PhenoComb

PhenoComb enables agnostic exploration of complex phenotypes in high-dimensional cytometry datasets by enumerating all possible combinations of marker expression states and quantifying cells per phenotype.

Key Features:

• Agnostic exploration: Evaluates every potential combination of markers in a dataset to discover novel phenotypic patterns.
• Signal intensity thresholding: Assigns discrete states (e.g., negative, low, high) to markers based on signal intensity thresholds and counts cells per sample across all marker combinations in a memory-efficient manner.
• Statistical comparisons and relevance evaluation: Provides multiple methods for statistical comparison of phenotypes, evaluation of biological relevance, and assessment of independence among identified phenotypes.
• Customizable analysis parameters: Supports specification of parent populations, filtering of low-frequency populations, and definition of a maximum phenotype complexity for analysis.
• Scalability and efficiency: Demonstrated processing of 16 markers in minutes and up to 26 markers within hours on synthetic data, enabling analysis of extensive marker combinations without large disk footprint.

Scientific Applications:

• HIV flow cytometry dataset: Analysis of a 12-marker, 421-sample dataset identified immune phenotypes associated with HIV seroconversion that corroborated the original publication.
• COVIDome CyTOF dataset: Analysis of a 40-marker, 99-sample CyTOF dataset revealed immune phenotypes with altered frequencies in infected individuals versus healthy controls.

Methodology:

Discretizes marker signal intensities into states (e.g., negative, low, high); enumerates all marker-state combinations and counts cells per sample in a memory-efficient manner; performs statistical comparisons between phenotypes and evaluates phenotype independence; supports parent-population specification, low-frequency filtering, and maximum-complexity constraints.