Harnessing the Kivvi tool to resolve the genetics of FSHD with HiFi sequencing

 

For families affected by facioscapulohumeral muscular dystrophy (FSHD), one of the hardest parts of the condition is not knowing what comes next. FSHD is one of the more common inherited muscular dystrophies in adults, and it is characterized by progressive weakness in the muscles of the face, shoulders, and upper arms. Yet people carrying similar genetic changes can have very different experiences. One may never notice symptoms, while another may gradually lose strength until a wheelchair becomes necessary. That unpredictability makes planning difficult for families and has left researchers asking why the disease varies so widely.

At the molecular level, FSHD is caused by chromatin relaxation and the abnormal expression of a gene called DUX4 in skeletal muscle. Resolving that region has typically required several separate assays.

In a recent preprint, scientists from PacBio, in collaboration with GeneDx and Leiden University Medical Center, presented the use of a new computational tool, Kivvi, to help address that challenge. The tool, released on GitHub, characterizes FSHD repeats from HiFi long-read sequencing data. Up against traditional methods in the study, Kivvi successfully identified every affected allele, demonstrating how a single HiFi workflow could replace multiple assays while supporting population-scale studies aimed at uncovering new genetic modifiers.

For a collection of impactful publications like this one that may help in the grant application support process, explore our grant support resources.

Get your grant support resources

The genetic basis of FSHD and existing testing technology

In contrast to repeat expansion disorders like Huntington disease where a repeated region of DNA can expand considerably past a safe threshold¹, most cases of FSHD are actually driven by the contraction of the repeat region D4Z4 that holds the DUX4 gene. Whereas unaffected individuals typically carry 11 to 100 copies of D4Z4 (each at 3.3 kb), individuals with FSHD Type 1 instead have between 1 and 10 copies.

In these instances amounting to 95% of FSHD cases, the contraction of this region loosens the chromatin that keeps DUX4 switched off. Alternatively, FSHD2 is typically associated with variants in chromatin regulators like SMCHD1 which reduce methylation across D4Z4 more broadly. Because D4Z4 is present on two different chromosomes (4 and 10) and various alleles can behave differently, characterizing FSHD depends on knowing repeat size, chromosome of origin, haplotype, and methylation together.

Traditional workflows only capture a part of that picture at one time. Southern blotting with pulsed-field gel electrophoresis (PFGE) has long been used to size D4Z4, but is labor-intensive and low-throughput. Optical genome mapping (OGM) has more recently been applied as an alternative, though its DNA extraction is highly time sensitive and it is often limited to blood samples. Labs may also need to wait for enough samples to run a batch. Because OGM cannot detect methylation or sequence-level changes, it is usually paired with bisulfite sequencing and separate sequencing of FSHD2-associated genes.

Table 1. Summary of methods used to characterize the genetics of FSHD, comparing traditional assays with HiFi WGS and Kivvi.

 

Long-read sequencing with Kivvi

In contrast to these traditional methods, long-read sequencing can capture repeat size, sequence, and methylation in a single assay. Starting from HiFi WGS reads, the Kivvi tool gathers the reads that overlap D4Z4, distinguishing individual repeat units by their subtle sequence differences, and assembles complete alleles by stitching together the consecutive units that each highly accurate HiFi read spans.

For every D4Z4 allele, Kivvi reports:

For every D4Z4 allele, Kivvi reports:

• Repeat size, the core FSHD1 measurement
• Chromosome of origin, separating pathogenic 4q alleles from homologous 10q alleles
• Allele type, including whether the 4qA polyadenylation signal is intact or disrupted
• Methylation level, read natively from 5mC in the same HiFi reads, with no bisulfite conversion

To assess performance, the team compared Kivvi against the same traditional methods described above, using Southern blot results for 27 control samples from the 1000 Genomes Project and OGM results for 10 FSHD1 samples characterized at GeneDx. Across both datasets, Kivvi detected all 11 contracted alleles and correctly classified 90% of noncontracted alleles.

Table 2. Table 2. Summary of Kivvi results benchmarked against Southern blot and optical genome mapping, as reported in Chen et al. (2026).

Methylation added another layer of insight. Alongside the FSHD1 samples, the team analyzed six additional samples carrying SMCHD1 variants associated with FSHD2, which were identified from the same HiFi WGS data using DeepVariant. In FSHD1 samples, reduced methylation was confined to the contracted allele, while samples carrying SMCHD1 pathogenic variants associated with FSHD2 showed lower methylation across nearly all D4Z4 alleles. This signature is reported by Kivvi and can help distinguish FSHD1 and FSHD2 subjects.

These results were generated from blood samples, but HiFi WGS data can also be generated from saliva, supported by validated protocols for extracting high molecular weight DNA from saliva and HiFi WGS data from saliva samples sequenced at roughly 30-fold coverage.

 

What Kivvi means for you

Clinical and reference labs. Replace a multi-assay FSHD workflow with standard HiFi WGS and an open-source analysis tool. One run generates repeat size, haplotype, methylation, and FSHD gene variants, with the ability to use the same data to answer other rare-disease questions.

Researchers and rare disease organizations. Resolve subjects who are clinically diagnosed but genetically unsolved, and study borderline alleles with methylation in hand. Population-scale data makes it possible to search for the genetic modifiers that explain why FSHD varies so much within families.

Pharma and biotech. Define trial cohorts more precisely by repeat size, haplotype, sequence variation and methylation from one dataset. Methylation is a candidate stratification marker, and a single workflow simplifies testing across trial sites.

 

Population-scale sequencing and the future of FSHD research

Kivvi was also used to profile D4Z4 in 601 individuals from five ancestral populations, revealing common haplotypes, hybrid repeat units, and alleles carrying signs of translocation events, with greater diversity observed in individuals of African ancestry. Complete sequences at this scale lay the groundwork for studies that could help identify genetic modifiers and, in turn, shed light on why FSHD affects family members so differently.

This more complete molecular view could also support pharma and biotech drug development in FSHD. By resolving D4Z4 repeat structure, haplotype, sequence variation, and methylation from a single HiFi dataset, Kivvi could help researchers better characterize FSHD populations, identify potential disease modifiers, and define more precise cohorts for therapeutic research and clinical trials.

FSHD2 is linked to variants in SMCHD1, DNMT3B and LRIF1⁴. LRIF1 and has also been catalogued as a distinct digenic form, FSHD3, and the classification is still evolving^5,6. These patients typically have normal-sized repeats, so size-based tests can often miss them. Because HiFi WGS captures both D4Z4 methylation and these genes in one dataset, labs can investigate FSHD2 and FSHD3 without reflex testing or a second sample.

The team is also extending Kivvi to other clinically relevant repeats with similarly large repeat units. Kivvi already includes support for the Kringle IV-type 2 (KIV-2) repeat in the LPA gene, where copy number is associated with cardiovascular disease risk⁴.

Together, these results point toward a future in which repeat size, haplotype, methylation, and sequence variants can be assessed from a single HiFi WGS dataset, simplifying FSHD research and opening new paths into a remarkably complex region of the genome.

---

To learn about other computational tools like Kivvi, visit the PacBio computational page and explore the grant support resources.

Get your grant support resources

 

References

1. Chen, Z., Morris, H. R., Polke, J., Wood, N. W., Gandhi, S., Ryten, M., Houlden, H., & Tucci, A. (2025). Repeat expansion disorders. Practical Neurology, 25(3), 204-216. https://pn.bmj.com/content/25/3/204
2. Xie, Q., Ma, G., & Song, Y. (2024). Therapeutic strategy and clinical path of facioscapulohumeral muscular dystrophy: review of the current literature. Applied Sciences, 14(18), 8222. https://doi.org/10.3390/app14188222
3. Chen, X., et al. (2026). Resolution of the D4Z4 repeat responsible for facioscapulohumeral muscular dystrophy with HiFi sequencing. bioRxiv, 2026-04. https://doi.org/10.64898/2026.04.10.717730
4. Facioscapulohumeral muscular dystrophy type 2: an update on the clinical, genetic, and molecular findings. (2021). Neuromuscular Disorders. https://doi.org/10.1016/j.nmd.2021.09.010
5. OMIM #619477. Facioscapulohumeral muscular dystrophy 3, digenic; FSHD3. https://omim.org/entry/619477
6. Hamanaka, K., et al. (2020). Homozygous nonsense variant in LRIF1 associated with facioscapulohumeral muscular dystrophy. Neurology. https://doi.org/10.1212/WNL.0000000000009617
7. Clarke, R., et al. (2009). Genetic variants associated with Lp(a) lipoprotein level and coronary disease. New England Journal of Medicine, 361(26), 2518-2528. https://www.nejm.org/doi/full/10.1056/NEJMoa0902604