SEQdata-BEACON

SEQdata-BEACON aggregates and analyzes lane-level sequencing performance data from BGISEQ-500 (DNBSEQ) runs to evaluate and predict run yield and quality.

Key Features:

• Data Collection and Organization: Aggregates 2236 lane-level entries from 60 BGISEQ-500 instruments at BGI-Wuhan (November 2018–April 2019), each with a unique lane ID and 65 metrics covering sample characteristics, yield, quality, machine state, and supplies.
• Metric Clustering: Uses a correlation matrix to cluster 52 numerical metrics into three groups—yield-quality, machine state, and sequencing calibration—to reveal metric relationships.
• Yield Simulation Model: Implements a linear regression model on cycle-by-cycle data (up to 200 cycles) to predict final yield, achieving R² = 0.81 at cycle 15 and R² = 0.97 at cycle 200, with external validation on a May 2019 test set yielding R² = 0.96.

Scientific Applications:

• Process Stabilization: Inform adjustments to BGISEQ-500 run parameters and supplies to stabilize sequencing performance.
• Performance Prediction and Optimization: Enable early prediction of final yield to guide run decisions and parameter optimization using the regression model.
• Troubleshooting and Interpretation: Identify anomalous runs and interpret DNBSEQ metrics through clustering and correlation analyses.

Methodology:

Systematic aggregation of lane-level metrics from BGISEQ-500 runs, construction of a correlation matrix to cluster 52 numerical metrics, and development of a linear regression model trained on cycle-by-cycle data (up to 200 cycles) with validation on a May 2019 test set.