energy-like functions.
• Integration of these novel methods into an open-source Python software package (DCS).

Main Results:

• The DCS toolkit provides a suite of methods for scRNA-seq data analysis, including novel algorithms.
• Demonstrated capability using large datasets of peripheral blood mononuclear cells (PBMC) and bone marrow plasma cells.
• Successfully validated algorithms for deconvolving cell mixtures and detecting anomalous cells in PBMC data.

Conclusions:

• The DCS software offers a robust and integrated toolkit for scRNA-seq analysis.
• The novel algorithms contribute to improved cell type identification, anomaly detection, and data visualization.
• DCS is available as an open-source Python package, facilitating its adoption in the research community.