Evaluating genome assemblies with HMM-Flagger.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Genome assembly is crucial for understanding genetic variation and disease.
• Detecting structural errors in haplotype-resolved assemblies remains a challenge.
• Reference-free methods are needed to assess assembly accuracy without a gold standard.

Purpose of the Study:

• To develop and validate HMM-Flagger, a novel reference-free tool for detecting structural errors in genome assemblies.
• To assess the performance of HMM-Flagger using both simulated and real sequencing data.
• To evaluate improvements in genome assembly quality over time using HMM-Flagger.

Main Methods:

• HMM-Flagger models read coverage using a hidden Markov model (HMM) with a Gaussian autoregressive process.
• The tool classifies coverage anomalies into erroneous blocks, false duplications, or collapsed blocks.
• Performance was evaluated using synthetic misassemblies and real data from Pacific Biosciences HiFi and Oxford Nanopore Technologies.

Main Results:

• HMM-Flagger achieved F1 scores of 78.4% and 60.4% for synthetic errors with PacBio HiFi and Oxford Nanopore data, respectively.
• The tool identified large misassemblies, including false duplications and collapse events, in human satellite regions of HG002 assemblies.
• Application to Human Pangenome Reference Consortium (HPRC) assemblies showed a significant reduction in error rates from 0.94% (release 1) to 0.38% (release 2).

Conclusions:

• HMM-Flagger is an effective reference-free tool for identifying structural errors in haplotype-resolved genome assemblies.
• The tool's performance highlights advancements in long-read sequencing technologies and assembly pipelines.
• HMM-Flagger successfully validated complex genomic regions, such as NOTCH2NL assemblies, demonstrating its utility in assessing high-quality genome references.