Clair

Clair calls small genetic variants from single-molecule sequencing pileup data using deep neural networks to enable accurate germline small-variant detection.

Key Features:

• High Accuracy and Efficiency: Achieves higher precision and recall than Clairvoyante, Longshot, and Medaka for small-variant calling from single-molecule sequencing data.
• Speed Optimization: Optimized for fast processing on conventional CPUs without requiring specialized hardware.
• Approaching Accuracy Limits: Benchmarking and analysis of missed variants indicate performance approaching theoretical accuracy limits of current deep neural network methods applied to pileup data.

Scientific Applications:

• Germline Variant Calling: Provides highly accurate small-variant detection for studies of inherited genetic variation.
• Structural Variant Analysis and Genome Assembly: Although focused on small variants, its precision can contribute to structural variant calling efforts and complex genome assembly projects.
• Epigenetic Research: Processing of single-molecule sequencing data supports detection of epigenetic marks relevant to gene regulation studies.

Methodology:

Employs a deep neural network architecture tailored for analyzing pileup data from single-molecule sequencing, trained on extensive datasets and optimized to distinguish true genetic variants from sequencing noise.