Comprehensive Error Profiling of NovaSeq6000, NovaSeqX, and Salus Pro Using Overlapping Paired-End Reads

Accurate base calling is essential for ensuring the reliability of next-generation sequencing (NGS) data. However, current approaches to error profiling often rely on reference genomes or variant-calling pipelines, which can introduce platform-specific biases or overlook systematic sequencing errors. Here, we present ErrorProfiler, a reference-free, paired-end overlap-based workflow for quantifying and characterizing base-call errors. We applied this approach to profile three high-throughput sequencing platforms - NovaSeq6000, NovaSeqX, and Salus Pro - using Escherichia coli datasets. By analyzing mismatches within fully overlapping 150 bp read regions, ErrorProfiler effectively distinguished sequencing-specific errors from PCR or library preparation artifacts.

NovaSeqX exhibited the lowest median sequencing error rate (0.089‰), with minimal cycle-dependent drift. Salus PRO followed (0.124‰), outperforming NovaSeq6000 (0.185‰). Base quality score calibration analysis revealed systematic underestimation of base-call accuracy across all platforms. Illumina’s binned Q-score models (Q11/Q25/Q37; Q12/Q24/Q40) masked substantial intra-bin variability, while Salus Pro’s continuous Q-score system offered finer granularity but still overestimated error probabilities in key bins.

A previously unreported “upstream effect” was observed, where the empirical accuracy of a base-call was significantly influenced by the Q-score of its preceding base. This effect was most pronounced in high-Q bins and consistent across all platforms. Substitution error spectra revealed distinct platform-specific and sequence-context-dependent biases, driven by optical systems and detection chemistry.

Our findings provide a rigorous, platform-agnostic framework for sequencing error analysis and quality assessment. ErrorProfiler is open-source and compatible with diverse sequencing platforms, offering a valuable tool for improving calibration, variant calling, and downstream bioinformatics workflows.