Gdaphen

Gdaphen identifies key qualitative and quantitative predictor variables in phenotypic datasets to support analysis of genetic and acquired disease effects while addressing low sample sizes, high-dimensional data, and unbalanced experimental designs.

Key Features:

• Integrated Statistical Framework: Incorporates Multiple Factor Analysis (MFA) to manage groups of variables recorded from the same individuals or to anonymize genotype-based recordings and selects optimized inputs based on a 30% correlation threshold.
• Advanced Analytical Techniques: Employs General Linear Model (GLM)-based classifiers to identify variables predicting gene dosage effects and Random Forest (RF) implementations to detect the most discriminative variables in non-linear distributions.
• Comprehensive Pre-processing Capabilities: Includes data imputation and genotype anonymization for handling missing values and sensitive genotype information.
• Visualization Tools: Generates plots to depict classifier performance and the significance or discriminative power of identified predictor variables.

Scientific Applications:

• Disease versus non-disease classification: Identifying phenotypic predictors that distinguish disease states from controls in clinical and preclinical datasets.
• Treatment response assessment: Detecting variables associated with treatment effects or gene dosage responses.
• Sexual dimorphism studies: Exploring sex-related differences in phenotypic measures.
• Phenotypic predictor discovery: Extracting significant qualitative and quantitative predictors from complex, high-dimensional phenotypic datasets.

Methodology:

Data pre-processing with missing-data imputation and genotype anonymization; dimensionality reduction and correlation management using Multiple Factor Analysis (MFA) and Principal Component Analysis (PCA); variable selection and classification via General Linear Model (GLM)-based classifiers and Random Forest (RF); and generation of result visualizations.