A reminder: QUAST only works when the assembly in evaluation and the reference genome are derived from the same sample; otherwise it reports truth structural variations as assembly errors and greatly penalizes better assemblies.

Try it. It saves time.

I am looking for one or two postdocs or research associates to join my group. We are currently working on sequence alignment, phased de novo assembly, pangenome analysis and a few smaller projects. DM or email me if interested. Thank you. hlilab.github.io/vacancies

A few frequently asked questions about hifiasm assembly of large genomes (a thread):

Minimap2 v2.22 released with a short preprint associated with it, more formally describing recent improvements since v2.19. Thank Chirag Jain, Arang Rhie and all users for their invaluable feedback. arxiv.org/abs/2108.03515

Impressive performance of SLOW5 on I/O! I noticed fast5 loading was extremely slow when I developed and tested the tools. ONT should really consider changing their file format at some time point.

Preprint on the recent Hi-C assembly module in hifiasm that can produce a pair of haplotype-resolved assemblies for a diploid individual without mixing parental haplotypes in each contig. One executable, one command line for the entire assembly step.

Our lab Heng Li developed a metagenome assembler, hifiasm-meta (arxiv.org/abs/2110.08457), which reconstructs PacBio Hifi libraries to tens to hundreds of complete circular bacterial genomes. We hope these high-quality contigs to offer a new means to study microbial communities.

Thanks for this great opportunity to share my works and also learn more about the recent progress of others at #GenomeInformatics

Our newest work Chromap for fast chromatin profiling analysis is published on Nature Communications today! Chromap significantly reduces the pain of waiting in the era of deep sequencing and single-cell analysis. Haowen Zhang Xiaole Shirley Liu Heng Li nature.com/articles/s4146…

Kar-Tong Tan found Oxford Nanopore basecallers often systematically call telomere repeats (TTAGGG)n into other repeats with similar signal profiles. He tuned the bonito basecaller and developed a pipeline to greatly reduce such errors. Read the thread and our preprint for details.

Table 2 of arxiv.org/pdf/2007.01547… shows what's wrong with ML research. Papers got in by providing a theorem (checked by reviewers) and "significant improvement" (not checked). Significant improvement disappeared when tested by third party

A proof-of-concept implementation of the graph wavefront alignment algorithm (GWFA) for aligning a sequence to a sequence graph. Only support edit distance and no traceback. Fast at small edit distance. github.com/lh3/gwfa

Most long-read assembly tools assume that the string graph model [Myers 2005] encodes every valid assembly, but this is unfortunately false. Excited to share new results on how to address this.

The previous hifiasm paper described haplotype-resolved assembly with trio data. In this new paper, Haoyu Cheng achieves chromosome-long phased assembly without parents. Tested for human and non-human species. A single executable and one command line for the entire process.

Xiaowen will give a webinar about HiFi metagenome assembly with hifiasm-meta tomorrow at 1pm ET. Preprint available at arxiv.org/abs/2110.08457. Most numbers updated but the conclusion stays the same.

1/3 We released Chromap v0.2.3 with the ultimate flexible barcode specification (hopefully) Haowen Zhang ! It now supports the complex non-consecutive barcode specification, a very useful feature for SPLiT-seq data, e.g.: github.com/haowenz/chroma…

Our next big thing! 🎉 "Verkko: telomere-to-telomere assembly of diploid chromosomes" from Mikko Rautiainen Sergey Nurk 🇺🇦 Sergey Koren generates beautiful graphs and automatically assembles 20/46 human chromosomes from HG002 #T2T without gaps biorxiv.org/content/10.110…

Two openings from my group: a) postdoc either on developing methods for sequencing and pangenomic data, OR on the analysis of these data types. b) staff scientist on high-performance algorithm development. For details, see hlilab.github.io/vacancies