iiinova — where your digital self keeps evolving

Chris Saunders

@ctsa.bsky.social

Rare disease and cancer analysis models for sequencing data. Art school survivor. Views my own.

Chris Saunders @ctsa.bsky.social · 20/02/2026

We also demonstrate the potential for improved CNV inference in a segmental duplication region where assembly-based read mapping substantially clarifies copy number inference compared to a conventional mapping approach. (4/5)

110

Chris Saunders @ctsa.bsky.social · 20/02/2026

We demonstrate that variant calling shows improved accuracy given these assembly-based read mappings. For instance, given the same set of input reads, DeepVariant calling from portello-mapped reads shows a 47% reduction in errors compared to calls from a conventional mapping process. (3/5)

Small variant basepair errors comparison between conventionally mapped reads and portello-mapped reads.

111

Chris Saunders @ctsa.bsky.social · 20/02/2026

Portello facilitates an assembly-based approach to read mapping, wherein reads mapped to their own diploid assembly contigs are transferred back to a reference genome. This approach allows standard secondary analysis tools and viewers to operate on reads mapped in the context of the diploid assembly

120

Chris Saunders @ctsa.bsky.social · 20/02/2026

What if you could improve small variant accuracy, CNV inference, and interpretability of your HiFi WGS data by taking a different approach to read mapping? Our new preprint describes portello, a method which demonstrates the potential for such improvements. (1/5)

Comparison of read mappings at HG002 chr4:40,294,825-40,295,700, showing conventional (pbmm2) read mappings (above) and portello mappings (below). The same set of unaligned input reads were input into each mapping process.

12511

Chris Saunders @ctsa.bsky.social · 05/10/2025

In Stockholm for work and got a great recommendation to Klättercentret Telefonplan for a boulder break last night. Great gym all around and easy hop on the metro from city center.

000

Chris Saunders @ctsa.bsky.social · 20/05/2025

Just released a major update to the sawfish SV caller, which adds CNV calling and integration. Assessment on a set of pathogenic CNVs from Gross et al. shows this integrated-call strategy can substantially improve (single-method) recall, especially at lower HiFi sequencing depths

161

Chris Saunders @ctsa.bsky.social · 10/04/2025

To assess multi-sample calling, we examined genotype pedigree concordance across SV calls from the 2nd and 3rd generations of CEPH-1463, finding that sawfish has over 9000 more pedigree-concordant calls compared to other methods while retaining the highest proportion of concordant SVs:

120

Chris Saunders @ctsa.bsky.social · 10/04/2025

We assessed sawfish on the recent draft GIAB SV benchmark from the T2T-HG002-Q100 assembly, finding that it has the highest F1-score among evaluated callers on every tested SV size range and depth level. Other callers required at least 30x depth to reach sawfish’s accuracy at 15x:

110