Element AVITI™ system delivers the most accurate short-read calls
In genomics, accuracy matters—especially when it comes to identifying genetic variants that impact health, ancestry, or research outcomes. That’s why we’re excited about a recent preprint, Pangenome-aware DeepVariant, from researchers at Google and the UC Santa Cruz Genomics Institute that combines a pangenome reference with advanced deep learning for improvements in variant calling accuracy across multiple platforms. Among them, AVITI sequencing consistently outperformed other technologies, reinforcing its reputation for accuracy and innovation.
Better variant calling with the power of pangenomes
Reference genomes provide a blueprint for read alignment, variant calling, annotation, comparative genomics, and allow us to systematically interpret sequencing data. However, basing variant calling pipelines on a single, linear reference genome introduces bias, especially in regions with high diversity or structural complexity, which can skew downstream data analysis.
To overcome these biases, the Human Pangenome Reference Consortium assembled a pangenome graph to improve read mapping and variant calling. Using this dataset as a foundation, in this study, the authors introduce and assess pangenome-aware DeepVariant, a variant caller using advanced deep learning to identify candidate variant sites and distinguish true variant signals from noise. Overall, the pangenome aware DeepVariant reduced errors and achieved over 20% more accurate variant calling compared to existing methods.
Importantly, this method is generalizable and supports reads from multiple sequencing technologies—and that’s where our AVITI sequencing stands out.
AVITI sequencing data sets the accuracy benchmark
The study evaluated performance of Illumina and Element reads using the benchmark truth set, T2T-Q100-v1.1. Across every mapper and variant caller tested on this truth set, Element AVITI outperformed Illumina in generating accurate variant call sets. The most accurate short-read call set achieved 99.65% precision and 99.1% recall when using AVITI data with vg giraffe mapping and pangenome-aware DeepVariant calling.

A powerful validation of our technology, this echoes what we’ve shown before: our avidite base chemistry delivers improved accuracy, especially in difficult genomics regions like homopolymers and short tandem repeats.
The paper also calls attention to how the choice of benchmark matters. The Platinum truth set, built mostly on Illumina data, masked some advantages of AVITI, whereas the T2T-Q100 benchmark highlights AVITI’s accuracy.
Future-proof variant calling for precision genomics
Regardless of input data, pangenome-aware DeepVariant consistently reduced errors, especially in segmental duplications where short reads are prone to mismapping and is designed to keep pace with ever-growing reference datasets.
As the field scales up to richer pangenome references including personalized pangenomes and future HPRC references containing up to 1,000 haplotypes, it’s clear that high-quality sequencing from platforms like the AVITI will play a key role in accurate variant calling.
Final takeaway? This study reaffirms what we’ve believed from the start—high-accuracy sequencing with Element isn’t just better. It’s leading the way.