FEATS

FEATS performs downstream analysis of single-cell RNA sequencing (scRNA-seq) datasets to identify cellular heterogeneity through clustering, outlier detection, and data integration.

Key Features:

• Clustering Capabilities: Performs clustering on scRNA-seq data using a univariate feature selection approach that selects top informative features for clustering, inspired by manual cell-type determination with known marker genes.
• Cluster Estimation: Estimates the optimal number of clusters using algorithms for precise identification of distinct cellular populations and evaluates performance with metrics such as the adjusted rand index (ARI).
• Outlier Detection: Detects and manages anomalous cells or data points within scRNA-seq datasets to improve downstream analysis reliability.
• Data Integration: Integrates scRNA-seq data from multiple experiments to combine datasets across conditions or experimental setups.
• Computational Efficiency: Implements computational optimizations to handle large-scale single-cell RNA-seq datasets without compromising speed or accuracy.
• Implementation: Provided as a Python-based software package for scRNA-seq downstream analysis.

Scientific Applications:

• Cellular heterogeneity characterization: Enables identification and characterization of distinct cell populations within heterogeneous tissues using scRNA-seq clustering.
• Developmental biology: Facilitates analysis of cell-state transitions and lineage relationships in developmental systems.
• Cancer research: Supports profiling of tumor heterogeneity and identification of malignant and non-malignant cell populations.
• Immunology: Assists in profiling immune cell subsets and states from single-cell transcriptomic data.
• Comparative studies: Allows integration of multiple experiments to compare cellular compositions across conditions or treatments.

Methodology:

Uses a univariate feature selection approach to prioritize informative features for clustering, algorithms for estimating the optimal number of clusters, outlier detection mechanisms, data integration functions, and evaluation with the adjusted rand index (ARI).