BayesCall

BayesCall performs Bayesian base-calling from raw fluorescence images to extract sequence information and assign base-specific quality scores for Illumina high-throughput sequencing.

Key Features:

• Bayesian framework: Employs a Bayesian framework to incorporate features intrinsic to the sequencing process.
• Model-based algorithm: Uses a model-based algorithm leveraging statistical learning techniques for base inference.
• Time-dependent parameter integration: Integrates time-dependent parameters and models residual effects to enhance adaptability and accuracy.
• Raw fluorescence image processing: Extracts sequence information directly from raw fluorescence images generated during Illumina sequencing.
• Per-base probability and quality scoring: Computes the probability of observing each base and converts these probabilities into high-discrimination base-specific quality scores.
• Improved accuracy versus Bustard: Reduces average per-base error rate by approximately 51% on 76-cycle phiX174 data compared to Illumina's Bustard, with notable improvement in later cycles.
• Throughput and cost impact: Increases throughput per run, with implications for reducing overall sequencing costs.

Scientific Applications:

• Illumina base-calling: Improves base-calling accuracy for Illumina high-throughput sequencing platforms.
• Quality-aware downstream analysis: Provides base-specific quality scores to inform downstream genomic analyses and confidence assessment.
• Sequencing benchmarking: Enables benchmarking of sequencing performance using standard samples such as 76-cycle phiX174 data.
• Throughput optimization: Supports strategies aimed at increasing run throughput to reduce per-run sequencing costs.

Methodology:

Applies a Bayesian model-based algorithm and statistical learning to raw fluorescence images, integrating time-dependent parameters and modeling residual effects to compute per-base observation probabilities that are converted into quality scores.