Finding long tandem repeats in long noisy reads

A comparison of Oxford nanopore library strategies for bacterial genomics

BACKGROUND

Oxford nanopore Technologies (ONT) provides three main library preparation strategies to sequence bacterial genomes. These include tagmentation (TAG), ligation (LIG) and amplification (PCR). Despite ONT's recommendations, making an informed decision for preparation choice remains difficult without a side-by-side comparison. Here, we sequenced 12 bacterial strains to examine the overall output of these strategies, including sequencing noise, barcoding efficiency and assembly quality based on mapping to curated genomes established herein.

RESULTS

Average read length ranged closely for TAG and LIG (> 5,000 bp), while being drastically smaller for PCR (< 1,100 bp). LIG produced the largest output with 33.62 Gbp vs. 11.72 Gbp for TAG and 4.79 Gbp for PCR. PCR produced the most sequencing noise with only 22.7% of reads mappable to the curated genomes, vs. 92.9% for LIG and 87.3% for TAG. Output per channel was most homogenous in LIG and most variable in PCR, while intermediate in TAG. Artifactual tandem content was most abundant in PCR (22.5%) and least in LIG and TAG (0.9% and 2.2%). Basecalling and demultiplexing of barcoded libraries resulted in ~ 20% data loss as unclassified reads and 1.5% read leakage.

CONCLUSION

The output of LIG was best (low noise, high read numbers of long lengths), intermediate in TAG (some noise, moderate read numbers of long lengths) and less desirable in PCR (high noise, high read numbers of short lengths). Overall, users should not accept assembly results at face value without careful replicon verification, including the detection of plasmids assembled from leaked reads.

A landscape of complex tandem repeats within individual human genomes

Markedly expanded tandem repeats (TRs) have been correlated with ~60 diseases. TR diversity has been considered a clue toward understanding missing heritability. However, haplotype-resolved long TRs remain mostly hidden or blacked out because their complex structures (TRs composed of various units and minisatellites containing >10-bp units) make them difficult to determine accurately with existing methods. Here, using a high-precision algorithm to determine complex TR structures from long, accurate reads of PacBio HiFi, an investigation of 270 Japanese control samples yields several genome-wide findings. Approximately 322,000 TRs are difficult to impute from the surrounding single-nucleotide variants. Greater genetic divergence of TR loci is significantly correlated with more events of younger replication slippage. Complex TRs are more abundant than single-unit TRs, and a tendency for complex TRs to consist of <10-bp units and single-unit TRs to be minisatellites is statistically significant at loci with ≥500-bp TRs. Of note, 8909 loci with extended TRs (>100b longer than the mode) contain several known disease-associated TRs and are considered candidates for association with disorders. Overall, complex TRs and minisatellites are found to be abundant and diverse, even in genetically small Japanese populations, yielding insights into the landscape of long TRs.