SIMBSIG

SIMBSIG performs out-of-core, GPU-accelerated similarity searches, principal component analysis (PCA), and clustering to analyze large-scale bioinformatics datasets such as biobank, statistical genetics, and single-cell data.

Key Features:

• Scalability: Handles biobank-scale datasets that exceed main memory capacity.
• GPU Acceleration: Uses GPU-enabled computations to accelerate similarity search, PCA, and clustering.
• Batched Similarity Search: Implements batched similarity searches as indicated by the SIMmilarity Batched Search Integrated GPU name.
• Out-of-Core Processing: Supports processing of data that does not fit into RAM via out-of-core computation.
• Principal Component Analysis (PCA): Provides PCA for dimensionality reduction of high-dimensional genetic data.
• Clustering: Provides clustering methods for identifying patterns and groupings in large datasets.
• PyTorch Backend: Built on a PyTorch backend to enable modularity and extensibility.

Scientific Applications:

• Biobank-scale data analysis: Enables similarity searches, PCA, and clustering on large biobank datasets to identify population structure and relationships.
• Statistical genetics: Facilitates analysis of large-cohort genetic data where memory-efficient similarity search and dimensionality reduction are required.
• Single-cell analysis: Supports processing and dimensionality reduction of single-cell datasets that exceed in-memory limits.

Methodology:

GPU-enabled computations, out-of-core processing techniques, batched similarity searches, and a PyTorch backend are used to perform similarity search, PCA, and clustering.