Mao-Jan Lin @maojanlin.bsky.social · 03/08/20267/ Huge thanks to my advisor @benlangmead.bsky.social for the guidance and support, and my collaborator @vikramshivakumar.bsky.social! 020
Mao-Jan Lin @maojanlin.bsky.social · 03/08/20266/ Although runtime scales linearly with the number of haplotypes, ImpuT2T is highly parallelizable: - Runs in ~15 hours (48 threads) on the full 470-haplotype pangenome. - If your compute node can handle de novo assembly, it can run ImpuT2T with HPRC2. 120
Mao-Jan Lin @maojanlin.bsky.social · 03/08/20265/ Adding more haplotypes consistently improves patched assembly quality, with top performance using the full HPRC Release 2 (470 haplotypes). While population-matched pangenomes achieve better result with lower number of haplotypes, the full pangenome still achieves the best overall results. 120
Mao-Jan Lin @maojanlin.bsky.social · 03/08/20264/ Where do the patches come from? Across all evaluated datasets, the plurality of selected patches match the sample's ancestral superpopulation: - EAS for CN1 & HG005 - AFR for PAN027 - EUR for HG002 120
Mao-Jan Lin @maojanlin.bsky.social · 03/08/20263/ We benchmarked 5 test sets with scaffolding (RagTag), single-reference (GPatch, ImpuT2T-1), and HPRC2 pangenome-based (ImpuT2T-470) patching. Target: High coverage (X) + Low errors (Y) → Bottom-right is ideal. Pangenome patching recovers the most coverage while introducing the fewest errors. 120
Mao-Jan Lin @maojanlin.bsky.social · 03/08/20262/ How does ImpuT2T work? A. Scaffolding: A logistic regression model predicts connection likelihoods between contig pairs to resolve ordering layout. B. Patching: The optimal reference sequence is selected by evaluating sequence identity between candidate references and contig flanking regions. 130
Mao-Jan Lin @maojanlin.bsky.social · 03/08/20261/ Excited to share our preprint ImpuT2T: a pangenome-based assembly patching tool ImpuT2T scaffolds draft human assemblies with a pangenome (HPRC2) and patches gaps by leveraging linkage disequilibrium and sequence identity between contigs and reference haplotypes Code: github.com/maojanlin/Im...biorxiv.orgImpuT2T: Pangenome-Based Patching for Human Genome AssembliesWith improvements in sequencing and assembly have come many high-quality telomere-to-telomere assemblies and reference pangenomes. However, the long-read sequencing recipes needed for high quality ass... 1229
Reposted by Mao-Jan LinVikram Shivakumar @vikramshivakumar.bsky.social · 15/07/20261/ Excited to share the newest tool in the pangenome MUMiverse: Shredtools! Shredtools enables a user to navigate the pangenome coordinate system with multi-MUMs. More in the thread🧵 Code: github.com/vikshiv/shredtools Interactive tool for querying HPRC assemblies: vikshiv.github.io/shredtoolsbiorxiv.orgNavigating the pangenome coordinate system with ShredtoolsExisting notions of pangenome coordinates rely on hard-to-compute multiple sequence alignments. On the other hand, pangenome-wide exact unique matches (multi-MUMs) can be computed efficiently, and rep... 13822
Reposted by Mao-Jan LinAdam Phillippy @aphillippy.bsky.social · 02/02/2026Time for a thread on our Christmas preprint “Origin and evolution of acrocentric chromosomes in human and great apes”. I had so much fun with this project and paper. It will be hard to summarize in a thread, but I’ll try www.biorxiv.org/content/10.6... [1/21] 14229
Reposted by Mao-Jan LinVikram Shivakumar @vikramshivakumar.bsky.social · 06/01/2025Excited to share a preprint for (w/ @benlangmead.bsky.social) our new tool, Mumemto, on biorxiv! Mumemto finds multi-MUMs across pangenomes (i.e. mummer but for pangenomes). It can rapidly visualize synteny, identify misassemblies, and accelerate core genome and multiple alignment, highlighting SVs.biorxiv.orgMumemto: efficient maximal matching across pangenomesAligning genomes into common coordinates is central to pangenome analysis and construction, but it is also computationally expensive. Multi-sequence maximal unique matches (multi-MUMs) are guideposts ... 13429