Research use only. Not for use in diagnostic procedures. © 2025 Pacific Biosciences of California, Inc. (“PacBio”). All rights reserved. Information in this document is subject to change without notice. PacBio assumes no responsibility for any errors or omissions in this document. Certain notices, terms, conditions and/or use restrictions may pertain to your use of PacBio products and/or third-party products. Refer to the applicable PacBio terms and conditions of sale and to the applicable license terms at pacb.com/license. Pacific Biosciences, the PacBio logo, PacBio, Circulomics, Omniome, SMRT, SMRTbell, Iso-Seq, Sequel, Nanobind, SBB, Revio, Onso, Apton, Kinnex, PureTarget, SPRQ, and Vega are trademarks of PacBio. Portello: Making global assembly more effective for rare-disease WGS Christopher T. Saunders, Zev Kronenberg, James M. Holt, William J. Rowell, Michael A. Eberle PacBio, 1305 O’Brien Drive, Menlo Park, CA 94025 Global assembly of HiFi long reads is a powerful approach to human rare-disease analysis, especially for rare variants that may be highly diverged from reference assemblies. However, integrating global assembly into a practical rare-disease analysis workflow faces significant challenges relative to traditional read mapping approaches: •Small variant accuracy •Somatic variation •Methylation •Assessment of assembly error/compression •Annotation Portello provides a way to reduce these challenges, by taking alignments of long-reads to their own assembly contigs and transferring them to an annotated reference assembly. Partitioning germline variation and sequencing error into two separate alignment problems, improves alignment consistency and interpretation. •Identical HiFi read set processed through pbmm2 mapping and verkko assembly with portello GRCh38 mapping. •Small variant calls from DeepVariant41.9, using standard PacBio model. No retraining for portello. •Assessment on Platinum Pedigree small variant benchmark5 v1.2 by aardvark6(similar trends on HG002/T2T benchmark) Summary Portello improves read alignment Portello improves small variant calling •Documentation, binary releases (Linux x64), and source code for portello are available on GitHub: https://github.com/PacificBiosciences/portello •Portello is still under development. For questions or new use-case requests please reach out to us (
[email protected]). Availability / Contact latest release References FN FP Recall Precision F1 pbmm2 34684 34445 0.992 0.992 0.992 portello 21232 29682 0.995 0.993 0.994 Read mapping with portello leads to a 39% reduction in small variant false negatives and 14% reduction in false positives. Workflow Overview 1. Global assembly of sample HiFi reads - Either partially-phased/dual assembly or fully-phased assembly via methods like trio binning - Tested with both hifiasm1 and verkko2 2. Map reads back to assembly contigs (pbmm2) 3. Map contigs to reference (minimap2) 4. Portello transfers read mappings from assembly to reference assembly contig1 reads mapped to assembly contig1 assembly contig2 reads mapped to assembly contig2 … reference assembly contig1 assembly contig2 reference reads mapped to assembly contig1 reads mapped to assembly contig2 Portello •Final alignment output unifies long-range global assembly information with read-level detail on somatic variation, methylation, basecall quality, etc. •Standard read-based variant callers can process this mapping output without any special modifications. portello reads grouped by their assembly contig portello pbmm2 •HG002 read alignments to VNTR region on GRCh38 •The same reads are input to the portello assembly/mapping process (top) and the pbmm2 conventional mapping process (bottom) •Portello reads are tagged by assembly contig segment, enabling read grouping by haplotype. Reduced read mapping artifact portello pbmm2 poorly mapped reads from conventional mapping can be assembled to contigs acting as samplespecific decoys in portello mapping instead •Example shows HG002 read alignments on GRCh38 •The same reads are input to the portello assembly/mapping process (top) and the pbmm2 conventional mapping process (bottom) Clarified CNVs in clinically-relevant samples Microarray CNV call (Gross et al) pbmm2 SegDup annotation portello IGV views of NA21886, a clinical CNV sample from Gross 20193, The same HiFi read data is mapped with pbmm2 (above) and portello (below). Both views show an identical chr22 span containing a verified CN3 duplication from the Gross study. Copy number inference and interpretation are greatly simplified in the portello-mapped view of this locus. Microarray CNV call (Gross et al) SegDup annotation Unified contig and read-based inference An additional complication of applying global assembly to raredisease WGS is unification of results with those from standard read-mapping. Sample HiFi WGS reads assembler Fasta/ graph portello portello GRCh38 BAM New hybrid contig/read tools (portello-CNV) Consistent sample state inference Current readbased variant tools (DV) Current contigbased variant tools (PAV) For instance, contigbased SV-calling tools typically provide superior results to methods based on mapped-reads7,8 (e.g. on GIAB T2T benchmark, shown right), but can be difficult to synchronize with read-based results. Portello enables workflows (below) which can synchronize results from existing contig and read-based callers, while allowing for the development of new joint contig/read methods, such as for CNV. GIAB T2T SV benchmark assessment 1. Cheng, Haoyu et al. “Haplotype-resolved de novo assembly using phased assembly graphs with hifiasm.” Nature methods vol. 18,2 (2021) 2. Rautiainen, Mikko et al. “Telomere-to-telomere assembly of diploid chromosomes with Verkko.” Nature biotechnology vol. 41,10 (2023) 3. Gross, Andrew M. et al. Copy-number variants in clinical genome sequencing: deployment and interpretation for rare and undiagnosed disease, Genetics in Medicine, vol. 21,5 (2019) 4. Poplin, Ryan et al. “A universal SNP and small-indel variant caller using deep neural networks.” Nature biotechnology vol. 36,10 (2018) 5. Kronenberg, Zev et al. “The Platinum Pedigree: a long-read benchmark for genetic variants.” Nature methods vol. 22,8 (2025) 6. Holt, James M. et al. “Aardvark: Sifting through differences in a mound of variants.” bioRxiv 2025.10.03.680257 7. Saunders, Christopher T et al. “Sawfish: improving long-read structural variant discovery and genotyping with local haplotype modeling.” Bioinformatics (Oxford, England) vol. 41,4 (2025) 8. Ebert, Peter et al. “Haplotype-resolved diverse human genomes and integrated analysis of structural variation.” Science (New York, N.Y.) vol. 372,6537 (2021) Copy number can be accurately inferred from portello alignments using both assembly contig depth and read coverage over each contig
