DAPCy

DAPCy implements the Discriminant Analysis of Principal Components (DAPC) method to infer population structure from genetic markers with enhanced scalability and efficiency.

Key Features:

• Large Dataset Support: Utilizes compressed sparse matrices and truncated singular value decomposition for dimensionality reduction to handle large genomic datasets.
• Cross-Validation: Implements training-test cross-validation for robust model evaluation.
• De Novo Genetic Clustering: Includes modules for de novo genetic clustering.
• Visualization and Reporting: Provides extensive visualization and reporting capabilities.
• Performance Optimization: Processes genomic datasets with thousands of samples and features in less computational time and with reduced memory usage compared to the original R implementation.

Scientific Applications:

• Population Structure Inference: Assesses the genetic structure of populations using genetic markers, focusing on the description of variation between genetic clusters.
• Large-Scale Genomic Analysis: Enables population genetic analyses on large genomic datasets such as the Plasmodium falciparum dataset from MalariaGEN and the 1000 Genomes Project.

Methodology:

DAPCy leverages the scikit-learn library and implements principal component analysis combined with linear discriminant analysis. Dimensionality reduction is performed using truncated singular value decomposition on compressed sparse matrices, followed by training-test cross-validation for model evaluation.