AdpFunChisq

AdpFunChisq applies an adapted functional chi-squared test to quantify model-free functional dependencies by evaluating the direction and strength of biological functions for causal inference in complex and single-cell multiomics datasets.

Key Features:

• Model-Free Analysis: Operates without predefined parametric models, enabling application across diverse datasets and experimental conditions.
• Functional Dependency Measure: Quantifies functional patterns by assessing both direction and strength to distinguish functional from non-functional or independent patterns.
• Bias Mitigation: Emphasizes functional evidence to reduce biases related to marginal distributions and sample size that can produce spurious causal relationships.
• Robust Performance: Demonstrated superior performance relative to ten alternative methods on synthetic and biological datasets, maintaining consistency despite data variability.
• Implementation: Provided as part of the R package FunChisq (version 2.5.2 or above).

Scientific Applications:

• Single-cell genomics and multiomics: Infers molecular interactions and prioritizes hypotheses for experimental validation from dynamically rich single-cell multiomics data.
• Acute leukemia analysis: Applied to single-cell multiomics data from acute leukemia phenotypes to identify a potential causal relationship between the T-cell surface glycoprotein CD3 delta chain and specific genes involved in viral carcinogenesis.

Methodology:

Implements an Adapted Functional Chi-squared test under the causality-by-functionality principle, performing model-free assessment by scoring direction and strength of functional patterns and rewarding patterns indicative of causality.