Complete Genomics Reports 73% Fewer Genome Analysis Errors with AI Models

Complete Genomics announced in a press release that AI trained variant calling models reduced genome analysis errors by up to 73 percent compared to standard pipelines when applied to DNBSEQ sequencing data. The study used data from the company’s T1+, T7, and T7+ platforms.

The research showed that integrating AI models trained on DNBSEQ data with high quality sequencing improved accuracy and consistency across the genome, including in complex and medically relevant regions. The models achieved high accuracy for single nucleotide variants and small insertions or deletions.

According to the company, combining its PCR free library preparation and clonal error free sequencing with AI analysis enables more accurate detection of genetic variation. The study also found that performance in difficult genomic regions approached that of long read sequencing technologies.

The AI framework used in the study, called PanVariants, is being released as an open source resource to support further development in genomic analysis. Complete Genomics will host a webinar on April 30 to discuss the results and applications of AI in scalable variant calling.