Genomic Variant Calling: Companies, Startups & Research

The facts list 314 companies and 19 startups in genomic variant calling. Illumina, Guardant Health, Natera, Intellia Therapeutics, Personalis and Tempus AI are among the patent holders. Illumina also appears among the leading organizations by recent activity. These companies span sequencing, diagnostics and therapeutics, so their involvement does not necessarily mean they sell standalone variant-calling software.

University of California leads the listed organizations with 576 recent-window documents, followed by Harvard University with 475 and Chinese Academy of Sciences CAS with 414. Other active groups include Fudan University, University of Washington and Stanford University. Their work ranges from short-read calling and caller benchmarking to long-read detection and somatic variant analysis.

The facts record $55,139,495 across 122 grants since 2020 and $59,068,787,094 across 44 funding rounds. These totals reflect funding associated with the topic’s broad organization and document set, not necessarily funding dedicated solely to variant-calling products. The rounds may include large organizations and activities beyond software development, so the totals are best treated as broad indicators.

Illumina leads the listed patent holders with 43 patents in the facts, followed by Guardant Health with 34 and Natera with 17. Intellia Therapeutics has 11, while Personalis has 10. The list also includes Life Technologies and Tempus AI with 8 each, and Foundation Medicine with 7.

No: the facts report 579 patents in the 24-month window and a 53.8% decline versus the previous 24 months. This contrasts with continued publication activity, which was down 2.0% in the last-year comparison. Patent counts alone do not establish whether commercial innovation has slowed; they can also reflect filing timing and the scope of technologies included.

Recent work includes ContextSV for detecting large copy-number variants from long reads, published in September 2026, and AccuSNV, a deep-learning method for bacterial SNV calling, also reported in September 2026. VCboost, reported in August 2026, applies a post-calling filter to reduce false positives in long-read SNP and indel detection. Other work explores nanopore signal information and synthetic benchmarks for somatic calling.

The facts identify 5 approaches: pileup-based calling, local haplotype assembly, deep-learning callers, graph-aware calling and long-read calling. Pileup methods evaluate read evidence at reference positions, while local assembly reconstructs nearby haplotypes. Learning-based callers model patterns in data; graph methods use multiple genome paths, and long-read methods use extended reads to capture complex variation. Each approach has different accuracy, data and compute trade-offs.

The United States leads the country list with 47,213 documents, followed by China with 39,296 and the United Kingdom with 15,993. Germany follows with 9,832. These figures indicate where topic-related activity is associated, but they do not measure research quality or commercial adoption.

The facts identify population genetics, agriculture and animal breeding, pathogen genomics, cancer genomics and rare disease diagnostics as applications. Population genetics has 4,836 documents, while agriculture and animal breeding has 2,443. Recent work also applies calling to bacterial transmission and antimicrobial-resistance research, crop improvement, and the detection of low-frequency variants in disease studies.