The GIAB genomic stratifications resource for human reference genomes.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• No single sequencing or variant-calling workflow excels across the entire human genome.
• Understanding performance variations in different genomic contexts is crucial for optimizing sequencing strategies.
• Newer reference genomes, like T2T-CHM13, include more complex regions impacting tool performance.

Purpose of the Study:

• To introduce a set of genomic stratifications (BED files) for evaluating sequencing and variant-calling tool performance.
• To define these stratifications for established (GRCh37/38) and novel (T2T-CHM13) human genome references.
• To analyze performance differences and challenges introduced by new regions in the T2T-CHM13 reference.

Main Methods:

• Development of BED files defining distinct genomic contexts (stratifications).
• Definition of stratifications for GRCh37/38 and T2T-CHM13 human genome references.
• Benchmarking of sequencing performance across different references and platform iterations.

Main Results:

• Identification of increased hard-to-map and GC-rich regions in the T2T-CHM13 reference.
• Quantification of a performance penalty associated with these difficult regions in CHM13.
• Demonstration of stratifications' utility in tracking platform-specific performance improvements (e.g., Oxford Nanopore Technologies).

Conclusions:

• Genomic stratifications provide critical context for assessing sequencing tool performance.
• The T2T-CHM13 reference presents new challenges due to expanded difficult-to-sequence regions.
• These stratifications enable informed decisions for building robust sequencing pipelines and calculating risk-reward tradeoffs.