Expanding the reach of the pangenome era with COSIGT

Researchers at Human Technopole have developed a novel pangenome-based pipeline that captures structural genomic diversity from low-coverage sequencing data at the population level. Their innovative work, including a step-by-step protocol, is published in Genome Biology.

Imagine reading a thick book with colourful pages, each filled with different combinations of words in a four-letter code: that’s your genome. Every person owns a similar volume, but a more careful look at the pages reveals that each colour, while maintaining the same order, can come in a distinct shade. For example, a green page may exist in aquamarine in one book or in emerald in another one, a blue page in turquoise or indigo, and so on. If we just listed the main colours and forgot about the different shades, we could find a common plot to all our stories, but we would not capture the diversity of the characters involved.

This lack of diversity was a limitation of the first human reference genome released more than two decades ago. This achievement undoubtedly revolutionised genomics and pushed the boundaries of our knowledge to a new level. However, as pioneering and groundbreaking as it was, the first reference genome could not fairly represent global diversity. A solution came with the innovative idea of assembling multiple genomes into a single reference, which launched the pangenome era.

This approach made it possible to resolve structurally complex genomic regions, containing multi-kilobase insertions, deletions, duplications, and copy-number variations. Indeed, the release of the current human pangenome reference revealed several previously inaccessible loci.

Despite bringing to light a variety of shades never captured before, a challenge remained: accessing low-coverage sequencing data. The accuracy of pangenome-based strategies still depended on sequencing depth, making it hard, if not impossible, to accurately analyse low-coverage samples. These are often found in population studies, biobank cohorts, and ancient DNA (aDNA), whose limited quality is caused by postmortem DNA degradation and microbial contamination. In other words, no tool could yet separate the shades of faded colours.

To address this issue, the Soranzo Group at Human Technopole developed COSIGT (Cosine Similarity-based GenoTyper), a pangenome-based pipeline for targeted genotyping that can be successfully employed at variable sequencing depths.

Rather than relying on a high number of sequencing reads for a given genomic locus, COSIGT evaluates how they are distributed across the region, maintaining accuracy even on low-coverage samples and outperforming existing pangenome-based tools.

The authors benchmarked COSIGT against the gold-standard pangenome-based genotyping method and proved its superiority at low coverage, even on aDNA. Moreover, they assessed its performance across different sequence lengths, complexity levels, zygosity states, and structural variant types, observing great stability. Finally, they demonstrated its scalability at the population level by successfully applying it to thousands of modern and ancient human genomes.

“Traditional pangenome tools often struggle when sequencing depth drops, leaving valuable datasets like large biobanks or ancient DNA underutilised,” explain Davide Bolognini, lead author of the study, and Nicole Soranzo, Head of the Genomics Research Centre – Population & Medical Genomics at Human Technopole. “With COSIGT, we shifted the focus from counting reads to evaluating coverage patterns across local pangenome graphs. By extending pangenomics to low-coverage data, COSIGT helps democratise access to complex genomic variation and unlocks hidden diversity across both modern population cohorts and ancient samples.“

Previously impossible population-scale biomedical and evolutionary analyses are now within reach. COSIGT extends the benefits of accessing pangenome references to a variety of data, including historical datasets and archaeological samples, taking another step towards highlighting global genomic diversity.

By revealing unprecedented shades of the human genome, the novel COSIGT pipeline promises to be a useful, groundbreaking tool for stable and scalable analysis of sequencing data with heterogeneous depth at reduced experimental costs.

Image: Visual representation of the complex LPA gene locus modeled as a pangenome graph using Graphoscope. The intricate network highlights distinct genomic paths in different colors (such as red, blue, and green), illustrating alternative haplotypes and variations across individuals.

Bolognini, D., Guarracino, A., Paleni, C. et al. COSIGT: population-scalable genotyping of complex loci from low-coverage sequencing data using pangenome graphs. Genome Biol 27, 286 (2026). https://doi.org/10.1186/s13059-026-04242-4